diff options
| author | Xiang, Haihao <haihao.xiang@intel.com> | 2010-03-26 12:48:50 +0800 |
|---|---|---|
| committer | Xiang, Haihao <haihao.xiang@intel.com> | 2010-03-26 12:48:50 +0800 |
| commit | 7ea27f59d6cb1f5045ab482f82ec27226d934a71 (patch) | |
| tree | 50ca8b4593335bc2a2277780dd4ecf37bfd0b06e | |
| parent | be803262934445e16e44bb63f2e4ba35b9d22948 (diff) | |
| download | libva-intel-driver-7ea27f59d6cb1f5045ab482f82ec27226d934a71.tar.gz libva-intel-driver-7ea27f59d6cb1f5045ab482f82ec27226d934a71.tar.bz2 libva-intel-driver-7ea27f59d6cb1f5045ab482f82ec27226d934a71.zip | |
i965_drv_video: add support for H264 on Clarkdale/Arrandale
190 files changed, 25606 insertions, 147 deletions
diff --git a/Makefile.am b/Makefile.am index 88b5590..bb160ae 100644 --- a/Makefile.am +++ b/Makefile.am @@ -32,21 +32,29 @@ i965_drv_video_la_LIBADD = ../va/libva-x11.la -lpthread i965_drv_video_la_SOURCES = \ object_heap.c \ intel_batchbuffer.c \ + intel_batchbuffer_dump.c\ intel_memman.c \ intel_driver.c \ i965_media.c \ i965_media_mpeg2.c \ + i965_media_h264.c \ i965_render.c \ - i965_drv_video.c + i965_drv_video.c \ + i965_avc_bsd.c \ + i965_avc_hw_scoreboard.c noinst_HEADERS = \ object_heap.h \ intel_batchbuffer.h \ + intel_batchbuffer_dump.h\ intel_memman.h \ intel_driver.h \ i965_media.h \ i965_media_mpeg2.h \ + i965_media_h264.h \ i965_render.h \ i965_drv_video.h \ i965_defines.h \ - i965_structs.h + i965_structs.h \ + i965_avc_bsd.h \ + i965_avc_hw_scoreboard.h diff --git a/i965_avc_bsd.c b/i965_avc_bsd.c new file mode 100644 index 0000000..9770207 --- /dev/null +++ b/i965_avc_bsd.c @@ -0,0 +1,801 @@ +/* + * Copyright © 2010 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Xiang Haihao <haihao.xiang@intel.com> + * + */ +#include <stdio.h> +#include <string.h> +#include <assert.h> + +#include "va_backend.h" + +#include "intel_batchbuffer.h" +#include "intel_driver.h" + +#include "i965_defines.h" +#include "i965_drv_video.h" +#include "i965_avc_bsd.h" +#include "i965_media_h264.h" +#include "i965_media.h" + +static void +i965_bsd_ind_obj_base_address(VADriverContextP ctx, struct decode_state *decode_state) +{ + dri_bo *ind_bo = decode_state->slice_data->bo; + + BEGIN_BCS_BATCH(ctx, 3); + OUT_BCS_BATCH(ctx, CMD_BSD_IND_OBJ_BASE_ADDR | (3 - 2)); + OUT_BCS_RELOC(ctx, ind_bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0); + OUT_BCS_BATCH(ctx, 0); + ADVANCE_BCS_BATCH(ctx); +} + +static void +i965_avc_bsd_img_state(VADriverContextP ctx, struct decode_state *decode_state) +{ + int qm_present_flag; + int img_struct; + int mbaff_frame_flag; + unsigned int avc_it_command_header; + unsigned int width_in_mbs, height_in_mbs; + VAPictureParameterBufferH264 *pic_param; + + if (decode_state->iq_matrix && decode_state->iq_matrix->buffer) + qm_present_flag = 1; + else + qm_present_flag = 0; /* built-in QM matrices */ + + assert(decode_state->pic_param && decode_state->pic_param->buffer); + pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; + + assert(!(pic_param->CurrPic.flags & VA_PICTURE_H264_INVALID)); + + if (pic_param->CurrPic.flags & VA_PICTURE_H264_TOP_FIELD) + img_struct = 1; + else if (pic_param->CurrPic.flags & VA_PICTURE_H264_BOTTOM_FIELD) + img_struct = 3; + else + img_struct = 0; + + if ((img_struct & 0x1) == 0x1) { + assert(pic_param->pic_fields.bits.field_pic_flag == 0x1); + } else { + assert(pic_param->pic_fields.bits.field_pic_flag == 0x0); + } + + if (pic_param->seq_fields.bits.frame_mbs_only_flag) { /* a frame containing only frame macroblocks */ + assert(pic_param->seq_fields.bits.mb_adaptive_frame_field_flag == 0); + assert(pic_param->pic_fields.bits.field_pic_flag == 0); + } else { + assert(pic_param->seq_fields.bits.direct_8x8_inference_flag == 1); /* see H.264 spec */ + } + + mbaff_frame_flag = (pic_param->seq_fields.bits.mb_adaptive_frame_field_flag && + !pic_param->pic_fields.bits.field_pic_flag); + + width_in_mbs = ((pic_param->picture_width_in_mbs_minus1 + 1) & 0xff); + height_in_mbs = ((pic_param->picture_height_in_mbs_minus1 + 1) & 0xff); /* frame height */ + + assert(!((width_in_mbs * height_in_mbs) & 0x8000)); /* hardware requirement */ + + /* BSD unit doesn't support 4:2:2 and 4:4:4 picture */ + assert(pic_param->seq_fields.bits.chroma_format_idc == 0 || /* monochrome picture */ + pic_param->seq_fields.bits.chroma_format_idc == 1); /* 4:2:0 */ + assert(pic_param->seq_fields.bits.residual_colour_transform_flag == 0); /* only available for 4:4:4 */ + + avc_it_command_header = (CMD_MEDIA_OBJECT_EX | (12 - 2)); + + BEGIN_BCS_BATCH(ctx, 6); + OUT_BCS_BATCH(ctx, CMD_AVC_BSD_IMG_STATE | (6 - 2)); + OUT_BCS_BATCH(ctx, + ((width_in_mbs * height_in_mbs) & 0x7fff)); + OUT_BCS_BATCH(ctx, + (height_in_mbs << 16) | + (width_in_mbs << 0)); + OUT_BCS_BATCH(ctx, + (pic_param->second_chroma_qp_index_offset << 24) | + (pic_param->chroma_qp_index_offset << 16) | + (SCAN_RASTER_ORDER << 15) | /* AVC ILDB Data */ + (SCAN_SPECIAL_ORDER << 14) | /* AVC IT Command */ + (SCAN_RASTER_ORDER << 13) | /* AVC IT Data */ + (1 << 12) | /* always 1, hardware requirement */ + (qm_present_flag << 10) | + (img_struct << 8) | + (16 << 0)); /* FIXME: always support 16 reference frames ??? */ + OUT_BCS_BATCH(ctx, + (RESIDUAL_DATA_OFFSET << 24) | /* residual data offset */ + (0 << 17) | /* don't overwrite SRT */ + (0 << 16) | /* Un-SRT (Unsynchronized Root Thread) */ + (0 << 12) | /* FIXME: no 16MV ??? */ + (pic_param->seq_fields.bits.chroma_format_idc << 10) | + (1 << 8) | /* Enable ILDB writing output */ + (pic_param->pic_fields.bits.entropy_coding_mode_flag << 7) | + ((!pic_param->pic_fields.bits.reference_pic_flag) << 6) | + (pic_param->pic_fields.bits.constrained_intra_pred_flag << 5) | + (pic_param->seq_fields.bits.direct_8x8_inference_flag << 4) | + (pic_param->pic_fields.bits.transform_8x8_mode_flag << 3) | + (pic_param->seq_fields.bits.frame_mbs_only_flag << 2) | + (mbaff_frame_flag << 1) | + (pic_param->pic_fields.bits.field_pic_flag << 0)); + OUT_BCS_BATCH(ctx, avc_it_command_header); + ADVANCE_BCS_BATCH(ctx); +} + +static void +i965_avc_bsd_qm_state(VADriverContextP ctx, struct decode_state *decode_state) +{ + int cmd_len; + VAIQMatrixBufferH264 *iq_matrix; + VAPictureParameterBufferH264 *pic_param; + + if (!decode_state->iq_matrix || !decode_state->iq_matrix->buffer) + return; + + iq_matrix = (VAIQMatrixBufferH264 *)decode_state->iq_matrix->buffer; + + assert(decode_state->pic_param && decode_state->pic_param->buffer); + pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; + + cmd_len = 2 + 6 * 4; /* always load six 4x4 scaling matrices */ + + if (pic_param->pic_fields.bits.transform_8x8_mode_flag) + cmd_len += 2 * 16; /* load two 8x8 scaling matrices */ + + BEGIN_BCS_BATCH(ctx, cmd_len); + OUT_BCS_BATCH(ctx, CMD_AVC_BSD_QM_STATE | (cmd_len - 2)); + + if (pic_param->pic_fields.bits.transform_8x8_mode_flag) + OUT_BCS_BATCH(ctx, + (0x0 << 8) | /* don't use default built-in matrices */ + (0xff << 0)); /* six 4x4 and two 8x8 scaling matrices */ + else + OUT_BCS_BATCH(ctx, + (0x0 << 8) | /* don't use default built-in matrices */ + (0x3f << 0)); /* six 4x4 scaling matrices */ + + intel_batchbuffer_data_bcs(ctx, &iq_matrix->ScalingList4x4[0][0], 6 * 4 * 4); + + if (pic_param->pic_fields.bits.transform_8x8_mode_flag) + intel_batchbuffer_data_bcs(ctx, &iq_matrix->ScalingList8x8[0][0], 2 * 16 * 4); + + ADVANCE_BCS_BATCH(ctx); +} + +static void +i965_avc_bsd_slice_state(VADriverContextP ctx, + VAPictureParameterBufferH264 *pic_param, + VASliceParameterBufferH264 *slice_param) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context = (struct i965_h264_context *)media_state->private_context; + int present_flag, cmd_len, list, j; + struct { + unsigned char bottom_idc:1; + unsigned char frame_store_index:4; + unsigned char field_picture:1; + unsigned char long_term:1; + unsigned char non_exist:1; + } refs[32]; + char weightoffsets[32 * 6]; + + /* don't issue SLICE_STATE for intra-prediction decoding */ + if (slice_param->slice_type == SLICE_TYPE_I) + return; + + cmd_len = 2; + + if (slice_param->slice_type == SLICE_TYPE_P) { + present_flag = PRESENT_REF_LIST0; + cmd_len += 8; + } else { + present_flag = PRESENT_REF_LIST0 | PRESENT_REF_LIST1; + cmd_len += 16; + } + + if (slice_param->luma_weight_l0_flag | slice_param->chroma_weight_l0_flag) { + present_flag |= PRESENT_WEIGHT_OFFSET_L0; + cmd_len += 48; + assert((pic_param->pic_fields.bits.weighted_pred_flag == 1) || /* P slice */ + (pic_param->pic_fields.bits.weighted_bipred_idc == 1)); /* B slice */ + } + + if (slice_param->luma_weight_l1_flag | slice_param->chroma_weight_l1_flag) { + present_flag |= PRESENT_WEIGHT_OFFSET_L1; + cmd_len += 48; + assert(slice_param->slice_type == SLICE_TYPE_B); + assert(pic_param->pic_fields.bits.weighted_bipred_idc == 1); + } + + BEGIN_BCS_BATCH(ctx, cmd_len); + OUT_BCS_BATCH(ctx, CMD_AVC_BSD_SLICE_STATE | (cmd_len - 2)); + OUT_BCS_BATCH(ctx, present_flag); + + for (list = 0; list < 2; list++) { + int flag; + VAPictureH264 *va_pic; + + if (list == 0) { + flag = PRESENT_REF_LIST0; + va_pic = slice_param->RefPicList0; + } else { + flag = PRESENT_REF_LIST1; + va_pic = slice_param->RefPicList1; + } + + if (!(present_flag & flag)) + continue; + + for (j = 0; j < 32; j++) { + if (va_pic->flags & VA_PICTURE_H264_INVALID) { + refs[j].non_exist = 1; + refs[j].long_term = 1; + refs[j].field_picture = 1; + refs[j].frame_store_index = 0xf; + refs[j].bottom_idc = 1; + } else { + int frame_idx; + + for (frame_idx = 0; frame_idx < 16; frame_idx++) { + VAPictureH264 *ref_pic = &pic_param->ReferenceFrames[frame_idx]; + + if (!(ref_pic->flags & VA_PICTURE_H264_INVALID)) { + if (ref_pic->picture_id == va_pic->picture_id) + break; + } + } + + assert(frame_idx < 16); + + refs[j].non_exist = 0; + refs[j].long_term = !!(va_pic->flags & VA_PICTURE_H264_LONG_TERM_REFERENCE); + refs[j].field_picture = !!(va_pic->flags & + (VA_PICTURE_H264_TOP_FIELD | + VA_PICTURE_H264_BOTTOM_FIELD)); + refs[j].frame_store_index = frame_idx; + refs[j].bottom_idc = !!(va_pic->flags & VA_PICTURE_H264_BOTTOM_FIELD); + } + + va_pic++; + } + + intel_batchbuffer_data_bcs(ctx, refs, sizeof(refs)); + } + + i965_h264_context->weight128_luma_l0 = 0; + i965_h264_context->weight128_luma_l1 = 0; + i965_h264_context->weight128_chroma_l0 = 0; + i965_h264_context->weight128_chroma_l1 = 0; + + i965_h264_context->weight128_offset0_flag = 0; + i965_h264_context->weight128_offset0 = 0; + + if (present_flag & PRESENT_WEIGHT_OFFSET_L0) { + for (j = 0; j < 32; j++) { + weightoffsets[j * 6 + 0] = slice_param->luma_offset_l0[j]; + weightoffsets[j * 6 + 1] = slice_param->luma_weight_l0[j]; + weightoffsets[j * 6 + 2] = slice_param->chroma_offset_l0[j][0]; + weightoffsets[j * 6 + 3] = slice_param->chroma_weight_l0[j][0]; + weightoffsets[j * 6 + 4] = slice_param->chroma_offset_l0[j][1]; + weightoffsets[j * 6 + 5] = slice_param->chroma_weight_l0[j][1]; + + if (pic_param->pic_fields.bits.weighted_bipred_idc == 1) { + if (i965_h264_context->use_hw_w128) { + if (slice_param->luma_weight_l0[j] == 128) + i965_h264_context->weight128_luma_l0 |= (1 << j); + + if (slice_param->chroma_weight_l0[j][0] == 128 || + slice_param->chroma_weight_l0[j][1] == 128) + i965_h264_context->weight128_chroma_l0 |= (1 << j); + } else { + /* FIXME: workaround for weight 128 */ + if (slice_param->luma_weight_l0[j] == 128 || + slice_param->chroma_weight_l0[j][0] == 128 || + slice_param->chroma_weight_l0[j][1] == 128) + i965_h264_context->weight128_offset0_flag = 1; + } + } + } + + intel_batchbuffer_data_bcs(ctx, weightoffsets, sizeof(weightoffsets)); + } + + if (present_flag & PRESENT_WEIGHT_OFFSET_L1) { + for (j = 0; j < 32; j++) { + weightoffsets[j * 6 + 0] = slice_param->luma_offset_l1[j]; + weightoffsets[j * 6 + 1] = slice_param->luma_weight_l1[j]; + weightoffsets[j * 6 + 2] = slice_param->chroma_offset_l1[j][0]; + weightoffsets[j * 6 + 3] = slice_param->chroma_weight_l1[j][0]; + weightoffsets[j * 6 + 4] = slice_param->chroma_offset_l1[j][1]; + weightoffsets[j * 6 + 5] = slice_param->chroma_weight_l1[j][1]; + + if (pic_param->pic_fields.bits.weighted_bipred_idc == 1) { + if (i965_h264_context->use_hw_w128) { + if (slice_param->luma_weight_l1[j] == 128) + i965_h264_context->weight128_luma_l1 |= (1 << j); + + if (slice_param->chroma_weight_l1[j][0] == 128 || + slice_param->chroma_weight_l1[j][1] == 128) + i965_h264_context->weight128_chroma_l1 |= (1 << j); + } else { + if (slice_param->luma_weight_l0[j] == 128 || + slice_param->chroma_weight_l0[j][0] == 128 || + slice_param->chroma_weight_l0[j][1] == 128) + i965_h264_context->weight128_offset0_flag = 1; + } + } + } + + intel_batchbuffer_data_bcs(ctx, weightoffsets, sizeof(weightoffsets)); + } + + ADVANCE_BCS_BATCH(ctx); +} + +static void +i965_avc_bsd_buf_base_state(VADriverContextP ctx, struct decode_state *decode_state) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context; + struct i965_avc_bsd_context *i965_avc_bsd_context; + int i; + VAPictureParameterBufferH264 *pic_param; + VAPictureH264 *va_pic; + struct object_surface *obj_surface; + + assert(decode_state->pic_param && decode_state->pic_param->buffer); + pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; + + assert(media_state->private_context); + i965_h264_context = (struct i965_h264_context *)media_state->private_context; + i965_avc_bsd_context = &i965_h264_context->i965_avc_bsd_context; + + BEGIN_BCS_BATCH(ctx, 74); + OUT_BCS_BATCH(ctx, CMD_AVC_BSD_BUF_BASE_STATE | (74 - 2)); + OUT_BCS_RELOC(ctx, i965_avc_bsd_context->bsd_raw_store.bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + 0); + OUT_BCS_RELOC(ctx, i965_avc_bsd_context->mpr_row_store.bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + 0); + OUT_BCS_RELOC(ctx, i965_h264_context->avc_it_command_mb_info.bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + i965_h264_context->avc_it_command_mb_info.mbs * i965_h264_context->use_avc_hw_scoreboard * MB_CMD_IN_BYTES); + OUT_BCS_RELOC(ctx, i965_h264_context->avc_it_data.bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + (i965_h264_context->avc_it_data.write_offset << 6)); + OUT_BCS_RELOC(ctx, i965_avc_bsd_context->ildb_data.bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + 0); + + for (i = 0; i < 16; i++) { + va_pic = &pic_param->ReferenceFrames[i]; + + if (!(va_pic->flags & VA_PICTURE_H264_INVALID)) { + obj_surface = SURFACE(va_pic->picture_id); + assert(obj_surface); + OUT_BCS_RELOC(ctx, obj_surface->direct_mv_wr_top_bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0); + + if (pic_param->pic_fields.bits.field_pic_flag && + !pic_param->seq_fields.bits.direct_8x8_inference_flag) + OUT_BCS_RELOC(ctx, obj_surface->direct_mv_wr_bottom_bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0); + else + OUT_BCS_RELOC(ctx, obj_surface->direct_mv_wr_top_bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0); + } else { + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + } + } + + va_pic = &pic_param->CurrPic; + assert(!(va_pic->flags & VA_PICTURE_H264_INVALID)); + obj_surface = SURFACE(va_pic->picture_id); + assert(obj_surface); + OUT_BCS_RELOC(ctx, obj_surface->direct_mv_wr_top_bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + 0); + + if (pic_param->pic_fields.bits.field_pic_flag && + !pic_param->seq_fields.bits.direct_8x8_inference_flag) + OUT_BCS_RELOC(ctx, obj_surface->direct_mv_wr_bottom_bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + 0); + else + OUT_BCS_RELOC(ctx, obj_surface->direct_mv_wr_top_bo, + I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, + 0); + + /* POC List */ + for (i = 0; i < 16; i++) { + va_pic = &pic_param->ReferenceFrames[i]; + if (!(va_pic->flags & VA_PICTURE_H264_INVALID)) { + OUT_BCS_BATCH(ctx, va_pic->TopFieldOrderCnt); + OUT_BCS_BATCH(ctx, va_pic->BottomFieldOrderCnt); + } else { + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + } + } + + OUT_BCS_BATCH(ctx, va_pic->TopFieldOrderCnt); + OUT_BCS_BATCH(ctx, va_pic->BottomFieldOrderCnt); + + ADVANCE_BCS_BATCH(ctx); +} + +static void +g4x_avc_bsd_object(VADriverContextP ctx, + struct decode_state *decode_state, + VAPictureParameterBufferH264 *pic_param, + VASliceParameterBufferH264 *slice_param) +{ + int width_in_mbs = pic_param->picture_width_in_mbs_minus1 + 1; + int height_in_mbs = pic_param->picture_height_in_mbs_minus1 + 1; /* frame height */ + + if (slice_param) { + int encrypted, counter_value, cmd_len; + int slice_hor_pos, slice_ver_pos; + int num_ref_idx_l0, num_ref_idx_l1; + int field_or_mbaff_picture = (pic_param->pic_fields.bits.field_pic_flag || + pic_param->seq_fields.bits.mb_adaptive_frame_field_flag); + int slice_data_bit_offset; + int weighted_pred_idc = 0; + + encrypted = 0; /* FIXME: which flag in VAAPI is used for encryption? */ + + if (encrypted) { + cmd_len = 9; + counter_value = 0; /* FIXME: ??? */ + } else + cmd_len = 8; + + slice_data_bit_offset = slice_param->slice_data_bit_offset; + + if (pic_param->pic_fields.bits.entropy_coding_mode_flag == ENTROPY_CABAC) + slice_data_bit_offset = ALIGN(slice_data_bit_offset, 0x8); + + if (slice_param->slice_type == SLICE_TYPE_I) { + assert(slice_param->num_ref_idx_l0_active_minus1 == 0); + assert(slice_param->num_ref_idx_l1_active_minus1 == 0); + num_ref_idx_l0 = 0; + num_ref_idx_l1 = 0; + } else if (slice_param->slice_type == SLICE_TYPE_P) { + assert(slice_param->num_ref_idx_l1_active_minus1 == 0); + num_ref_idx_l0 = slice_param->num_ref_idx_l0_active_minus1 + 1; + num_ref_idx_l1 = 0; + } else { + num_ref_idx_l0 = slice_param->num_ref_idx_l0_active_minus1 + 1; + num_ref_idx_l1 = slice_param->num_ref_idx_l1_active_minus1 + 1; + } + + if (slice_param->slice_type == SLICE_TYPE_P) + weighted_pred_idc = pic_param->pic_fields.bits.weighted_pred_flag; + else if (slice_param->slice_type == SLICE_TYPE_B) + weighted_pred_idc = pic_param->pic_fields.bits.weighted_bipred_idc; + + slice_hor_pos = slice_param->first_mb_in_slice % width_in_mbs; + slice_ver_pos = slice_param->first_mb_in_slice / width_in_mbs; + slice_ver_pos <<= (1 + field_or_mbaff_picture); /* FIXME: right ??? */ + + BEGIN_BCS_BATCH(ctx, cmd_len); + OUT_BCS_BATCH(ctx, CMD_AVC_BSD_OBJECT | (cmd_len - 2)); + OUT_BCS_BATCH(ctx, + (encrypted << 31) | + ((slice_param->slice_data_size - (slice_data_bit_offset >> 3)) << 0)); + OUT_BCS_BATCH(ctx, + (slice_param->slice_data_offset + + (slice_data_bit_offset >> 3))); + OUT_BCS_BATCH(ctx, + (0 << 31) | /* concealment mode: 0->intra 16x16 prediction, 1->inter P Copy */ + (0 << 14) | /* ignore BSDPrematureComplete Error handling */ + (0 << 13) | /* FIXME: ??? */ + (0 << 12) | /* ignore MPR Error handling */ + (0 << 10) | /* ignore Entropy Error handling */ + (0 << 8) | /* ignore MB Header Error handling */ + (slice_param->slice_type << 0)); + OUT_BCS_BATCH(ctx, + (num_ref_idx_l1 << 24) | + (num_ref_idx_l0 << 16) | + (slice_param->chroma_log2_weight_denom << 8) | + (slice_param->luma_log2_weight_denom << 0)); + OUT_BCS_BATCH(ctx, + (weighted_pred_idc << 30) | + (slice_param->direct_spatial_mv_pred_flag << 29) | + (slice_param->disable_deblocking_filter_idc << 27) | + (slice_param->cabac_init_idc << 24) | + ((pic_param->pic_init_qp_minus26 + 26 + slice_param->slice_qp_delta) << 16) | + ((slice_param->slice_beta_offset_div2 & 0xf) << 8) | + ((slice_param->slice_alpha_c0_offset_div2 & 0xf) << 0)); + OUT_BCS_BATCH(ctx, + (slice_ver_pos << 24) | + (slice_hor_pos << 16) | + (slice_param->first_mb_in_slice << 0)); + OUT_BCS_BATCH(ctx, + (0 << 7) | /* FIXME: ??? */ + ((0x7 - (slice_data_bit_offset & 0x7)) << 0)); + + if (encrypted) { + OUT_BCS_BATCH(ctx, counter_value); + } + + ADVANCE_BCS_BATCH(ctx); + } else { + BEGIN_BCS_BATCH(ctx, 8); + OUT_BCS_BATCH(ctx, CMD_AVC_BSD_OBJECT | (8 - 2)); + OUT_BCS_BATCH(ctx, 0); /* indirect data length for phantom slice is 0 */ + OUT_BCS_BATCH(ctx, 0); /* indirect data start address for phantom slice is 0 */ + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, width_in_mbs * height_in_mbs / (1 + !!pic_param->pic_fields.bits.field_pic_flag)); + OUT_BCS_BATCH(ctx, 0); + ADVANCE_BCS_BATCH(ctx); + } +} + +static void +ironlake_avc_bsd_object(VADriverContextP ctx, + struct decode_state *decode_state, + VAPictureParameterBufferH264 *pic_param, + VASliceParameterBufferH264 *slice_param) +{ + int width_in_mbs = pic_param->picture_width_in_mbs_minus1 + 1; + int height_in_mbs = pic_param->picture_height_in_mbs_minus1 + 1; /* frame height */ + + if (slice_param) { + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context = (struct i965_h264_context *)media_state->private_context; + int encrypted, counter_value; + int slice_hor_pos, slice_ver_pos; + int num_ref_idx_l0, num_ref_idx_l1; + int field_or_mbaff_picture = (pic_param->pic_fields.bits.field_pic_flag || + pic_param->seq_fields.bits.mb_adaptive_frame_field_flag); + int slice_data_bit_offset; + int weighted_pred_idc = 0; + + encrypted = 0; /* FIXME: which flag in VAAPI is used for encryption? */ + + if (encrypted) { + counter_value = 0; /* FIXME: ??? */ + } else + counter_value = 0; + + slice_data_bit_offset = slice_param->slice_data_bit_offset; + + if (pic_param->pic_fields.bits.entropy_coding_mode_flag == ENTROPY_CABAC) + slice_data_bit_offset = ALIGN(slice_data_bit_offset, 0x8); + + if (slice_param->slice_type == SLICE_TYPE_I) { + assert(slice_param->num_ref_idx_l0_active_minus1 == 0); + assert(slice_param->num_ref_idx_l1_active_minus1 == 0); + num_ref_idx_l0 = 0; + num_ref_idx_l1 = 0; + } else if (slice_param->slice_type == SLICE_TYPE_P) { + assert(slice_param->num_ref_idx_l1_active_minus1 == 0); + num_ref_idx_l0 = slice_param->num_ref_idx_l0_active_minus1 + 1; + num_ref_idx_l1 = 0; + } else { + num_ref_idx_l0 = slice_param->num_ref_idx_l0_active_minus1 + 1; + num_ref_idx_l1 = slice_param->num_ref_idx_l1_active_minus1 + 1; + } + + if (slice_param->slice_type == SLICE_TYPE_P) + weighted_pred_idc = pic_param->pic_fields.bits.weighted_pred_flag; + else if (slice_param->slice_type == SLICE_TYPE_B) + weighted_pred_idc = pic_param->pic_fields.bits.weighted_bipred_idc; + + slice_hor_pos = slice_param->first_mb_in_slice % width_in_mbs; + slice_ver_pos = slice_param->first_mb_in_slice / width_in_mbs; + slice_ver_pos <<= (1 + field_or_mbaff_picture); /* FIXME: right ??? */ + + BEGIN_BCS_BATCH(ctx, 16); + OUT_BCS_BATCH(ctx, CMD_AVC_BSD_OBJECT | (16 - 2)); + OUT_BCS_BATCH(ctx, + (encrypted << 31) | + (0 << 30) | /* FIXME: packet based bit stream */ + (0 << 29) | /* FIXME: packet format */ + ((slice_param->slice_data_size - (slice_data_bit_offset >> 3)) << 0)); + OUT_BCS_BATCH(ctx, + (slice_param->slice_data_offset + + (slice_data_bit_offset >> 3))); + OUT_BCS_BATCH(ctx, + (0 << 31) | /* concealment mode: 0->intra 16x16 prediction, 1->inter P Copy */ + (0 << 14) | /* ignore BSDPrematureComplete Error handling */ + (0 << 13) | /* FIXME: ??? */ + (0 << 12) | /* ignore MPR Error handling */ + (0 << 10) | /* ignore Entropy Error handling */ + (0 << 8) | /* ignore MB Header Error handling */ + (slice_param->slice_type << 0)); + OUT_BCS_BATCH(ctx, + (num_ref_idx_l1 << 24) | + (num_ref_idx_l0 << 16) | + (slice_param->chroma_log2_weight_denom << 8) | + (slice_param->luma_log2_weight_denom << 0)); + OUT_BCS_BATCH(ctx, + (weighted_pred_idc << 30) | + (slice_param->direct_spatial_mv_pred_flag << 29) | + (slice_param->disable_deblocking_filter_idc << 27) | + (slice_param->cabac_init_idc << 24) | + ((pic_param->pic_init_qp_minus26 + 26 + slice_param->slice_qp_delta) << 16) | + ((slice_param->slice_beta_offset_div2 & 0xf) << 8) | + ((slice_param->slice_alpha_c0_offset_div2 & 0xf) << 0)); + OUT_BCS_BATCH(ctx, + (slice_ver_pos << 24) | + (slice_hor_pos << 16) | + (slice_param->first_mb_in_slice << 0)); + OUT_BCS_BATCH(ctx, + (0 << 7) | /* FIXME: ??? */ + ((0x7 - (slice_data_bit_offset & 0x7)) << 0)); + OUT_BCS_BATCH(ctx, counter_value); + + /* FIXME: dw9-dw11 */ + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, i965_h264_context->weight128_luma_l0); + OUT_BCS_BATCH(ctx, i965_h264_context->weight128_luma_l1); + OUT_BCS_BATCH(ctx, i965_h264_context->weight128_chroma_l0); + OUT_BCS_BATCH(ctx, i965_h264_context->weight128_chroma_l1); + + ADVANCE_BCS_BATCH(ctx); + } else { + BEGIN_BCS_BATCH(ctx, 16); + OUT_BCS_BATCH(ctx, CMD_AVC_BSD_OBJECT | (16 - 2)); + OUT_BCS_BATCH(ctx, 0); /* indirect data length for phantom slice is 0 */ + OUT_BCS_BATCH(ctx, 0); /* indirect data start address for phantom slice is 0 */ + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, width_in_mbs * height_in_mbs / (1 + !!pic_param->pic_fields.bits.field_pic_flag)); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + OUT_BCS_BATCH(ctx, 0); + ADVANCE_BCS_BATCH(ctx); + } +} + +static void +i965_avc_bsd_object(VADriverContextP ctx, + struct decode_state *decode_state, + VAPictureParameterBufferH264 *pic_param, + VASliceParameterBufferH264 *slice_param) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + + if (IS_IRONLAKE(i965->intel.device_id)) + ironlake_avc_bsd_object(ctx, decode_state, pic_param, slice_param); + else + g4x_avc_bsd_object(ctx, decode_state, pic_param, slice_param); +} + +static void +i965_avc_bsd_phantom_slice(VADriverContextP ctx, + struct decode_state *decode_state, + VAPictureParameterBufferH264 *pic_param) +{ + i965_avc_bsd_object(ctx, decode_state, pic_param, NULL); +} + +void +i965_avc_bsd_pipeline(VADriverContextP ctx, struct decode_state *decode_state) +{ + int i; + VAPictureParameterBufferH264 *pic_param; + VASliceParameterBufferH264 *slice_param; + + assert(decode_state->pic_param && decode_state->pic_param->buffer); + pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; + assert(decode_state->slice_param && decode_state->slice_param->buffer); + slice_param = (VASliceParameterBufferH264 *)decode_state->slice_param->buffer; + + intel_batchbuffer_start_atomic_bcs(ctx, 0x1000); + i965_bsd_ind_obj_base_address(ctx, decode_state); + + assert(decode_state->num_slices == 1); /* FIXME: */ + for (i = 0; i < decode_state->num_slices; i++) { + assert(slice_param->slice_data_flag == VA_SLICE_DATA_FLAG_ALL); + assert((slice_param->slice_type == SLICE_TYPE_I) || + (slice_param->slice_type == SLICE_TYPE_P) || + (slice_param->slice_type == SLICE_TYPE_B)); /* hardware requirement */ + + if (i == 0) { + i965_avc_bsd_img_state(ctx, decode_state); + i965_avc_bsd_qm_state(ctx, decode_state); + } + + i965_avc_bsd_slice_state(ctx, pic_param, slice_param); + i965_avc_bsd_buf_base_state(ctx, decode_state); + i965_avc_bsd_object(ctx, decode_state, pic_param, slice_param); + slice_param++; + } + + i965_avc_bsd_phantom_slice(ctx, decode_state, pic_param); + intel_batchbuffer_emit_mi_flush_bcs(ctx); + intel_batchbuffer_end_atomic_bcs(ctx); + intel_batchbuffer_flush_bcs(ctx); +} + +void +i965_avc_bsd_decode_init(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context = (struct i965_h264_context *)media_state->private_context; + struct i965_avc_bsd_context *i965_avc_bsd_context; + dri_bo *bo; + + assert(i965_h264_context); + i965_avc_bsd_context = &i965_h264_context->i965_avc_bsd_context; + + dri_bo_unreference(i965_avc_bsd_context->bsd_raw_store.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "bsd raw store", + 0x4000, /* at least 11520 bytes to support 120 MBs per row */ + 64); + assert(bo); + i965_avc_bsd_context->bsd_raw_store.bo = bo; + + dri_bo_unreference(i965_avc_bsd_context->mpr_row_store.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "mpr row store", + 0x2000, /* at least 7680 bytes to support 120 MBs per row */ + 64); + assert(bo); + i965_avc_bsd_context->mpr_row_store.bo = bo; + + dri_bo_unreference(i965_avc_bsd_context->ildb_data.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "ildb data", + 0x100000, /* at least 1044480 bytes */ + 64); + assert(bo); + i965_avc_bsd_context->ildb_data.bo = bo; +} + +Bool +i965_avc_bsd_ternimate(struct i965_avc_bsd_context *i965_avc_bsd_context) +{ + dri_bo_unreference(i965_avc_bsd_context->bsd_raw_store.bo); + dri_bo_unreference(i965_avc_bsd_context->mpr_row_store.bo); + dri_bo_unreference(i965_avc_bsd_context->ildb_data.bo); + + return True; +} diff --git a/i965_avc_bsd.h b/i965_avc_bsd.h new file mode 100644 index 0000000..7c373f3 --- /dev/null +++ b/i965_avc_bsd.h @@ -0,0 +1,61 @@ +/* + * Copyright © 2010 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Xiang Haihao <haihao.xiang@intel.com> + * + */ + +#ifndef __I965_AVC_BSD_H__ +#define __I965_AVC_BSD_H__ + +struct i965_avc_bsd_context +{ + struct { + dri_bo *bo; + } bsd_raw_store; + + struct { + dri_bo *bo; + } mpr_row_store; + + struct { + dri_bo *bo; + } avc_it_command_mb_info; + + struct { + dri_bo *bo; + long write_offset; + } avc_it_data; + + struct { + dri_bo *bo; + } ildb_data; +}; + +void i965_avc_bsd_pipeline(VADriverContextP, struct decode_state *); +void i965_avc_bsd_decode_init(VADriverContextP); +Bool i965_avc_bsd_ternimate(struct i965_avc_bsd_context *); + +#endif /* __I965_AVC_BSD_H__ */ + diff --git a/i965_avc_hw_scoreboard.c b/i965_avc_hw_scoreboard.c new file mode 100644 index 0000000..9d6d973 --- /dev/null +++ b/i965_avc_hw_scoreboard.c @@ -0,0 +1,420 @@ +/* + * Copyright © 2010 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Xiang Haihao <haihao.xiang@intel.com> + * + */ + +#include <stdio.h> +#include <string.h> +#include <assert.h> + +#include "va_backend.h" + +#include "intel_batchbuffer.h" +#include "intel_driver.h" + +#include "i965_defines.h" +#include "i965_drv_video.h" +#include "i965_avc_hw_scoreboard.h" +#include "i965_media_h264.h" +#include "i965_media.h" + +extern struct media_kernel *h264_avc_kernels; + +/* On Ironlake */ +#include "shaders/h264/mc/export.inc.gen5" + +enum { + AVC_HW_SCOREBOARD = 0, + AVC_HW_SCOREBOARD_MBAFF +}; + +static unsigned long avc_hw_scoreboard_kernel_offset[] = { + SETHWSCOREBOARD_IP_GEN5, + SETHWSCOREBOARD_MBAFF_IP_GEN5 +}; + +static unsigned int avc_hw_scoreboard_constants[] = { + 0x08040201, + 0x00000010, + 0x08000210, + 0x00000000, + 0x08040201, + 0x08040210, + 0x01000010, + 0x08040200 +}; + +static void +i965_avc_hw_scoreboard_surface_state(struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + struct i965_surface_state *ss; + dri_bo *bo; + + bo = avc_hw_scoreboard_context->surface.ss_bo; + dri_bo_map(bo, 1); + assert(bo->virtual); + ss = bo->virtual; + memset(ss, 0, sizeof(*ss)); + ss->ss0.surface_type = I965_SURFACE_BUFFER; + ss->ss1.base_addr = avc_hw_scoreboard_context->surface.s_bo->offset; + ss->ss2.width = ((avc_hw_scoreboard_context->surface.total_mbs * MB_CMD_IN_OWS - 1) & 0x7f); + ss->ss2.height = (((avc_hw_scoreboard_context->surface.total_mbs * MB_CMD_IN_OWS - 1) >> 7) & 0x1fff); + ss->ss3.depth = (((avc_hw_scoreboard_context->surface.total_mbs * MB_CMD_IN_OWS - 1) >> 20) & 0x7f); + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER, + 0, + offsetof(struct i965_surface_state, ss1), + avc_hw_scoreboard_context->surface.s_bo); + dri_bo_unmap(bo); +} + +static void +i965_avc_hw_scoreboard_interface_descriptor_table(struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + struct i965_interface_descriptor *desc; + dri_bo *bo; + + bo = avc_hw_scoreboard_context->idrt.bo; + dri_bo_map(bo, 1); + assert(bo->virtual); + desc = bo->virtual; + memset(desc, 0, sizeof(*desc)); + desc->desc0.grf_reg_blocks = 7; + desc->desc0.kernel_start_pointer = (avc_hw_scoreboard_context->hw_kernel.bo->offset + + avc_hw_scoreboard_context->hw_kernel.offset) >> 6; /* reloc */ + desc->desc1.const_urb_entry_read_offset = 0; + desc->desc1.const_urb_entry_read_len = 1; + desc->desc3.binding_table_entry_count = 0; + desc->desc3.binding_table_pointer = + avc_hw_scoreboard_context->binding_table.bo->offset >> 5; /*reloc */ + + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + desc->desc0.grf_reg_blocks + avc_hw_scoreboard_context->hw_kernel.offset, + offsetof(struct i965_interface_descriptor, desc0), + avc_hw_scoreboard_context->hw_kernel.bo); + + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + desc->desc3.binding_table_entry_count, + offsetof(struct i965_interface_descriptor, desc3), + avc_hw_scoreboard_context->binding_table.bo); + + dri_bo_unmap(bo); +} + +static void +i965_avc_hw_scoreboard_binding_table(struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + unsigned int *binding_table; + dri_bo *bo = avc_hw_scoreboard_context->binding_table.bo; + + dri_bo_map(bo, 1); + assert(bo->virtual); + binding_table = bo->virtual; + memset(binding_table, 0, bo->size); + binding_table[0] = avc_hw_scoreboard_context->surface.ss_bo->offset; + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0, + 0, + avc_hw_scoreboard_context->surface.ss_bo); + dri_bo_unmap(bo); +} + +static void +i965_avc_hw_scoreboard_vfe_state(struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + struct i965_vfe_state *vfe_state; + dri_bo *bo; + + bo = avc_hw_scoreboard_context->vfe_state.bo; + dri_bo_map(bo, 1); + assert(bo->virtual); + vfe_state = bo->virtual; + memset(vfe_state, 0, sizeof(*vfe_state)); + vfe_state->vfe1.max_threads = avc_hw_scoreboard_context->urb.num_vfe_entries - 1; + vfe_state->vfe1.urb_entry_alloc_size = avc_hw_scoreboard_context->urb.size_vfe_entry - 1; + vfe_state->vfe1.num_urb_entries = avc_hw_scoreboard_context->urb.num_vfe_entries; + vfe_state->vfe1.vfe_mode = VFE_GENERIC_MODE; + vfe_state->vfe1.children_present = 0; + vfe_state->vfe2.interface_descriptor_base = + avc_hw_scoreboard_context->idrt.bo->offset >> 4; /* reloc */ + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0, + offsetof(struct i965_vfe_state, vfe2), + avc_hw_scoreboard_context->idrt.bo); + dri_bo_unmap(bo); +} + +static void +i965_avc_hw_scoreboard_upload_constants(struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + unsigned char *constant_buffer; + + dri_bo_map(avc_hw_scoreboard_context->curbe.bo, 1); + assert(avc_hw_scoreboard_context->curbe.bo->virtual); + constant_buffer = avc_hw_scoreboard_context->curbe.bo->virtual; + memcpy(constant_buffer, avc_hw_scoreboard_constants, sizeof(avc_hw_scoreboard_constants)); + dri_bo_unmap(avc_hw_scoreboard_context->curbe.bo); +} + +static void +i965_avc_hw_scoreboard_states_setup(struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + i965_avc_hw_scoreboard_surface_state(avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_binding_table(avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_interface_descriptor_table(avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_vfe_state(avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_upload_constants(avc_hw_scoreboard_context); +} + +static void +i965_avc_hw_scoreboard_pipeline_select(VADriverContextP ctx) +{ + BEGIN_BATCH(ctx, 1); + OUT_BATCH(ctx, CMD_PIPELINE_SELECT | PIPELINE_SELECT_MEDIA); + ADVANCE_BATCH(ctx); +} + +static void +i965_avc_hw_scoreboard_urb_layout(VADriverContextP ctx, struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + unsigned int vfe_fence, cs_fence; + + vfe_fence = avc_hw_scoreboard_context->urb.cs_start; + cs_fence = URB_SIZE((&i965->intel)); + + BEGIN_BATCH(ctx, 3); + OUT_BATCH(ctx, CMD_URB_FENCE | UF0_VFE_REALLOC | UF0_CS_REALLOC | 1); + OUT_BATCH(ctx, 0); + OUT_BATCH(ctx, + (vfe_fence << UF2_VFE_FENCE_SHIFT) | /* VFE_SIZE */ + (cs_fence << UF2_CS_FENCE_SHIFT)); /* CS_SIZE */ + ADVANCE_BATCH(ctx); +} + +static void +i965_avc_hw_scoreboard_state_base_address(VADriverContextP ctx) +{ + BEGIN_BATCH(ctx, 8); + OUT_BATCH(ctx, CMD_STATE_BASE_ADDRESS | 6); + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + ADVANCE_BATCH(ctx); +} + +static void +i965_avc_hw_scoreboard_state_pointers(VADriverContextP ctx, struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + BEGIN_BATCH(ctx, 3); + OUT_BATCH(ctx, CMD_MEDIA_STATE_POINTERS | 1); + OUT_BATCH(ctx, 0); + OUT_RELOC(ctx, avc_hw_scoreboard_context->vfe_state.bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); + ADVANCE_BATCH(ctx); +} + +static void +i965_avc_hw_scoreboard_cs_urb_layout(VADriverContextP ctx, struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + BEGIN_BATCH(ctx, 2); + OUT_BATCH(ctx, CMD_CS_URB_STATE | 0); + OUT_BATCH(ctx, + ((avc_hw_scoreboard_context->urb.size_cs_entry - 1) << 4) | /* URB Entry Allocation Size */ + (avc_hw_scoreboard_context->urb.num_cs_entries << 0)); /* Number of URB Entries */ + ADVANCE_BATCH(ctx); +} + +static void +i965_avc_hw_scoreboard_constant_buffer(VADriverContextP ctx, struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + BEGIN_BATCH(ctx, 2); + OUT_BATCH(ctx, CMD_CONSTANT_BUFFER | (1 << 8) | (2 - 2)); + OUT_RELOC(ctx, avc_hw_scoreboard_context->curbe.bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + avc_hw_scoreboard_context->urb.size_cs_entry - 1); + ADVANCE_BATCH(ctx); +} + +static void +i965_avc_hw_scoreboard_objects(VADriverContextP ctx, struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + BEGIN_BATCH(ctx, 6); + OUT_BATCH(ctx, CMD_MEDIA_OBJECT | 4); + OUT_BATCH(ctx, 0); /* interface descriptor offset: 0 */ + OUT_BATCH(ctx, 0); /* no indirect data */ + OUT_BATCH(ctx, 0); + OUT_BATCH(ctx, ((avc_hw_scoreboard_context->inline_data.num_mb_cmds << 16) | + (avc_hw_scoreboard_context->inline_data.starting_mb_number << 0))); + OUT_BATCH(ctx, avc_hw_scoreboard_context->inline_data.pic_width_in_mbs); + ADVANCE_BATCH(ctx); +} + +static void +i965_avc_hw_scoreboard_pipeline_setup(VADriverContextP ctx, struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + intel_batchbuffer_start_atomic(ctx, 0x1000); + intel_batchbuffer_emit_mi_flush(ctx); + i965_avc_hw_scoreboard_pipeline_select(ctx); + i965_avc_hw_scoreboard_state_base_address(ctx); + i965_avc_hw_scoreboard_state_pointers(ctx, avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_urb_layout(ctx, avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_cs_urb_layout(ctx, avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_constant_buffer(ctx, avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_objects(ctx, avc_hw_scoreboard_context); + intel_batchbuffer_end_atomic(ctx); +} + +void +i965_avc_hw_scoreboard(VADriverContextP ctx, struct decode_state *decode_state) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context = (struct i965_h264_context *)media_state->private_context; + struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context; + + if (i965_h264_context->use_avc_hw_scoreboard) { + assert(i965_h264_context != NULL); + avc_hw_scoreboard_context = &i965_h264_context->avc_hw_scoreboard_context; + + avc_hw_scoreboard_context->inline_data.num_mb_cmds = i965_h264_context->avc_it_command_mb_info.mbs; + avc_hw_scoreboard_context->inline_data.starting_mb_number = i965_h264_context->avc_it_command_mb_info.mbs; + avc_hw_scoreboard_context->inline_data.pic_width_in_mbs = i965_h264_context->picture.width_in_mbs; + avc_hw_scoreboard_context->surface.total_mbs = i965_h264_context->avc_it_command_mb_info.mbs * 2; + + dri_bo_unreference(avc_hw_scoreboard_context->hw_kernel.bo); + avc_hw_scoreboard_context->hw_kernel.bo = h264_avc_kernels[H264_AVC_COMBINED].bo; + assert(avc_hw_scoreboard_context->hw_kernel.bo != NULL); + dri_bo_reference(avc_hw_scoreboard_context->hw_kernel.bo); + + if (i965_h264_context->picture.mbaff_frame_flag) + avc_hw_scoreboard_context->hw_kernel.offset = avc_hw_scoreboard_kernel_offset[AVC_HW_SCOREBOARD_MBAFF]; + else + avc_hw_scoreboard_context->hw_kernel.offset = avc_hw_scoreboard_kernel_offset[AVC_HW_SCOREBOARD]; + + i965_avc_hw_scoreboard_states_setup(avc_hw_scoreboard_context); + i965_avc_hw_scoreboard_pipeline_setup(ctx, avc_hw_scoreboard_context); + } +} + +void +i965_avc_hw_scoreboard_decode_init(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context = (struct i965_h264_context *)media_state->private_context; + struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context; + dri_bo *bo; + + if (i965_h264_context->use_avc_hw_scoreboard) { + assert(i965_h264_context != NULL); + avc_hw_scoreboard_context = &i965_h264_context->avc_hw_scoreboard_context; + + dri_bo_unreference(avc_hw_scoreboard_context->curbe.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "constant buffer", + 4096, 64); + assert(bo); + avc_hw_scoreboard_context->curbe.bo = bo; + + dri_bo_unreference(avc_hw_scoreboard_context->surface.s_bo); + avc_hw_scoreboard_context->surface.s_bo = i965_h264_context->avc_it_command_mb_info.bo; + assert(avc_hw_scoreboard_context->surface.s_bo != NULL); + dri_bo_reference(avc_hw_scoreboard_context->surface.s_bo); + + dri_bo_unreference(avc_hw_scoreboard_context->surface.ss_bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "surface state", + sizeof(struct i965_surface_state), 32); + assert(bo); + avc_hw_scoreboard_context->surface.ss_bo = bo; + + dri_bo_unreference(avc_hw_scoreboard_context->binding_table.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "binding table", + MAX_MEDIA_SURFACES * sizeof(unsigned int), 32); + assert(bo); + avc_hw_scoreboard_context->binding_table.bo = bo; + + dri_bo_unreference(avc_hw_scoreboard_context->idrt.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "interface discriptor", + MAX_INTERFACE_DESC * sizeof(struct i965_interface_descriptor), 16); + assert(bo); + avc_hw_scoreboard_context->idrt.bo = bo; + + dri_bo_unreference(avc_hw_scoreboard_context->vfe_state.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "vfe state", + sizeof(struct i965_vfe_state), 32); + assert(bo); + avc_hw_scoreboard_context->vfe_state.bo = bo; + + avc_hw_scoreboard_context->urb.num_vfe_entries = 1; + avc_hw_scoreboard_context->urb.size_vfe_entry = 2; + avc_hw_scoreboard_context->urb.num_cs_entries = 1; + avc_hw_scoreboard_context->urb.size_cs_entry = 1; + avc_hw_scoreboard_context->urb.vfe_start = 0; + avc_hw_scoreboard_context->urb.cs_start = avc_hw_scoreboard_context->urb.vfe_start + + avc_hw_scoreboard_context->urb.num_vfe_entries * avc_hw_scoreboard_context->urb.size_vfe_entry; + assert(avc_hw_scoreboard_context->urb.cs_start + + avc_hw_scoreboard_context->urb.num_cs_entries * avc_hw_scoreboard_context->urb.size_cs_entry <= URB_SIZE((&i965->intel))); + } +} + +Bool +i965_avc_hw_scoreboard_ternimate(struct i965_avc_hw_scoreboard_context *avc_hw_scoreboard_context) +{ + dri_bo_unreference(avc_hw_scoreboard_context->curbe.bo); + avc_hw_scoreboard_context->curbe.bo = NULL; + + dri_bo_unreference(avc_hw_scoreboard_context->surface.ss_bo); + avc_hw_scoreboard_context->surface.ss_bo = NULL; + + dri_bo_unreference(avc_hw_scoreboard_context->surface.s_bo); + avc_hw_scoreboard_context->surface.s_bo = NULL; + + dri_bo_unreference(avc_hw_scoreboard_context->binding_table.bo); + avc_hw_scoreboard_context->binding_table.bo = NULL; + + dri_bo_unreference(avc_hw_scoreboard_context->idrt.bo); + avc_hw_scoreboard_context->idrt.bo = NULL; + + dri_bo_unreference(avc_hw_scoreboard_context->vfe_state.bo); + avc_hw_scoreboard_context->vfe_state.bo = NULL; + + dri_bo_unreference(avc_hw_scoreboard_context->hw_kernel.bo); + avc_hw_scoreboard_context->hw_kernel.bo = NULL; + + return True; +} diff --git a/i965_avc_hw_scoreboard.h b/i965_avc_hw_scoreboard.h new file mode 100644 index 0000000..66ee9fb --- /dev/null +++ b/i965_avc_hw_scoreboard.h @@ -0,0 +1,84 @@ +/* + * Copyright © 2010 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Xiang Haihao <haihao.xiang@intel.com> + * + */ + +#ifndef __I965_AVC_HW_SCOREBOARD_H__ +#define __I965_AVC_HW_SCOREBOARD_H__ + +struct i965_avc_hw_scoreboard_context +{ + struct { + unsigned int num_mb_cmds; + unsigned int starting_mb_number; + unsigned int pic_width_in_mbs; + } inline_data; + + struct { + dri_bo *ss_bo; + dri_bo *s_bo; + unsigned int total_mbs; + } surface; + + struct { + dri_bo *bo; + } binding_table; + + struct { + dri_bo *bo; + } idrt; + + struct { + dri_bo *bo; + } vfe_state; + + struct { + dri_bo *bo; + } curbe; + + struct { + dri_bo *bo; + unsigned long offset; + } hw_kernel; + + struct { + unsigned int vfe_start; + unsigned int cs_start; + + unsigned int num_vfe_entries; + unsigned int num_cs_entries; + + unsigned int size_vfe_entry; + unsigned int size_cs_entry; + } urb; +}; + +void i965_avc_hw_scoreboard(VADriverContextP, struct decode_state *); +void i965_avc_hw_scoreboard_decode_init(VADriverContextP); +Bool i965_avc_hw_scoreboard_ternimate(struct i965_avc_hw_scoreboard_context *); + +#endif /* __I965_AVC_HW_SCOREBOARD_H__ */ + diff --git a/i965_defines.h b/i965_defines.h index 260d820..e50b41b 100644 --- a/i965_defines.h +++ b/i965_defines.h @@ -20,6 +20,13 @@ #define CMD_MEDIA_OBJECT CMD(2, 1, 0) #define CMD_MEDIA_OBJECT_EX CMD(2, 1, 1) +#define CMD_AVC_BSD_IMG_STATE CMD(2, 4, 0) +#define CMD_AVC_BSD_QM_STATE CMD(2, 4, 1) +#define CMD_AVC_BSD_SLICE_STATE CMD(2, 4, 2) +#define CMD_AVC_BSD_BUF_BASE_STATE CMD(2, 4, 3) +#define CMD_BSD_IND_OBJ_BASE_ADDR CMD(2, 4, 4) +#define CMD_AVC_BSD_OBJECT CMD(2, 4, 8) + #define CMD_PIPELINED_POINTERS CMD(3, 0, 0) #define CMD_BINDING_TABLE_POINTERS CMD(3, 0, 1) #define CMD_VERTEX_BUFFERS CMD(3, 0, 8) @@ -321,6 +328,32 @@ #define I965_TILEWALK_XMAJOR 0 #define I965_TILEWALK_YMAJOR 1 -#define URB_SIZE(intel) (IS_IGDNG(intel->device_id) ? 1024 : \ +#define SCAN_RASTER_ORDER 0 +#define SCAN_SPECIAL_ORDER 1 + +#define ENTROPY_CAVLD 0 +#define ENTROPY_CABAC 1 + +#define SLICE_TYPE_P 0 +#define SLICE_TYPE_B 1 +#define SLICE_TYPE_I 2 +#define SLICE_TYPE_SP 3 +#define SLICE_TYPE_SI 4 + +#define PRESENT_REF_LIST0 (1 << 0) +#define PRESENT_REF_LIST1 (1 << 1) +#define PRESENT_WEIGHT_OFFSET_L0 (1 << 2) +#define PRESENT_WEIGHT_OFFSET_L1 (1 << 3) + +#define RESIDUAL_DATA_OFFSET 48 + +#define PRESENT_NOMV 0 +#define PRESENT_NOWO 1 +#define PRESENT_MV_WO 3 + +#define SCOREBOARD_STALLING 0 +#define SCOREBOARD_NON_STALLING 1 + +#define URB_SIZE(intel) (IS_IRONLAKE(intel->device_id) ? 1024 : \ IS_G4X(intel->device_id) ? 384 : 256) #endif /* _I965_DEFINES_H_ */ diff --git a/i965_drv_video.c b/i965_drv_video.c index 6a60ecc..5033fe8 100644 --- a/i965_drv_video.c +++ b/i965_drv_video.c @@ -99,6 +99,9 @@ i965_QueryConfigProfiles(VADriverContextP ctx, profile_list[i++] = VAProfileMPEG2Simple; profile_list[i++] = VAProfileMPEG2Main; + profile_list[i++] = VAProfileH264Baseline; + profile_list[i++] = VAProfileH264Main; + profile_list[i++] = VAProfileH264High; /* If the assert fails then I965_MAX_PROFILES needs to be bigger */ assert(i <= I965_MAX_PROFILES); @@ -122,6 +125,13 @@ i965_QueryConfigEntrypoints(VADriverContextP ctx, entrypoint_list[0] = VAEntrypointVLD; break; + case VAProfileH264Baseline: + case VAProfileH264Main: + case VAProfileH264High: + *num_entrypoints = 1; + entrypoint_list[0] = VAEntrypointVLD; + break; + default: vaStatus = VA_STATUS_ERROR_UNSUPPORTED_PROFILE; *num_entrypoints = 0; @@ -217,6 +227,17 @@ i965_CreateConfig(VADriverContextP ctx, } break; + case VAProfileH264Baseline: + case VAProfileH264Main: + case VAProfileH264High: + if (VAEntrypointVLD == entrypoint) { + vaStatus = VA_STATUS_SUCCESS; + } else { + vaStatus = VA_STATUS_ERROR_UNSUPPORTED_ENTRYPOINT; + } + + break; + default: vaStatus = VA_STATUS_ERROR_UNSUPPORTED_PROFILE; break; @@ -305,6 +326,10 @@ i965_destroy_surface(struct object_heap *heap, struct object_base *obj) dri_bo_unreference(obj_surface->bo); obj_surface->bo = NULL; + dri_bo_unreference(obj_surface->direct_mv_wr_top_bo); + obj_surface->direct_mv_wr_top_bo = NULL; + dri_bo_unreference(obj_surface->direct_mv_wr_bottom_bo); + obj_surface->direct_mv_wr_bottom_bo = NULL; object_heap_free(heap, obj); } @@ -344,12 +369,17 @@ i965_CreateSurfaces(VADriverContextP ctx, "vaapi surface", obj_surface->size, 64); - assert(obj_surface->bo); - if (NULL == obj_surface->bo) { - vaStatus = VA_STATUS_ERROR_UNKNOWN; - break; - } + obj_surface->direct_mv_wr_top_bo = dri_bo_alloc(i965->intel.bufmgr, + "direct mv wr top", + 0x90000, + 64); + assert(obj_surface->direct_mv_wr_top_bo); + obj_surface->direct_mv_wr_bottom_bo = dri_bo_alloc(i965->intel.bufmgr, + "direct mv wr bottom", + 0x90000, + 64); + assert(obj_surface->direct_mv_wr_bottom_bo); } /* Error recovery */ @@ -881,6 +911,12 @@ i965_BeginPicture(VADriverContextP ctx, vaStatus = VA_STATUS_SUCCESS; break; + case VAProfileH264Baseline: + case VAProfileH264Main: + case VAProfileH264High: + vaStatus = VA_STATUS_SUCCESS; + break; + default: assert(0); vaStatus = VA_STATUS_ERROR_UNSUPPORTED_PROFILE; @@ -1014,6 +1050,7 @@ VAStatus i965_EndPicture(VADriverContextP ctx, VAContextID context) { struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_render_state *render_state = &i965->render_state; struct object_context *obj_context = CONTEXT(context); struct object_config *obj_config; VAContextID config; @@ -1026,6 +1063,18 @@ i965_EndPicture(VADriverContextP ctx, VAContextID context) config = obj_context->config_id; obj_config = CONFIG(config); assert(obj_config); + + switch (obj_config->profile) { + case VAProfileH264Baseline: + case VAProfileH264Main: + case VAProfileH264High: + render_state->interleaved_uv = 1; + break; + + default: + render_state->interleaved_uv = 0; + } + i965_media_decode_picture(ctx, obj_config->profile, &obj_context->decode_state); obj_context->decode_state.current_render_target = -1; obj_context->decode_state.num_slices = 0; @@ -1131,7 +1180,7 @@ i965_Init(VADriverContextP ctx) return VA_STATUS_ERROR_UNKNOWN; if (!IS_G4X(i965->intel.device_id) && - !IS_IGDNG(i965->intel.device_id)) + !IS_IRONLAKE(i965->intel.device_id)) return VA_STATUS_ERROR_UNKNOWN; if (i965_media_init(ctx) == False) diff --git a/i965_drv_video.h b/i965_drv_video.h index fa09869..605b386 100644 --- a/i965_drv_video.h +++ b/i965_drv_video.h @@ -97,6 +97,9 @@ struct object_surface int height; int size; dri_bo *bo; + /* FIXME: only for H.264 */ + dri_bo *direct_mv_wr_top_bo; + dri_bo *direct_mv_wr_bottom_bo; }; struct object_buffer diff --git a/i965_media.c b/i965_media.c index 82c5652..6a2971c 100644 --- a/i965_media.c +++ b/i965_media.c @@ -38,6 +38,7 @@ #include "i965_defines.h" #include "i965_media_mpeg2.h" +#include "i965_media_h264.h" #include "i965_media.h" #include "i965_drv_video.h" @@ -72,13 +73,21 @@ static void i965_media_state_base_address(VADriverContextP ctx) { struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; - if (IS_IGDNG(i965->intel.device_id)) { + if (IS_IRONLAKE(i965->intel.device_id)) { BEGIN_BATCH(ctx, 8); OUT_BATCH(ctx, CMD_STATE_BASE_ADDRESS | 6); OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); - OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + + if (media_state->indirect_object.bo) { + OUT_RELOC(ctx, media_state->indirect_object.bo, I915_GEM_DOMAIN_INSTRUCTION, 0, + media_state->indirect_object.offset | BASE_ADDRESS_MODIFY); + } else { + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + } + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); @@ -89,7 +98,14 @@ i965_media_state_base_address(VADriverContextP ctx) OUT_BATCH(ctx, CMD_STATE_BASE_ADDRESS | 4); OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); - OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + + if (media_state->indirect_object.bo) { + OUT_RELOC(ctx, media_state->indirect_object.bo, I915_GEM_DOMAIN_INSTRUCTION, 0, + media_state->indirect_object.offset | BASE_ADDRESS_MODIFY); + } else { + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); + } + OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); ADVANCE_BATCH(ctx); @@ -221,6 +237,12 @@ i965_media_decode_init(VADriverContextP ctx, VAProfile profile) case VAProfileMPEG2Main: i965_media_mpeg2_decode_init(ctx); break; + + case VAProfileH264Baseline: + case VAProfileH264Main: + case VAProfileH264High: + i965_media_h264_decode_init(ctx); + break; default: assert(0); @@ -237,8 +259,8 @@ i965_media_decode_picture(VADriverContextP ctx, struct i965_media_state *media_state = &i965->media_state; i965_media_decode_init(ctx, profile); - assert(media_state->states_setup); - media_state->states_setup(ctx, decode_state); + assert(media_state->media_states_setup); + media_state->media_states_setup(ctx, decode_state); i965_media_pipeline_setup(ctx, decode_state); intel_batchbuffer_flush(ctx); } @@ -247,6 +269,7 @@ Bool i965_media_init(VADriverContextP ctx) { i965_media_mpeg2_init(ctx); + i965_media_h264_init(ctx); return True; } @@ -277,7 +300,11 @@ i965_media_terminate(VADriverContextP ctx) dri_bo_unreference(media_state->curbe.bo); media_state->curbe.bo = NULL; + dri_bo_unreference(media_state->indirect_object.bo); + media_state->indirect_object.bo = NULL; + i965_media_mpeg2_ternimate(ctx); + i965_media_h264_ternimate(ctx); return True; } diff --git a/i965_media.h b/i965_media.h index e207f4e..b745777 100644 --- a/i965_media.h +++ b/i965_media.h @@ -38,7 +38,7 @@ #include "i965_structs.h" #define MAX_INTERFACE_DESC 16 -#define MAX_MEDIA_SURFACES 32 +#define MAX_MEDIA_SURFACES 34 #define MPEG_TOP_FIELD 1 #define MPEG_BOTTOM_FIELD 2 @@ -83,6 +83,11 @@ struct i965_media_state } curbe; struct { + dri_bo *bo; + unsigned long offset; + } indirect_object; + + struct { unsigned int vfe_start; unsigned int cs_start; @@ -93,7 +98,8 @@ struct i965_media_state unsigned int size_cs_entry; } urb; - void (*states_setup)(VADriverContextP ctx, struct decode_state *decode_state); + void *private_context; + void (*media_states_setup)(VADriverContextP ctx, struct decode_state *decode_state); void (*media_objects)(VADriverContextP ctx, struct decode_state *decode_state); }; diff --git a/i965_media_h264.c b/i965_media_h264.c new file mode 100644 index 0000000..c1b5626 --- /dev/null +++ b/i965_media_h264.c @@ -0,0 +1,882 @@ +#include <stdlib.h> +#include <stdio.h> +#include <string.h> +#include <assert.h> + +#include "va_backend.h" + +#include "intel_batchbuffer.h" +#include "intel_driver.h" + +#include "i965_defines.h" +#include "i965_drv_video.h" +#include "i965_media.h" +#include "i965_media_h264.h" + +enum { + INTRA_16X16 = 0, + INTRA_8X8, + INTRA_4X4, + INTRA_PCM, + FRAMEMB_MOTION, + FIELDMB_MOTION, + MBAFF_MOTION, +}; + +struct intra_kernel_header +{ + /* R1.0 */ + unsigned char intra_4x4_luma_mode_0_offset; + unsigned char intra_4x4_luma_mode_1_offset; + unsigned char intra_4x4_luma_mode_2_offset; + unsigned char intra_4x4_luma_mode_3_offset; + /* R1.1 */ + unsigned char intra_4x4_luma_mode_4_offset; + unsigned char intra_4x4_luma_mode_5_offset; + unsigned char intra_4x4_luma_mode_6_offset; + unsigned char intra_4x4_luma_mode_7_offset; + /* R1.2 */ + unsigned char intra_4x4_luma_mode_8_offset; + unsigned char pad0; + unsigned short top_reference_offset; + /* R1.3 */ + unsigned char intra_8x8_luma_mode_0_offset; + unsigned char intra_8x8_luma_mode_1_offset; + unsigned char intra_8x8_luma_mode_2_offset; + unsigned char intra_8x8_luma_mode_3_offset; + /* R1.4 */ + unsigned char intra_8x8_luma_mode_4_offset; + unsigned char intra_8x8_luma_mode_5_offset; + unsigned char intra_8x8_luma_mode_6_offset; + unsigned char intra_8x8_luma_mode_7_offset; + /* R1.5 */ + unsigned char intra_8x8_luma_mode_8_offset; + unsigned char pad1; + unsigned short const_reverse_data_transfer_intra_8x8; + /* R1.6 */ + unsigned char intra_16x16_luma_mode_0_offset; + unsigned char intra_16x16_luma_mode_1_offset; + unsigned char intra_16x16_luma_mode_2_offset; + unsigned char intra_16x16_luma_mode_3_offset; + /* R1.7 */ + unsigned char intra_chroma_mode_0_offset; + unsigned char intra_chroma_mode_1_offset; + unsigned char intra_chroma_mode_2_offset; + unsigned char intra_chroma_mode_3_offset; + /* R2.0 */ + unsigned int const_intra_16x16_plane_0; + /* R2.1 */ + unsigned int const_intra_16x16_chroma_plane_0; + /* R2.2 */ + unsigned int const_intra_16x16_chroma_plane_1; + /* R2.3 */ + unsigned int const_intra_16x16_plane_1; + /* R2.4 */ + unsigned int left_shift_count_reverse_dw_ordering; + /* R2.5 */ + unsigned int const_reverse_data_transfer_intra_4x4; + /* R2.6 */ + unsigned int intra_4x4_pred_mode_offset; +}; + +struct inter_kernel_header +{ + unsigned short weight_offset; + unsigned char weight_offset_flag; + unsigned char pad0; +}; + +#include "shaders/h264/mc/export.inc" +static unsigned long avc_mc_kernel_offset_gen4[] = { + INTRA_16x16_IP, + INTRA_8x8_IP, + INTRA_4x4_IP, + INTRA_PCM_IP, + FRAME_MB_IP, + FIELD_MB_IP, + MBAFF_MB_IP +}; + +#define INST_UNIT_GEN4 16 +struct intra_kernel_header intra_kernel_header_gen4 = { + 0 / INST_UNIT_GEN4, + (INTRA_4X4_HORIZONTAL_IP - INTRA_4X4_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_4X4_DC_IP - INTRA_4X4_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_4X4_DIAG_DOWN_LEFT_IP - INTRA_4X4_VERTICAL_IP) / INST_UNIT_GEN4, + + (INTRA_4X4_DIAG_DOWN_RIGHT_IP - INTRA_4X4_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_4X4_VERT_RIGHT_IP - INTRA_4X4_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_4X4_HOR_DOWN_IP - INTRA_4X4_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_4X4_VERT_LEFT_IP - INTRA_4X4_VERTICAL_IP) / INST_UNIT_GEN4, + + (INTRA_4X4_HOR_UP_IP - INTRA_4X4_VERTICAL_IP) / INST_UNIT_GEN4, + 0, + 0xFFFC, + + 0 / INST_UNIT_GEN4, + (INTRA_8X8_HORIZONTAL_IP - INTRA_8X8_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_8X8_DC_IP - INTRA_8X8_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_8X8_DIAG_DOWN_LEFT_IP - INTRA_8X8_VERTICAL_IP) / INST_UNIT_GEN4, + + (INTRA_8X8_DIAG_DOWN_RIGHT_IP - INTRA_8X8_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_8X8_VERT_RIGHT_IP - INTRA_8X8_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_8X8_HOR_DOWN_IP - INTRA_8X8_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_8X8_VERT_LEFT_IP - INTRA_8X8_VERTICAL_IP) / INST_UNIT_GEN4, + + (INTRA_8X8_HOR_UP_IP - INTRA_8X8_VERTICAL_IP) / INST_UNIT_GEN4, + 0, + 0x0001, + + 0 / INST_UNIT_GEN4, + (INTRA_16x16_HORIZONTAL_IP - INTRA_16x16_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_16x16_DC_IP - INTRA_16x16_VERTICAL_IP) / INST_UNIT_GEN4, + (INTRA_16x16_PLANE_IP - INTRA_16x16_VERTICAL_IP) / INST_UNIT_GEN4, + + 0 / INST_UNIT_GEN4, + (INTRA_CHROMA_HORIZONTAL_IP - INTRA_CHROMA_DC_IP) / INST_UNIT_GEN4, + (INTRA_CHROMA_VERTICAL_IP - INTRA_CHROMA_DC_IP) / INST_UNIT_GEN4, + (INTRA_Chroma_PLANE_IP - INTRA_CHROMA_DC_IP) / INST_UNIT_GEN4, + + 0xFCFBFAF9, + + 0x00FFFEFD, + + 0x04030201, + + 0x08070605, + + 0x18100800, + + 0x00020406, + + (intra_Pred_4x4_Y_IP - ADD_ERROR_SB3_IP) / INST_UNIT_GEN4 * 0x1000000 + + (intra_Pred_4x4_Y_IP - ADD_ERROR_SB2_IP) / INST_UNIT_GEN4 * 0x10000 + + (intra_Pred_4x4_Y_IP - ADD_ERROR_SB1_IP) / INST_UNIT_GEN4 * 0x100 + + (intra_Pred_4x4_Y_IP - ADD_ERROR_SB0_IP) / INST_UNIT_GEN4 +}; + +static uint32_t h264_avc_combined_gen4[][4] = { +#include "shaders/h264/mc/avc_mc.g4b" +}; + +static uint32_t h264_avc_null_gen4[][4] = { +#include "shaders/h264/mc/null.g4b" +}; + +static struct media_kernel h264_avc_kernels_gen4[] = { + { + "AVC combined kernel", + H264_AVC_COMBINED, + h264_avc_combined_gen4, + sizeof(h264_avc_combined_gen4), + NULL + }, + + { + "NULL kernel", + H264_AVC_NULL, + h264_avc_null_gen4, + sizeof(h264_avc_null_gen4), + NULL + } +}; + +/* On Ironlake */ +#include "shaders/h264/mc/export.inc.gen5" +static unsigned long avc_mc_kernel_offset_gen5[] = { + INTRA_16x16_IP_GEN5, + INTRA_8x8_IP_GEN5, + INTRA_4x4_IP_GEN5, + INTRA_PCM_IP_GEN5, + FRAME_MB_IP_GEN5, + FIELD_MB_IP_GEN5, + MBAFF_MB_IP_GEN5 +}; + +#define INST_UNIT_GEN5 8 + +struct intra_kernel_header intra_kernel_header_gen5 = { + 0 / INST_UNIT_GEN5, + (INTRA_4X4_HORIZONTAL_IP_GEN5 - INTRA_4X4_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_4X4_DC_IP_GEN5 - INTRA_4X4_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_4X4_DIAG_DOWN_LEFT_IP_GEN5 - INTRA_4X4_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + + (INTRA_4X4_DIAG_DOWN_RIGHT_IP_GEN5 - INTRA_4X4_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_4X4_VERT_RIGHT_IP_GEN5 - INTRA_4X4_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_4X4_HOR_DOWN_IP_GEN5 - INTRA_4X4_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_4X4_VERT_LEFT_IP_GEN5 - INTRA_4X4_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + + (INTRA_4X4_HOR_UP_IP_GEN5 - INTRA_4X4_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + 0, + 0xFFFC, + + 0 / INST_UNIT_GEN5, + (INTRA_8X8_HORIZONTAL_IP_GEN5 - INTRA_8X8_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_8X8_DC_IP_GEN5 - INTRA_8X8_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_8X8_DIAG_DOWN_LEFT_IP_GEN5 - INTRA_8X8_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + + (INTRA_8X8_DIAG_DOWN_RIGHT_IP_GEN5 - INTRA_8X8_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_8X8_VERT_RIGHT_IP_GEN5 - INTRA_8X8_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_8X8_HOR_DOWN_IP_GEN5 - INTRA_8X8_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_8X8_VERT_LEFT_IP_GEN5 - INTRA_8X8_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + + (INTRA_8X8_HOR_UP_IP_GEN5 - INTRA_8X8_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + 0, + 0x0001, + + 0 / INST_UNIT_GEN5, + (INTRA_16x16_HORIZONTAL_IP_GEN5 - INTRA_16x16_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_16x16_DC_IP_GEN5 - INTRA_16x16_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_16x16_PLANE_IP_GEN5 - INTRA_16x16_VERTICAL_IP_GEN5) / INST_UNIT_GEN5, + + 0 / INST_UNIT_GEN5, + (INTRA_CHROMA_HORIZONTAL_IP_GEN5 - INTRA_CHROMA_DC_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_CHROMA_VERTICAL_IP_GEN5 - INTRA_CHROMA_DC_IP_GEN5) / INST_UNIT_GEN5, + (INTRA_Chroma_PLANE_IP_GEN5 - INTRA_CHROMA_DC_IP_GEN5) / INST_UNIT_GEN5, + + 0xFCFBFAF9, + + 0x00FFFEFD, + + 0x04030201, + + 0x08070605, + + 0x18100800, + + 0x00020406, + + (intra_Pred_4x4_Y_IP_GEN5 - ADD_ERROR_SB3_IP_GEN5) / INST_UNIT_GEN5 * 0x1000000 + + (intra_Pred_4x4_Y_IP_GEN5 - ADD_ERROR_SB2_IP_GEN5) / INST_UNIT_GEN5 * 0x10000 + + (intra_Pred_4x4_Y_IP_GEN5 - ADD_ERROR_SB1_IP_GEN5) / INST_UNIT_GEN5 * 0x100 + + (intra_Pred_4x4_Y_IP_GEN5 - ADD_ERROR_SB0_IP_GEN5) / INST_UNIT_GEN5 +}; + +static uint32_t h264_avc_combined_gen5[][4] = { +#include "shaders/h264/mc/avc_mc.g4b.gen5" +}; + +static uint32_t h264_avc_null_gen5[][4] = { +#include "shaders/h264/mc/null.g4b.gen5" +}; + +static struct media_kernel h264_avc_kernels_gen5[] = { + { + "AVC combined kernel", + H264_AVC_COMBINED, + h264_avc_combined_gen5, + sizeof(h264_avc_combined_gen5), + NULL + }, + + { + "NULL kernel", + H264_AVC_NULL, + h264_avc_null_gen5, + sizeof(h264_avc_null_gen5), + NULL + } +}; + +#define NUM_H264_AVC_KERNELS (sizeof(h264_avc_kernels_gen4) / sizeof(h264_avc_kernels_gen4[0])) +struct media_kernel *h264_avc_kernels = NULL; + +#define NUM_AVC_MC_INTERFACES (sizeof(avc_mc_kernel_offset_gen4) / sizeof(avc_mc_kernel_offset_gen4[0])) +static unsigned long *avc_mc_kernel_offset = NULL; + +static struct intra_kernel_header *intra_kernel_header = NULL; + +static void +i965_media_h264_surface_state(VADriverContextP ctx, + int index, + struct object_surface *obj_surface, + unsigned long offset, + int w, int h, int pitch, + Bool is_dst, + int vert_line_stride, + int vert_line_stride_ofs, + int format) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_surface_state *ss; + dri_bo *bo; + uint32_t write_domain, read_domain; + + bo = dri_bo_alloc(i965->intel.bufmgr, + "surface state", + sizeof(struct i965_surface_state), 32); + assert(bo); + dri_bo_map(bo, 1); + assert(bo->virtual); + ss = bo->virtual; + memset(ss, 0, sizeof(*ss)); + ss->ss0.surface_type = I965_SURFACE_2D; + ss->ss0.surface_format = format; + ss->ss0.vert_line_stride = vert_line_stride; + ss->ss0.vert_line_stride_ofs = vert_line_stride_ofs; + ss->ss1.base_addr = obj_surface->bo->offset + offset; + ss->ss2.width = w - 1; + ss->ss2.height = h - 1; + ss->ss3.pitch = pitch - 1; + + if (is_dst) { + write_domain = I915_GEM_DOMAIN_RENDER; + read_domain = I915_GEM_DOMAIN_RENDER; + } else { + write_domain = 0; + read_domain = I915_GEM_DOMAIN_SAMPLER; + } + + dri_bo_emit_reloc(bo, + read_domain, write_domain, + offset, + offsetof(struct i965_surface_state, ss1), + obj_surface->bo); + dri_bo_unmap(bo); + + assert(index < MAX_MEDIA_SURFACES); + media_state->surface_state[index].bo = bo; +} + +static void +i965_media_h264_surfaces_setup(VADriverContextP ctx, + struct decode_state *decode_state) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct object_surface *obj_surface; + VAPictureParameterBufferH264 *pic_param; + VAPictureH264 *va_pic; + int i, w, h; + int field_picture; + + assert(decode_state->pic_param && decode_state->pic_param->buffer); + pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; + + /* Target Picture */ + va_pic = &pic_param->CurrPic; + assert(!(va_pic->flags & VA_PICTURE_H264_INVALID)); + obj_surface = SURFACE(va_pic->picture_id); + assert(obj_surface); + w = obj_surface->width; + h = obj_surface->height; + field_picture = !!(va_pic->flags & (VA_PICTURE_H264_TOP_FIELD | VA_PICTURE_H264_BOTTOM_FIELD)); + i965_media_h264_surface_state(ctx, 0, obj_surface, + 0, w / 4, h / (1 + field_picture), w, + 1, + field_picture, + !!(va_pic->flags & VA_PICTURE_H264_BOTTOM_FIELD), + I965_SURFACEFORMAT_R8_SINT); /* Y */ + i965_media_h264_surface_state(ctx, 1, obj_surface, + w * h, w / 4, h / 2 / (1 + field_picture), w, + 1, + field_picture, + !!(va_pic->flags & VA_PICTURE_H264_BOTTOM_FIELD), + I965_SURFACEFORMAT_R8G8_SINT); /* INTERLEAVED U/V */ + + /* Reference Pictures */ + for (i = 0; i < 16; i++) { + va_pic = &pic_param->ReferenceFrames[i]; + + if (!(va_pic->flags & VA_PICTURE_H264_INVALID)) { + obj_surface = SURFACE(va_pic->picture_id); + assert(obj_surface); + w = obj_surface->width; + h = obj_surface->height; + field_picture = !!(va_pic->flags & (VA_PICTURE_H264_TOP_FIELD | VA_PICTURE_H264_BOTTOM_FIELD)); + i965_media_h264_surface_state(ctx, 2 + i, obj_surface, + 0, w / 4, h / (1 + field_picture), w, + 0, + field_picture, + !!(va_pic->flags & VA_PICTURE_H264_BOTTOM_FIELD), + I965_SURFACEFORMAT_R8_SINT); /* Y */ + i965_media_h264_surface_state(ctx, 18 + i, obj_surface, + w * h, w / 4, h / 2 / (1 + field_picture), w, + 0, + field_picture, + !!(va_pic->flags & VA_PICTURE_H264_BOTTOM_FIELD), + I965_SURFACEFORMAT_R8G8_SINT); /* INTERLEAVED U/V */ + } + } +} + +static void +i965_media_h264_binding_table(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + int i; + unsigned int *binding_table; + dri_bo *bo = media_state->binding_table.bo; + + dri_bo_map(bo, 1); + assert(bo->virtual); + binding_table = bo->virtual; + memset(binding_table, 0, bo->size); + + for (i = 0; i < MAX_MEDIA_SURFACES; i++) { + if (media_state->surface_state[i].bo) { + binding_table[i] = media_state->surface_state[i].bo->offset; + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0, + i * sizeof(*binding_table), + media_state->surface_state[i].bo); + } + } + + dri_bo_unmap(media_state->binding_table.bo); +} + +static void +i965_media_h264_interface_descriptor_remap_table(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_interface_descriptor *desc; + int i; + dri_bo *bo; + + bo = media_state->idrt.bo; + dri_bo_map(bo, 1); + assert(bo->virtual); + desc = bo->virtual; + + for (i = 0; i < NUM_AVC_MC_INTERFACES; i++) { + int kernel_offset = avc_mc_kernel_offset[i]; + memset(desc, 0, sizeof(*desc)); + desc->desc0.grf_reg_blocks = 7; + desc->desc0.kernel_start_pointer = (h264_avc_kernels[H264_AVC_COMBINED].bo->offset + kernel_offset) >> 6; /* reloc */ + desc->desc1.const_urb_entry_read_offset = 0; + desc->desc1.const_urb_entry_read_len = 2; + desc->desc3.binding_table_entry_count = 0; + desc->desc3.binding_table_pointer = + media_state->binding_table.bo->offset >> 5; /*reloc */ + + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + desc->desc0.grf_reg_blocks + kernel_offset, + i * sizeof(*desc) + offsetof(struct i965_interface_descriptor, desc0), + h264_avc_kernels[H264_AVC_COMBINED].bo); + + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + desc->desc3.binding_table_entry_count, + i * sizeof(*desc) + offsetof(struct i965_interface_descriptor, desc3), + media_state->binding_table.bo); + desc++; + } + + dri_bo_unmap(bo); +} + +static void +i965_media_h264_vfe_state(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_vfe_state *vfe_state; + dri_bo *bo; + + bo = media_state->vfe_state.bo; + dri_bo_map(bo, 1); + assert(bo->virtual); + vfe_state = bo->virtual; + memset(vfe_state, 0, sizeof(*vfe_state)); + vfe_state->vfe0.extend_vfe_state_present = 1; + vfe_state->vfe1.max_threads = media_state->urb.num_vfe_entries - 1; + vfe_state->vfe1.urb_entry_alloc_size = media_state->urb.size_vfe_entry - 1; + vfe_state->vfe1.num_urb_entries = media_state->urb.num_vfe_entries; + vfe_state->vfe1.vfe_mode = VFE_AVC_IT_MODE; + vfe_state->vfe1.children_present = 0; + vfe_state->vfe2.interface_descriptor_base = + media_state->idrt.bo->offset >> 4; /* reloc */ + dri_bo_emit_reloc(bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + 0, + offsetof(struct i965_vfe_state, vfe2), + media_state->idrt.bo); + dri_bo_unmap(bo); +} + +static void +i965_media_h264_vfe_state_extension(VADriverContextP ctx, + struct decode_state *decode_state) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context; + struct i965_vfe_state_ex *vfe_state_ex; + VAPictureParameterBufferH264 *pic_param; + VASliceParameterBufferH264 *slice_param; + int mbaff_frame_flag; + + assert(media_state->private_context); + i965_h264_context = (struct i965_h264_context *)media_state->private_context; + + assert(decode_state->pic_param && decode_state->pic_param->buffer); + pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; + + assert(decode_state->slice_param && decode_state->slice_param->buffer); + slice_param = (VASliceParameterBufferH264 *)decode_state->slice_param->buffer; + + mbaff_frame_flag = (pic_param->seq_fields.bits.mb_adaptive_frame_field_flag && + !pic_param->pic_fields.bits.field_pic_flag); + + assert(media_state->extended_state.bo); + dri_bo_map(media_state->extended_state.bo, 1); + assert(media_state->extended_state.bo->virtual); + vfe_state_ex = media_state->extended_state.bo->virtual; + memset(vfe_state_ex, 0, sizeof(*vfe_state_ex)); + + /* + * Indirect data buffer: + * -------------------------------------------------------- + * | Motion Vectors | Weight/Offset data | Residual data | + * -------------------------------------------------------- + * R4-R7: Motion Vectors + * R8-R9: Weight/Offset + * R10-R33: Residual data + */ + vfe_state_ex->vfex1.avc.residual_data_fix_offset_flag = !!RESIDUAL_DATA_OFFSET; + vfe_state_ex->vfex1.avc.residual_data_offset = RESIDUAL_DATA_OFFSET; + + if (slice_param->slice_type == SLICE_TYPE_I) + vfe_state_ex->vfex1.avc.sub_field_present_flag = PRESENT_NOMV; /* NoMV */ + else + vfe_state_ex->vfex1.avc.sub_field_present_flag = PRESENT_MV_WO; /* Both MV and W/O */ + + if (vfe_state_ex->vfex1.avc.sub_field_present_flag == 0) { + vfe_state_ex->vfex1.avc.weight_grf_offset = 0; + vfe_state_ex->vfex1.avc.residual_grf_offset = 0; + } else { + vfe_state_ex->vfex1.avc.weight_grf_offset = 4; + vfe_state_ex->vfex1.avc.residual_grf_offset = 6; + } + + if (!pic_param->pic_fields.bits.field_pic_flag) { + if (mbaff_frame_flag) { + vfe_state_ex->remap_table0.remap_index_0 = INTRA_16X16; + vfe_state_ex->remap_table0.remap_index_1 = INTRA_8X8; + vfe_state_ex->remap_table0.remap_index_2 = INTRA_4X4; + vfe_state_ex->remap_table0.remap_index_3 = INTRA_PCM; + vfe_state_ex->remap_table0.remap_index_4 = MBAFF_MOTION; + vfe_state_ex->remap_table0.remap_index_5 = MBAFF_MOTION; + vfe_state_ex->remap_table0.remap_index_6 = MBAFF_MOTION; + vfe_state_ex->remap_table0.remap_index_7 = MBAFF_MOTION; + + vfe_state_ex->remap_table1.remap_index_8 = MBAFF_MOTION; + vfe_state_ex->remap_table1.remap_index_9 = MBAFF_MOTION; + vfe_state_ex->remap_table1.remap_index_10 = MBAFF_MOTION; + vfe_state_ex->remap_table1.remap_index_11 = MBAFF_MOTION; + vfe_state_ex->remap_table1.remap_index_12 = MBAFF_MOTION; + vfe_state_ex->remap_table1.remap_index_13 = MBAFF_MOTION; + vfe_state_ex->remap_table1.remap_index_14 = MBAFF_MOTION; + vfe_state_ex->remap_table1.remap_index_15 = MBAFF_MOTION; + } else { + vfe_state_ex->remap_table0.remap_index_0 = INTRA_16X16; + vfe_state_ex->remap_table0.remap_index_1 = INTRA_8X8; + vfe_state_ex->remap_table0.remap_index_2 = INTRA_4X4; + vfe_state_ex->remap_table0.remap_index_3 = INTRA_PCM; + vfe_state_ex->remap_table0.remap_index_4 = FRAMEMB_MOTION; + vfe_state_ex->remap_table0.remap_index_5 = FRAMEMB_MOTION; + vfe_state_ex->remap_table0.remap_index_6 = FRAMEMB_MOTION; + vfe_state_ex->remap_table0.remap_index_7 = FRAMEMB_MOTION; + + vfe_state_ex->remap_table1.remap_index_8 = FRAMEMB_MOTION; + vfe_state_ex->remap_table1.remap_index_9 = FRAMEMB_MOTION; + vfe_state_ex->remap_table1.remap_index_10 = FRAMEMB_MOTION; + vfe_state_ex->remap_table1.remap_index_11 = FRAMEMB_MOTION; + vfe_state_ex->remap_table1.remap_index_12 = FRAMEMB_MOTION; + vfe_state_ex->remap_table1.remap_index_13 = FRAMEMB_MOTION; + vfe_state_ex->remap_table1.remap_index_14 = FRAMEMB_MOTION; + vfe_state_ex->remap_table1.remap_index_15 = FRAMEMB_MOTION; + } + } else { + vfe_state_ex->remap_table0.remap_index_0 = INTRA_16X16; + vfe_state_ex->remap_table0.remap_index_1 = INTRA_8X8; + vfe_state_ex->remap_table0.remap_index_2 = INTRA_4X4; + vfe_state_ex->remap_table0.remap_index_3 = INTRA_PCM; + vfe_state_ex->remap_table0.remap_index_4 = FIELDMB_MOTION; + vfe_state_ex->remap_table0.remap_index_5 = FIELDMB_MOTION; + vfe_state_ex->remap_table0.remap_index_6 = FIELDMB_MOTION; + vfe_state_ex->remap_table0.remap_index_7 = FIELDMB_MOTION; + + vfe_state_ex->remap_table1.remap_index_8 = FIELDMB_MOTION; + vfe_state_ex->remap_table1.remap_index_9 = FIELDMB_MOTION; + vfe_state_ex->remap_table1.remap_index_10 = FIELDMB_MOTION; + vfe_state_ex->remap_table1.remap_index_11 = FIELDMB_MOTION; + vfe_state_ex->remap_table1.remap_index_12 = FIELDMB_MOTION; + vfe_state_ex->remap_table1.remap_index_13 = FIELDMB_MOTION; + vfe_state_ex->remap_table1.remap_index_14 = FIELDMB_MOTION; + vfe_state_ex->remap_table1.remap_index_15 = FIELDMB_MOTION; + } + + if (i965_h264_context->use_avc_hw_scoreboard) { + vfe_state_ex->scoreboard0.enable = 1; + vfe_state_ex->scoreboard0.type = SCOREBOARD_STALLING; + vfe_state_ex->scoreboard0.mask = 0xff; + + vfe_state_ex->scoreboard1.delta_x0 = -1; + vfe_state_ex->scoreboard1.delta_y0 = 0; + vfe_state_ex->scoreboard1.delta_x1 = 0; + vfe_state_ex->scoreboard1.delta_y1 = -1; + vfe_state_ex->scoreboard1.delta_x2 = 1; + vfe_state_ex->scoreboard1.delta_y2 = -1; + vfe_state_ex->scoreboard1.delta_x3 = -1; + vfe_state_ex->scoreboard1.delta_y3 = -1; + + vfe_state_ex->scoreboard2.delta_x4 = -1; + vfe_state_ex->scoreboard2.delta_y4 = 1; + vfe_state_ex->scoreboard2.delta_x5 = 0; + vfe_state_ex->scoreboard2.delta_y5 = -2; + vfe_state_ex->scoreboard2.delta_x6 = 1; + vfe_state_ex->scoreboard2.delta_y6 = -2; + vfe_state_ex->scoreboard2.delta_x7 = -1; + vfe_state_ex->scoreboard2.delta_y7 = -2; + } + + dri_bo_unmap(media_state->extended_state.bo); +} + +static void +i965_media_h264_upload_constants(VADriverContextP ctx, struct decode_state *decode_state) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context; + unsigned char *constant_buffer; + VASliceParameterBufferH264 *slice_param; + + assert(media_state->private_context); + i965_h264_context = (struct i965_h264_context *)media_state->private_context; + + assert(decode_state->slice_param && decode_state->slice_param->buffer); + slice_param = (VASliceParameterBufferH264 *)decode_state->slice_param->buffer; + + dri_bo_map(media_state->curbe.bo, 1); + assert(media_state->curbe.bo->virtual); + constant_buffer = media_state->curbe.bo->virtual; + + /* HW solution for W=128 */ + if (i965_h264_context->use_hw_w128) { + memcpy(constant_buffer, intra_kernel_header, sizeof(*intra_kernel_header)); + } else { + if (slice_param->slice_type == SLICE_TYPE_I) { + memcpy(constant_buffer, intra_kernel_header, sizeof(*intra_kernel_header)); + } else { + /* FIXME: Need to upload CURBE data to inter kernel interface + * to support weighted prediction work-around + */ + *(short *)constant_buffer = i965_h264_context->weight128_offset0; + constant_buffer += 2; + *(char *)constant_buffer = i965_h264_context->weight128_offset0_flag; + constant_buffer++; + *constant_buffer = 0; + } + } + + dri_bo_unmap(media_state->curbe.bo); +} + +static void +i965_media_h264_states_setup(VADriverContextP ctx, struct decode_state *decode_state) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context; + VAPictureParameterBufferH264 *pic_param; + unsigned int *object_command; + + assert(media_state->private_context); + i965_h264_context = (struct i965_h264_context *)media_state->private_context; + assert(decode_state->pic_param && decode_state->pic_param->buffer); + pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; + + i965_h264_context->picture.width_in_mbs = ((pic_param->picture_width_in_mbs_minus1 + 1) & 0xff); + i965_h264_context->picture.height_in_mbs = ((pic_param->picture_height_in_mbs_minus1 + 1) & 0xff) / + (1 + !!pic_param->pic_fields.bits.field_pic_flag); /* picture height */ + i965_h264_context->picture.mbaff_frame_flag = (pic_param->seq_fields.bits.mb_adaptive_frame_field_flag && + !pic_param->pic_fields.bits.field_pic_flag); + + i965_h264_context->avc_it_command_mb_info.mbs = (i965_h264_context->picture.width_in_mbs * + i965_h264_context->picture.height_in_mbs); + + i965_avc_bsd_pipeline(ctx, decode_state); + + dri_bo_map(i965_h264_context->avc_it_command_mb_info.bo, True); + assert(i965_h264_context->avc_it_command_mb_info.bo->virtual); + object_command = i965_h264_context->avc_it_command_mb_info.bo->virtual; + memset(object_command, 0, i965_h264_context->avc_it_command_mb_info.mbs * i965_h264_context->use_avc_hw_scoreboard * MB_CMD_IN_BYTES); + object_command += i965_h264_context->avc_it_command_mb_info.mbs * (1 + i965_h264_context->use_avc_hw_scoreboard) * MB_CMD_IN_DWS; + *object_command = MI_BATCH_BUFFER_END; + dri_bo_unmap(i965_h264_context->avc_it_command_mb_info.bo); + + i965_avc_hw_scoreboard(ctx, decode_state); + + i965_media_h264_surfaces_setup(ctx, decode_state); + i965_media_h264_binding_table(ctx); + i965_media_h264_interface_descriptor_remap_table(ctx); + i965_media_h264_vfe_state_extension(ctx, decode_state); + i965_media_h264_vfe_state(ctx); + i965_media_h264_upload_constants(ctx, decode_state); +} + +static void +i965_media_h264_objects(VADriverContextP ctx, struct decode_state *decode_state) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context; + + assert(media_state->private_context); + i965_h264_context = (struct i965_h264_context *)media_state->private_context; + + BEGIN_BATCH(ctx, 2); + OUT_BATCH(ctx, MI_BATCH_BUFFER_START | (2 << 6)); + OUT_RELOC(ctx, i965_h264_context->avc_it_command_mb_info.bo, + I915_GEM_DOMAIN_COMMAND, 0, + 0); + ADVANCE_BATCH(ctx); +} + +void +i965_media_h264_decode_init(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context; + dri_bo *bo; + + assert(media_state->private_context); + i965_h264_context = media_state->private_context; + + dri_bo_unreference(i965_h264_context->avc_it_command_mb_info.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "avc it command mb info", + 0x80000 * (1 + i965_h264_context->use_avc_hw_scoreboard), /* at least 522240 bytes */ + 0x1000); + assert(bo); + i965_h264_context->avc_it_command_mb_info.bo = bo; + + dri_bo_unreference(i965_h264_context->avc_it_data.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "avc it data", + 0x1000000, /* at least 16711680 bytes */ + 4096); + assert(bo); + i965_h264_context->avc_it_data.bo = bo; + i965_h264_context->avc_it_data.write_offset = 0; + dri_bo_unreference(media_state->indirect_object.bo); + media_state->indirect_object.bo = bo; + dri_bo_reference(media_state->indirect_object.bo); + media_state->indirect_object.offset = i965_h264_context->avc_it_data.write_offset; + + /* bsd pipeline */ + i965_avc_bsd_decode_init(ctx); + + /* HW scoreboard */ + i965_avc_hw_scoreboard_decode_init(ctx); + + /* for Media pipeline */ + media_state->extended_state.enabled = 1; + dri_bo_unreference(media_state->extended_state.bo); + bo = dri_bo_alloc(i965->intel.bufmgr, + "extened vfe state", + sizeof(struct i965_vfe_state_ex), 32); + assert(bo); + media_state->extended_state.bo = bo; + + /* URB */ + if (IS_IRONLAKE(i965->intel.device_id)) { + media_state->urb.num_vfe_entries = 63; + } else { + media_state->urb.num_vfe_entries = 23; + } + + media_state->urb.size_vfe_entry = 16; + + media_state->urb.num_cs_entries = 1; + media_state->urb.size_cs_entry = 1; + + media_state->urb.vfe_start = 0; + media_state->urb.cs_start = media_state->urb.vfe_start + + media_state->urb.num_vfe_entries * media_state->urb.size_vfe_entry; + assert(media_state->urb.cs_start + + media_state->urb.num_cs_entries * media_state->urb.size_cs_entry <= URB_SIZE((&i965->intel))); + + /* hook functions */ + media_state->media_states_setup = i965_media_h264_states_setup; + media_state->media_objects = i965_media_h264_objects; +} + +Bool +i965_media_h264_init(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context; + int i; + + i965_h264_context = calloc(1, sizeof(struct i965_h264_context)); + + /* kernel */ + assert(NUM_H264_AVC_KERNELS == (sizeof(h264_avc_kernels_gen5) / + sizeof(h264_avc_kernels_gen5[0]))); + assert(NUM_AVC_MC_INTERFACES == (sizeof(avc_mc_kernel_offset_gen5) / + sizeof(avc_mc_kernel_offset_gen5[0]))); + + if (IS_IRONLAKE(i965->intel.device_id)) { + h264_avc_kernels = h264_avc_kernels_gen5; + avc_mc_kernel_offset = avc_mc_kernel_offset_gen5; + intra_kernel_header = &intra_kernel_header_gen5; + i965_h264_context->use_avc_hw_scoreboard = 1; + i965_h264_context->use_hw_w128 = 1; + } else { + h264_avc_kernels = h264_avc_kernels_gen4; + avc_mc_kernel_offset = avc_mc_kernel_offset_gen4; + intra_kernel_header = &intra_kernel_header_gen4; + i965_h264_context->use_avc_hw_scoreboard = 0; + i965_h264_context->use_hw_w128 = 0; + } + + for (i = 0; i < NUM_H264_AVC_KERNELS; i++) { + struct media_kernel *kernel = &h264_avc_kernels[i]; + kernel->bo = dri_bo_alloc(i965->intel.bufmgr, + kernel->name, + kernel->size, 64); + assert(kernel->bo); + dri_bo_subdata(kernel->bo, 0, kernel->size, kernel->bin); + } + + media_state->private_context = i965_h264_context; + return True; +} + +Bool +i965_media_h264_ternimate(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_media_state *media_state = &i965->media_state; + struct i965_h264_context *i965_h264_context; + int i; + + if (media_state->private_context) { + i965_h264_context = (struct i965_h264_context *)media_state->private_context; + i965_avc_hw_scoreboard_ternimate(&i965_h264_context->avc_hw_scoreboard_context); + i965_avc_bsd_ternimate(&i965_h264_context->i965_avc_bsd_context); + dri_bo_unreference(i965_h264_context->avc_it_command_mb_info.bo); + dri_bo_unreference(i965_h264_context->avc_it_data.bo); + free(i965_h264_context); + media_state->private_context = NULL; + } + + for (i = 0; i < NUM_H264_AVC_KERNELS; i++) { + struct media_kernel *kernel = &h264_avc_kernels[i]; + + dri_bo_unreference(kernel->bo); + kernel->bo = NULL; + } + + return True; +} diff --git a/i965_media_h264.h b/i965_media_h264.h new file mode 100644 index 0000000..3001b2e --- /dev/null +++ b/i965_media_h264.h @@ -0,0 +1,52 @@ +#ifndef _I965_MEDIA_H264_H_ +#define _I965_MEDIA_H264_H_ + +#include "i965_avc_bsd.h" +#include "i965_avc_hw_scoreboard.h" + +#define MB_CMD_IN_BYTES 64 +#define MB_CMD_IN_DWS 16 +#define MB_CMD_IN_OWS 4 + +enum { + H264_AVC_COMBINED = 0, + H264_AVC_NULL +}; + +struct i965_h264_context +{ + struct { + dri_bo *bo; + unsigned int mbs; + } avc_it_command_mb_info; + + struct { + dri_bo *bo; + long write_offset; + } avc_it_data; + + struct { + unsigned int width_in_mbs; + unsigned int height_in_mbs; + int mbaff_frame_flag; + } picture; + + int use_avc_hw_scoreboard; + + int use_hw_w128; + unsigned int weight128_luma_l0; + unsigned int weight128_luma_l1; + unsigned int weight128_chroma_l0; + unsigned int weight128_chroma_l1; + char weight128_offset0_flag; + short weight128_offset0; + + struct i965_avc_bsd_context i965_avc_bsd_context; + struct i965_avc_hw_scoreboard_context avc_hw_scoreboard_context; +}; + +Bool i965_media_h264_init(VADriverContextP ctx); +Bool i965_media_h264_ternimate(VADriverContextP ctx); +void i965_media_h264_decode_init(VADriverContextP ctx); + +#endif /* _I965_MEDIA_H264_H_ */ diff --git a/i965_media_mpeg2.c b/i965_media_mpeg2.c index 83ee505..eaed1e9 100644 --- a/i965_media_mpeg2.c +++ b/i965_media_mpeg2.c @@ -280,7 +280,7 @@ static struct media_kernel mpeg2_vld_kernels_gen4[] = { } }; -/* On IGDNG */ +/* On IRONLAKE */ static uint32_t frame_intra_kernel_gen5[][4] = { #include "shaders/mpeg2/vld/frame_intra.g4b.gen5" }; @@ -886,6 +886,7 @@ i965_media_mpeg2_decode_init(VADriverContextP ctx) dri_bo *bo; media_state->extended_state.enabled = 1; + media_state->indirect_object.bo = NULL; dri_bo_unreference(media_state->extended_state.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "vld state", @@ -907,7 +908,7 @@ i965_media_mpeg2_decode_init(VADriverContextP ctx) media_state->urb.num_cs_entries * media_state->urb.size_cs_entry <= URB_SIZE((&i965->intel))); /* hook functions */ - media_state->states_setup = i965_media_mpeg2_states_setup; + media_state->media_states_setup = i965_media_mpeg2_states_setup; media_state->media_objects = i965_media_mpeg2_objects; } @@ -923,7 +924,7 @@ i965_media_mpeg2_init(VADriverContextP ctx) sizeof(mpeg2_vld_kernels_gen5[0]))); assert(NUM_MPEG2_VLD_KERNELS <= MAX_INTERFACE_DESC); - if (IS_IGDNG(i965->intel.device_id)) + if (IS_IRONLAKE(i965->intel.device_id)) mpeg2_vld_kernels = mpeg2_vld_kernels_gen5; else mpeg2_vld_kernels = mpeg2_vld_kernels_gen4; diff --git a/i965_render.c b/i965_render.c index 4a4f041..08fa582 100644 --- a/i965_render.c +++ b/i965_render.c @@ -75,7 +75,7 @@ static const unsigned int ps_subpic_kernel_static[][4] = #include "shaders/render/exa_wm_write.g4b" }; -/* On IGDNG */ +/* On IRONLAKE */ static const unsigned int sf_kernel_static_gen5[][4] = { #include "shaders/render/exa_sf.g4b.gen5" @@ -183,8 +183,8 @@ static struct render_kernel *render_kernels = NULL; #define URB_SF_ENTRIES 1 #define URB_SF_ENTRY_SIZE 2 -#define URB_CS_ENTRIES 0 -#define URB_CS_ENTRY_SIZE 0 +#define URB_CS_ENTRIES 1 +#define URB_CS_ENTRY_SIZE 1 static void i965_render_vs_unit(VADriverContextP ctx) @@ -198,7 +198,7 @@ i965_render_vs_unit(VADriverContextP ctx) vs_state = render_state->vs.state->virtual; memset(vs_state, 0, sizeof(*vs_state)); - if (IS_IGDNG(i965->intel.device_id)) + if (IS_IRONLAKE(i965->intel.device_id)) vs_state->thread4.nr_urb_entries = URB_VS_ENTRIES >> 2; else vs_state->thread4.nr_urb_entries = URB_VS_ENTRIES; @@ -312,7 +312,7 @@ i965_subpic_render_wm_unit(VADriverContextP ctx) wm_state->thread1.single_program_flow = 1; /* XXX */ - if (IS_IGDNG(i965->intel.device_id)) + if (IS_IRONLAKE(i965->intel.device_id)) wm_state->thread1.binding_table_entry_count = 0; /* hardware requirement */ else wm_state->thread1.binding_table_entry_count = 7; @@ -329,7 +329,7 @@ i965_subpic_render_wm_unit(VADriverContextP ctx) wm_state->wm4.stats_enable = 0; wm_state->wm4.sampler_state_pointer = render_state->wm.sampler->offset >> 5; - if (IS_IGDNG(i965->intel.device_id)) + if (IS_IRONLAKE(i965->intel.device_id)) wm_state->wm4.sampler_count = 0; /* hardware requirement */ else wm_state->wm4.sampler_count = (render_state->wm.sampler_count + 3) / 4; @@ -375,7 +375,7 @@ i965_render_wm_unit(VADriverContextP ctx) wm_state->thread1.single_program_flow = 1; /* XXX */ - if (IS_IGDNG(i965->intel.device_id)) + if (IS_IRONLAKE(i965->intel.device_id)) wm_state->thread1.binding_table_entry_count = 0; /* hardware requirement */ else wm_state->thread1.binding_table_entry_count = 7; @@ -383,8 +383,8 @@ i965_render_wm_unit(VADriverContextP ctx) wm_state->thread2.scratch_space_base_pointer = 0; wm_state->thread2.per_thread_scratch_space = 0; /* 1024 bytes */ - wm_state->thread3.dispatch_grf_start_reg = 3; /* XXX */ - wm_state->thread3.const_urb_entry_read_length = 0; + wm_state->thread3.dispatch_grf_start_reg = 2; /* XXX */ + wm_state->thread3.const_urb_entry_read_length = 1; wm_state->thread3.const_urb_entry_read_offset = 0; wm_state->thread3.urb_entry_read_length = 1; /* XXX */ wm_state->thread3.urb_entry_read_offset = 0; /* XXX */ @@ -392,7 +392,7 @@ i965_render_wm_unit(VADriverContextP ctx) wm_state->wm4.stats_enable = 0; wm_state->wm4.sampler_state_pointer = render_state->wm.sampler->offset >> 5; - if (IS_IGDNG(i965->intel.device_id)) + if (IS_IRONLAKE(i965->intel.device_id)) wm_state->wm4.sampler_count = 0; /* hardware requirement */ else wm_state->wm4.sampler_count = (render_state->wm.sampler_count + 3) / 4; @@ -532,7 +532,8 @@ i965_render_src_surface_state(VADriverContextP ctx, int index, dri_bo *region, unsigned long offset, - int w, int h) + int w, int h, + int pitch, int format) { struct i965_driver_data *i965 = i965_driver_data(ctx); struct i965_render_state *render_state = &i965->render_state; @@ -548,7 +549,7 @@ i965_render_src_surface_state(VADriverContextP ctx, ss = ss_bo->virtual; memset(ss, 0, sizeof(*ss)); ss->ss0.surface_type = I965_SURFACE_2D; - ss->ss0.surface_format = I965_SURFACEFORMAT_R8_UNORM; + ss->ss0.surface_format = format; ss->ss0.writedisable_alpha = 0; ss->ss0.writedisable_red = 0; ss->ss0.writedisable_green = 0; @@ -566,7 +567,7 @@ i965_render_src_surface_state(VADriverContextP ctx, ss->ss2.mip_count = 0; ss->ss2.render_target_rotation = 0; - ss->ss3.pitch = w - 1; + ss->ss3.pitch = pitch - 1; dri_bo_emit_reloc(ss_bo, I915_GEM_DOMAIN_SAMPLER, 0, @@ -642,6 +643,7 @@ i965_render_src_surfaces_state(VADriverContextP ctx, VASurfaceID surface) { struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_render_state *render_state = &i965->render_state; struct object_surface *obj_surface; int w, h; dri_bo *region; @@ -653,12 +655,18 @@ i965_render_src_surfaces_state(VADriverContextP ctx, h = obj_surface->height; region = obj_surface->bo; - i965_render_src_surface_state(ctx, 1, region, 0, w, h); /* Y */ - i965_render_src_surface_state(ctx, 2, region, 0, w, h); - i965_render_src_surface_state(ctx, 3, region, w * h + w * h / 4, w / 2, h / 2); /* V */ - i965_render_src_surface_state(ctx, 4, region, w * h + w * h / 4, w / 2, h / 2); - i965_render_src_surface_state(ctx, 5, region, w * h, w / 2, h / 2); /* U */ - i965_render_src_surface_state(ctx, 6, region, w * h, w / 2, h / 2); + i965_render_src_surface_state(ctx, 1, region, 0, w, h, w, I965_SURFACEFORMAT_R8_UNORM); /* Y */ + i965_render_src_surface_state(ctx, 2, region, 0, w, h, w, I965_SURFACEFORMAT_R8_UNORM); + + if (render_state->interleaved_uv) { + i965_render_src_surface_state(ctx, 3, region, w * h, w / 2, h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM); /* UV */ + i965_render_src_surface_state(ctx, 4, region, w * h, w / 2, h / 2, w, I965_SURFACEFORMAT_R8G8_UNORM); + } else { + i965_render_src_surface_state(ctx, 3, region, w * h, w / 2, h / 2, w / 2, I965_SURFACEFORMAT_R8_UNORM); /* U */ + i965_render_src_surface_state(ctx, 4, region, w * h, w / 2, h / 2, w / 2, I965_SURFACEFORMAT_R8_UNORM); + i965_render_src_surface_state(ctx, 5, region, w * h + w * h / 4, w / 2, h / 2, w / 2, I965_SURFACEFORMAT_R8_UNORM); /* V */ + i965_render_src_surface_state(ctx, 6, region, w * h + w * h / 4, w / 2, h / 2, w / 2, I965_SURFACEFORMAT_R8_UNORM); + } } static void @@ -903,6 +911,25 @@ i965_render_upload_vertex(VADriverContextP ctx, } static void +i965_render_upload_constants(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_render_state *render_state = &i965->render_state; + unsigned short *constant_buffer; + + dri_bo_map(render_state->curbe.bo, 1); + assert(render_state->curbe.bo->virtual); + constant_buffer = render_state->curbe.bo->virtual; + + if (render_state->interleaved_uv) + *constant_buffer = 1; + else + *constant_buffer = 0; + + dri_bo_unmap(render_state->curbe.bo); +} + +static void i965_surface_render_state_setup(VADriverContextP ctx, VASurfaceID surface, short srcx, @@ -926,6 +953,7 @@ i965_surface_render_state_setup(VADriverContextP ctx, i965_render_upload_vertex(ctx, surface, srcx, srcy, srcw, srch, destx, desty, destw, desth); + i965_render_upload_constants(ctx); } static void i965_subpic_render_state_setup(VADriverContextP ctx, @@ -980,7 +1008,7 @@ i965_render_state_base_address(VADriverContextP ctx) { struct i965_driver_data *i965 = i965_driver_data(ctx); - if (IS_IGDNG(i965->intel.device_id)) { + if (IS_IRONLAKE(i965->intel.device_id)) { BEGIN_BATCH(ctx, 8); OUT_BATCH(ctx, CMD_STATE_BASE_ADDRESS | 6); OUT_BATCH(ctx, 0 | BASE_ADDRESS_MODIFY); @@ -1099,6 +1127,20 @@ i965_render_cs_urb_layout(VADriverContextP ctx) } static void +i965_render_constant_buffer(VADriverContextP ctx) +{ + struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_render_state *render_state = &i965->render_state; + + BEGIN_BATCH(ctx, 2); + OUT_BATCH(ctx, CMD_CONSTANT_BUFFER | (1 << 8) | (2 - 2)); + OUT_RELOC(ctx, render_state->curbe.bo, + I915_GEM_DOMAIN_INSTRUCTION, 0, + URB_CS_ENTRY_SIZE - 1); + ADVANCE_BATCH(ctx); +} + +static void i965_render_drawing_rectangle(VADriverContextP ctx) { struct i965_driver_data *i965 = i965_driver_data(ctx); @@ -1118,7 +1160,7 @@ i965_render_vertex_elements(VADriverContextP ctx) { struct i965_driver_data *i965 = i965_driver_data(ctx); - if (IS_IGDNG(i965->intel.device_id)) { + if (IS_IRONLAKE(i965->intel.device_id)) { BEGIN_BATCH(ctx, 5); OUT_BATCH(ctx, CMD_VERTEX_ELEMENTS | 3); /* offset 0: X,Y -> {X, Y, 1.0, 1.0} */ @@ -1206,7 +1248,7 @@ i965_render_startup(VADriverContextP ctx) ((4 * 4) << VB0_BUFFER_PITCH_SHIFT)); OUT_RELOC(ctx, render_state->vb.vertex_buffer, I915_GEM_DOMAIN_VERTEX, 0, 0); - if (IS_IGDNG(i965->intel.device_id)) + if (IS_IRONLAKE(i965->intel.device_id)) OUT_RELOC(ctx, render_state->vb.vertex_buffer, I915_GEM_DOMAIN_VERTEX, 0, 12 * 4); else OUT_BATCH(ctx, 3); @@ -1282,6 +1324,7 @@ i965_surface_render_pipeline_setup(VADriverContextP ctx) i965_render_pipelined_pointers(ctx); i965_render_urb_layout(ctx); i965_render_cs_urb_layout(ctx); + i965_render_constant_buffer(ctx); i965_render_drawing_rectangle(ctx); i965_render_vertex_elements(ctx); i965_render_startup(ctx); @@ -1445,13 +1488,14 @@ Bool i965_render_init(VADriverContextP ctx) { struct i965_driver_data *i965 = i965_driver_data(ctx); + struct i965_render_state *render_state = &i965->render_state; int i; /* kernel */ assert(NUM_RENDER_KERNEL == (sizeof(render_kernels_gen5) / sizeof(render_kernels_gen5[0]))); - if (IS_IGDNG(i965->intel.device_id)) + if (IS_IRONLAKE(i965->intel.device_id)) render_kernels = render_kernels_gen5; else render_kernels = render_kernels_gen4; @@ -1465,6 +1509,12 @@ i965_render_init(VADriverContextP ctx) dri_bo_subdata(kernel->bo, 0, kernel->size, kernel->bin); } + /* constant buffer */ + render_state->curbe.bo = dri_bo_alloc(i965->intel.bufmgr, + "constant buffer", + 4096, 64); + assert(render_state->curbe.bo); + return True; } @@ -1475,6 +1525,9 @@ i965_render_terminate(VADriverContextP ctx) struct i965_driver_data *i965 = i965_driver_data(ctx); struct i965_render_state *render_state = &i965->render_state; + dri_bo_unreference(render_state->curbe.bo); + render_state->curbe.bo = NULL; + for (i = 0; i < NUM_RENDER_KERNEL; i++) { struct render_kernel *kernel = &render_kernels[i]; diff --git a/i965_render.h b/i965_render.h index e3dce02..2643285 100644 --- a/i965_render.h +++ b/i965_render.h @@ -60,6 +60,11 @@ struct i965_render_state dri_bo *viewport; } cc; + struct { + dri_bo *bo; + } curbe; + + int interleaved_uv; struct intel_region *draw_region; }; diff --git a/i965_structs.h b/i965_structs.h index 93fa2f0..d133446 100644 --- a/i965_structs.h +++ b/i965_structs.h @@ -34,16 +34,20 @@ struct i965_vfe_state_ex unsigned int obj_id:24; } vfex0; - struct { - unsigned int residual_grf_offset:5; - unsigned int pad0:3; - unsigned int weight_grf_offset:5; - unsigned int pad1:3; - unsigned int residual_data_offset:8; - unsigned int sub_field_present_flag:2; - unsigned int residual_data_fix_offset:1; - unsigned int pad2:5; - }vfex1; + union { + struct { + unsigned int residual_grf_offset:5; + unsigned int pad0:3; + unsigned int weight_grf_offset:5; + unsigned int pad1:3; + unsigned int residual_data_offset:8; + unsigned int sub_field_present_flag:2; + unsigned int residual_data_fix_offset_flag:1; + unsigned int pad2:5; + } avc; + + unsigned int vc1; + } vfex1; struct { unsigned int remap_index_0:4; @@ -68,18 +72,32 @@ struct i965_vfe_state_ex } remap_table1; struct { - unsigned int scoreboard_mask:8; + unsigned int mask:8; unsigned int pad:22; unsigned int type:1; unsigned int enable:1; } scoreboard0; struct { - unsigned int ignore; + int delta_x0:4; + int delta_y0:4; + int delta_x1:4; + int delta_y1:4; + int delta_x2:4; + int delta_y2:4; + int delta_x3:4; + int delta_y3:4; } scoreboard1; struct { - unsigned int ignore; + int delta_x4:4; + int delta_y4:4; + int delta_x5:4; + int delta_y5:4; + int delta_x6:4; + int delta_y6:4; + int delta_x7:4; + int delta_y7:4; } scoreboard2; unsigned int pad; @@ -177,8 +195,9 @@ struct i965_surface_state unsigned int cube_neg_y:1; unsigned int cube_pos_x:1; unsigned int cube_neg_x:1; - unsigned int pad:3; + unsigned int pad:2; unsigned int render_cache_read_mode:1; + unsigned int cube_map_corner_mode:1; unsigned int mipmap_layout_mode:1; unsigned int vert_line_stride_ofs:1; unsigned int vert_line_stride:1; diff --git a/intel_batchbuffer.c b/intel_batchbuffer.c index 9fabf6a..4c622d6 100644 --- a/intel_batchbuffer.c +++ b/intel_batchbuffer.c @@ -38,16 +38,16 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) { struct intel_driver_data *intel = batch->intel; - if (batch->buffer != NULL) { - dri_bo_unreference(batch->buffer); - batch->buffer = NULL; - } + assert(batch->flag == ON_RENDER_RING || + batch->flag == ON_BSD_RING); - batch->buffer = dri_bo_alloc(intel->bufmgr, "batch buffer", + dri_bo_unreference(batch->buffer); + batch->buffer = dri_bo_alloc(intel->bufmgr, + batch->flag == ON_RENDER_RING ? "render batch buffer" : "bsd batch buffer", BATCH_SIZE, 0x1000); - assert(batch->buffer); dri_bo_map(batch->buffer, 1); + assert(batch->buffer->virtual); batch->map = batch->buffer->virtual; batch->size = BATCH_SIZE; batch->ptr = batch->map; @@ -58,11 +58,19 @@ Bool intel_batchbuffer_init(struct intel_driver_data *intel) { intel->batch = calloc(1, sizeof(*(intel->batch))); - assert(intel->batch); intel->batch->intel = intel; + intel->batch->flag = ON_RENDER_RING; + intel->batch->run = drm_intel_bo_mrb_exec; intel_batchbuffer_reset(intel->batch); + intel->batch_bcs = calloc(1, sizeof(*(intel->batch_bcs))); + assert(intel->batch_bcs); + intel->batch_bcs->intel = intel; + intel->batch_bcs->flag = ON_BSD_RING; + intel->batch_bcs->run = drm_intel_bo_mrb_exec; + intel_batchbuffer_reset(intel->batch_bcs); + return True; } @@ -80,14 +88,25 @@ intel_batchbuffer_terminate(struct intel_driver_data *intel) intel->batch = NULL; } + if (intel->batch_bcs) { + if (intel->batch_bcs->map) { + dri_bo_unmap(intel->batch_bcs->buffer); + intel->batch_bcs->map = NULL; + } + + dri_bo_unreference(intel->batch_bcs->buffer); + free(intel->batch_bcs); + intel->batch_bcs = NULL; + } + return True; } -Bool -intel_batchbuffer_flush(VADriverContextP ctx) +static Bool +intel_batchbuffer_flush_helper(VADriverContextP ctx, + struct intel_batchbuffer *batch) { - struct intel_driver_data *intel = intel_driver_data(ctx); - struct intel_batchbuffer *batch = intel->batch; + struct intel_driver_data *intel = batch->intel; unsigned int used = batch->ptr - batch->map; int is_locked = intel->locked; @@ -108,98 +127,228 @@ intel_batchbuffer_flush(VADriverContextP ctx) if (!is_locked) intel_lock_hardware(ctx); - dri_bo_exec(batch->buffer, used, 0, 0, 0); + batch->run(batch->buffer, used, 0, 0, 0, batch->flag); if (!is_locked) intel_unlock_hardware(ctx); - intel_batchbuffer_reset(intel->batch); + intel_batchbuffer_reset(batch); return True; } +Bool +intel_batchbuffer_flush(VADriverContextP ctx) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + return intel_batchbuffer_flush_helper(ctx, intel->batch); +} + +Bool +intel_batchbuffer_flush_bcs(VADriverContextP ctx) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + return intel_batchbuffer_flush_helper(ctx, intel->batch_bcs); +} + static unsigned int -intel_batchbuffer_space(struct intel_batchbuffer *batch) +intel_batchbuffer_space_helper(struct intel_batchbuffer *batch) { return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); } +static void +intel_batchbuffer_emit_dword_helper(struct intel_batchbuffer *batch, + unsigned int x) +{ + assert(intel_batchbuffer_space_helper(batch) >= 4); + *(unsigned int *)batch->ptr = x; + batch->ptr += 4; +} + void intel_batchbuffer_emit_dword(VADriverContextP ctx, unsigned int x) { struct intel_driver_data *intel = intel_driver_data(ctx); - struct intel_batchbuffer *batch = intel->batch; - assert(intel_batchbuffer_space(batch) >= 4); - *(unsigned int*)batch->ptr = x; - batch->ptr += 4; + intel_batchbuffer_emit_dword_helper(intel->batch, x); } void -intel_batchbuffer_emit_reloc(VADriverContextP ctx, dri_bo *bo, - uint32_t read_domains, uint32_t write_domains, - uint32_t delta) +intel_batchbuffer_emit_dword_bcs(VADriverContextP ctx, unsigned int x) { struct intel_driver_data *intel = intel_driver_data(ctx); - struct intel_batchbuffer *batch = intel->batch; + intel_batchbuffer_emit_dword_helper(intel->batch_bcs, x); +} + +static void +intel_batchbuffer_emit_reloc_helper(VADriverContextP ctx, + struct intel_batchbuffer *batch, + dri_bo *bo, + uint32_t read_domains, uint32_t write_domains, + uint32_t delta) +{ assert(batch->ptr - batch->map < batch->size); dri_bo_emit_reloc(batch->buffer, read_domains, write_domains, delta, batch->ptr - batch->map, bo); - intel_batchbuffer_emit_dword(ctx, bo->offset + delta); + intel_batchbuffer_emit_dword_helper(batch, bo->offset + delta); } void -intel_batchbuffer_require_space(VADriverContextP ctx, unsigned int size) +intel_batchbuffer_emit_reloc(VADriverContextP ctx, dri_bo *bo, + uint32_t read_domains, uint32_t write_domains, + uint32_t delta) { struct intel_driver_data *intel = intel_driver_data(ctx); - struct intel_batchbuffer *batch = intel->batch; + intel_batchbuffer_emit_reloc_helper(ctx, intel->batch, + bo, read_domains, write_domains, + delta); +} + +void +intel_batchbuffer_emit_reloc_bcs(VADriverContextP ctx, dri_bo *bo, + uint32_t read_domains, uint32_t write_domains, + uint32_t delta) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + intel_batchbuffer_emit_reloc_helper(ctx, intel->batch_bcs, + bo, read_domains, write_domains, + delta); +} + +static void +intel_batchbuffer_require_space_helper(VADriverContextP ctx, + struct intel_batchbuffer *batch, + unsigned int size) +{ assert(size < batch->size - 8); - if (intel_batchbuffer_space(batch) < size) { - intel_batchbuffer_flush(ctx); + if (intel_batchbuffer_space_helper(batch) < size) { + intel_batchbuffer_flush_helper(ctx, batch); } } void -intel_batchbuffer_data(VADriverContextP ctx, void *data, unsigned int size) +intel_batchbuffer_require_space(VADriverContextP ctx, unsigned int size) { struct intel_driver_data *intel = intel_driver_data(ctx); - struct intel_batchbuffer *batch = intel->batch; + intel_batchbuffer_require_space_helper(ctx, intel->batch, size); +} + +void +intel_batchbuffer_require_space_bcs(VADriverContextP ctx, unsigned int size) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + intel_batchbuffer_require_space_helper(ctx, intel->batch_bcs, size); +} + +static void +intel_batchbuffer_data_helper(VADriverContextP ctx, + struct intel_batchbuffer *batch, + void *data, + unsigned int size) +{ assert((size & 3) == 0); - intel_batchbuffer_require_space(ctx, size); + intel_batchbuffer_require_space_helper(ctx, batch, size); assert(batch->ptr); memcpy(batch->ptr, data, size); batch->ptr += size; } +void +intel_batchbuffer_data(VADriverContextP ctx, void *data, unsigned int size) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + intel_batchbuffer_data_helper(ctx, intel->batch, data, size); +} + +void +intel_batchbuffer_data_bcs(VADriverContextP ctx, void *data, unsigned int size) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + intel_batchbuffer_data_helper(ctx, intel->batch_bcs, data, size); +} + +static void +intel_batchbuffer_emit_mi_flush_helper(VADriverContextP ctx, + struct intel_batchbuffer *batch) +{ + intel_batchbuffer_require_space_helper(ctx, batch, 4); + intel_batchbuffer_emit_dword_helper(batch, + MI_FLUSH | STATE_INSTRUCTION_CACHE_INVALIDATE); +} + void intel_batchbuffer_emit_mi_flush(VADriverContextP ctx) { - intel_batchbuffer_require_space(ctx, 4); - intel_batchbuffer_emit_dword(ctx, MI_FLUSH | STATE_INSTRUCTION_CACHE_INVALIDATE); + struct intel_driver_data *intel = intel_driver_data(ctx); + + intel_batchbuffer_emit_mi_flush_helper(ctx, intel->batch); } void -intel_batchbuffer_start_atomic(VADriverContextP ctx, unsigned int size) +intel_batchbuffer_emit_mi_flush_bcs(VADriverContextP ctx) { struct intel_driver_data *intel = intel_driver_data(ctx); - struct intel_batchbuffer *batch = intel->batch; + intel_batchbuffer_emit_mi_flush_helper(ctx, intel->batch_bcs); +} + +void +intel_batchbuffer_start_atomic_helper(VADriverContextP ctx, + struct intel_batchbuffer *batch, + unsigned int size) +{ assert(!batch->atomic); - intel_batchbuffer_require_space(ctx, size); + intel_batchbuffer_require_space_helper(ctx, batch, size); batch->atomic = 1; } void -intel_batchbuffer_end_atomic(VADriverContextP ctx) +intel_batchbuffer_start_atomic(VADriverContextP ctx, unsigned int size) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + intel_batchbuffer_start_atomic_helper(ctx, intel->batch, size); +} + +void +intel_batchbuffer_start_atomic_bcs(VADriverContextP ctx, unsigned int size) { struct intel_driver_data *intel = intel_driver_data(ctx); - struct intel_batchbuffer *batch = intel->batch; + intel_batchbuffer_start_atomic_helper(ctx, intel->batch_bcs, size); +} +void +intel_batchbuffer_end_atomic_helper(struct intel_batchbuffer *batch) +{ assert(batch->atomic); batch->atomic = 0; } + +void +intel_batchbuffer_end_atomic(VADriverContextP ctx) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + intel_batchbuffer_end_atomic_helper(intel->batch); +} + +void +intel_batchbuffer_end_atomic_bcs(VADriverContextP ctx) +{ + struct intel_driver_data *intel = intel_driver_data(ctx); + + intel_batchbuffer_end_atomic_helper(intel->batch_bcs); +} + diff --git a/intel_batchbuffer.h b/intel_batchbuffer.h index c3a2457..99ab08d 100644 --- a/intel_batchbuffer.h +++ b/intel_batchbuffer.h @@ -16,10 +16,16 @@ struct intel_batchbuffer unsigned char *map; unsigned char *ptr; int atomic; + int flag; + + int (*run)(drm_intel_bo *bo, int used, + drm_clip_rect_t *cliprects, int num_cliprects, + int DR4, int ring_flag); }; Bool intel_batchbuffer_init(struct intel_driver_data *intel); Bool intel_batchbuffer_terminate(struct intel_driver_data *intel); + void intel_batchbuffer_emit_dword(VADriverContextP ctx, unsigned int x); void intel_batchbuffer_emit_reloc(VADriverContextP ctx, dri_bo *bo, uint32_t read_domains, uint32_t write_domains, @@ -31,6 +37,17 @@ void intel_batchbuffer_start_atomic(VADriverContextP ctx, unsigned int size); void intel_batchbuffer_end_atomic(VADriverContextP ctx); Bool intel_batchbuffer_flush(VADriverContextP ctx); +void intel_batchbuffer_emit_dword_bcs(VADriverContextP ctx, unsigned int x); +void intel_batchbuffer_emit_reloc_bcs(VADriverContextP ctx, dri_bo *bo, + uint32_t read_domains, uint32_t write_domains, + uint32_t delta); +void intel_batchbuffer_require_space_bcs(VADriverContextP ctx, unsigned int size); +void intel_batchbuffer_data_bcs(VADriverContextP ctx, void *data, unsigned int size); +void intel_batchbuffer_emit_mi_flush_bcs(VADriverContextP ctx); +void intel_batchbuffer_start_atomic_bcs(VADriverContextP ctx, unsigned int size); +void intel_batchbuffer_end_atomic_bcs(VADriverContextP ctx); +Bool intel_batchbuffer_flush_bcs(VADriverContextP ctx); + #define BEGIN_BATCH(ctx, n) do { \ intel_batchbuffer_require_space(ctx, (n) * 4); \ } while (0) @@ -48,4 +65,21 @@ Bool intel_batchbuffer_flush(VADriverContextP ctx); #define ADVANCE_BATCH(ctx) do { \ } while (0) +#define BEGIN_BCS_BATCH(ctx, n) do { \ + intel_batchbuffer_require_space_bcs(ctx, (n) * 4); \ +} while (0) + +#define OUT_BCS_BATCH(ctx, d) do { \ + intel_batchbuffer_emit_dword_bcs(ctx, d); \ +} while (0) + +#define OUT_BCS_RELOC(ctx, bo, read_domains, write_domain, delta) do { \ + assert((delta) >= 0); \ + intel_batchbuffer_emit_reloc_bcs(ctx, bo, \ + read_domains, write_domain, delta); \ +} while (0) + +#define ADVANCE_BCS_BATCH(ctx) do { \ +} while (0) + #endif /* _INTEL_BATCHBUFFER_H_ */ diff --git a/intel_batchbuffer_dump.c b/intel_batchbuffer_dump.c new file mode 100644 index 0000000..0732f0f --- /dev/null +++ b/intel_batchbuffer_dump.c @@ -0,0 +1,409 @@ +#include <stdio.h> +#include <stdarg.h> +#include <string.h> +#include <inttypes.h> + +#include "intel_driver.h" +#include "intel_batchbuffer_dump.h" + +#define BUFFER_FAIL(_count, _len, _name) do { \ + fprintf(gout, "Buffer size too small in %s (%d < %d)\n", \ + (_name), (_count), (_len)); \ + (*failures)++; \ + return count; \ +} while (0) + +static FILE *gout; + +static void +instr_out(unsigned int *data, unsigned int offset, unsigned int index, char *fmt, ...) +{ + va_list va; + + fprintf(gout, "0x%08x: 0x%08x:%s ", offset + index * 4, data[index], + index == 0 ? "" : " "); + va_start(va, fmt); + vfprintf(gout, fmt, va); + va_end(va); +} + + +static int +dump_mi(unsigned int *data, unsigned int offset, int count, unsigned int device, int *failures) +{ + unsigned int opcode; + int length, i; + + struct { + unsigned int opcode; + int mask_length; + int min_len; + int max_len; + char *name; + } mi_commands[] = { + { 0x00, 0, 1, 1, "MI_NOOP" }, + { 0x04, 0, 1, 1, "MI_FLUSH" }, + { 0x0a, 0, 1, 1, "MI_BATCH_BUFFER_END" }, + }; + + opcode = ((data[0] & MASK_MI_OPCODE) >> SHIFT_MI_OPCODE); + + for (i = 0; i < sizeof(mi_commands) / sizeof(mi_commands[0]); i++) { + if (opcode == mi_commands[i].opcode) { + int index; + + length = 1; + instr_out(data, offset, 0, "%s\n", mi_commands[i].name); + + if (mi_commands[i].max_len > 1) { + length = (data[0] & mi_commands[i].mask_length) + 2; + + if (length < mi_commands[i].min_len || + length > mi_commands[i].max_len) { + fprintf(gout, "Bad length (%d) in %s, [%d, %d]\n", + length, mi_commands[i].name, + mi_commands[i].min_len, + mi_commands[i].max_len); + } + } + + for (index = 1; index < length; index++) { + if (index >= count) + BUFFER_FAIL(count, length, mi_commands[i].name); + + instr_out(data, offset, index, "dword %d\n", index); + } + + return length; + } + } + + instr_out(data, offset, 0, "UNKNOWN MI COMMAND\n"); + (*failures)++; + return 1; +} + +static int +dump_gfxpipe_3d(unsigned int *data, unsigned int offset, int count, unsigned int device, int *failures) +{ + instr_out(data, offset, 0, "UNKNOWN 3D COMMAND\n"); + (*failures)++; + + return 1; +} + +static void +dump_avc_bsd_img_state(unsigned int *data, unsigned int offset, unsigned int device, int *failures) +{ + int img_struct = ((data[3] >> 8) & 0x3); + + instr_out(data, offset, 1, "frame size: %d\n", (data[1] & 0xffff)); + instr_out(data, offset, 2, "width: %d, height: %d\n", (data[2] & 0xff), (data[2] >> 16) & 0xff); + instr_out(data, offset, 3, + "second_chroma_qp_offset: %d," + "chroma_qp_offset: %d," + "QM present flag: %d," + "image struct: %s," + "img_dec_fs_idc: %d," + "\n", + (data[3] >> 24) & 0x1f, + (data[3] >> 16) & 0x1f, + (data[3] >> 10) & 0x1, + (img_struct == 0) ? "frame" : (img_struct == 2) ? "invalid" : (img_struct == 1) ? "top field" : "bottom field", + data[3] & 0xff); + instr_out(data, offset, 4, + "residual off: 0x%x," + "16MV: %d," + "chroma fmt: %d," + "CABAC: %d," + "non-ref: %d," + "constrained intra: %d," + "direct8x8: %d," + "trans8x8: %d," + "MB only: %d," + "MBAFF: %d," + "\n", + (data[4] >> 24) & 0xff, + (data[4] >> 12) & 0x1, + (data[4] >> 10) & 0x3, + (data[4] >> 7) & 0x1, + (data[4] >> 6) & 0x1, + (data[4] >> 5) & 0x1, + (data[4] >> 4) & 0x1, + (data[4] >> 3) & 0x1, + (data[4] >> 2) & 0x1, + (data[4] >> 1) & 0x1); + instr_out(data, offset, 5, "AVC-IT Command Header\n"); +} + +static void +dump_avc_bsd_qm_state(unsigned int *data, unsigned int offset, unsigned int device, int *failures) +{ + unsigned int length = ((data[0] & MASK_GFXPIPE_LENGTH) >> SHIFT_GFXPIPE_LENGTH) + 2; + int i; + + instr_out(data, offset, 1, "user default: %02x, QM list present: %02x\n", + (data[1] >> 8) & 0xff, data[1] & 0xff); + + for (i = 2; i < length; i++) { + instr_out(data, offset, i, "dword %d\n", i); + } +} + +static void +dump_avc_bsd_slice_state(unsigned int *data, unsigned int offset, unsigned int device, int *failures) +{ + +} + +static void +dump_avc_bsd_buf_base_state(unsigned int *data, unsigned int offset, unsigned int device, int *failures) +{ + int i; + + instr_out(data, offset, 1, "BSD row store base address\n"); + instr_out(data, offset, 2, "MPR row store base address\n"); + instr_out(data, offset, 3, "AVC-IT command buffer base address\n"); + instr_out(data, offset, 4, "AVC-IT data buffer: 0x%08x, write offset: 0x%x\n", + data[4] & 0xFFFFF000, data[4] & 0xFC0); + instr_out(data, offset, 5, "ILDB data buffer\n"); + + for (i = 6; i < 38; i++) { + instr_out(data, offset, i, "Direct MV read base address for reference frame %d\n", i - 6); + } + + instr_out(data, offset, 38, "direct mv wr0 top\n"); + instr_out(data, offset, 39, "direct mv wr0 bottom\n"); + + for (i = 40; i < 74; i++) { + instr_out(data, offset, i, "POC List %d\n", i - 40); + } +} + +static void +dump_bsd_ind_obj_base_addr(unsigned int *data, unsigned int offset, unsigned int device, int *failures) +{ + instr_out(data, offset, 1, "AVC indirect object base address\n"); + instr_out(data, offset, 2, "AVC Indirect Object Access Upper Bound\n"); +} + +static void +dump_ironlake_avc_bsd_object(unsigned int *data, unsigned int offset, int *failures) +{ + int slice_type = data[3] & 0xf; + int i, is_phantom = ((data[1] & 0x3fffff) == 0); + + if (!is_phantom) { + instr_out(data, offset, 1, "Encrypted: %d, bitsteam length: %d\n", data[1] >> 31, data[1] & 0x3fffff); + instr_out(data, offset, 2, "Indirect Data Start Address: %d\n", data[2] & 0x1fffffff); + instr_out(data, offset, 3, "%s Slice\n", slice_type == 0 ? "P" : slice_type == 1 ? "B" : "I"); + instr_out(data, offset, 4, + "Num_Ref_Idx_L1: %d," + "Num_Ref_Idx_L0: %d," + "Log2WeightDenomChroma: %d," + "Log2WeightDenomLuma: %d" + "\n", + (data[4] >> 24) & 0x3f, + (data[4] >> 16) & 0x3f, + (data[4] >> 8) & 0x3, + (data[4] >> 0) & 0x3); + instr_out(data, offset, 5, + "WeightedPredIdc: %d," + "DirectPredType: %d," + "DisableDeblockingFilter: %d," + "CabacInitIdc: %d," + "SliceQp: %d," + "SliceBetaOffsetDiv2: %d," + "SliceAlphaC0OffsetDiv2: %d" + "\n", + (data[5] >> 30) & 0x3, + (data[5] >> 29) & 0x1, + (data[5] >> 27) & 0x3, + (data[5] >> 24) & 0x3, + (data[5] >> 16) & 0x3f, + (data[5] >> 8) & 0xf, + (data[5] >> 0) & 0xf); + instr_out(data, offset, 6, + "Slice_MB_Start_Vert_Pos: %d," + "Slice_MB_Start_Hor_Pos: %d," + "Slice_Start_Mb_Num: %d" + "\n", + (data[6] >> 24) & 0xff, + (data[6] >> 16) & 0xff, + (data[6] >> 0) & 0x7fff); + instr_out(data, offset, 7, + "Fix_Prev_Mb_Skipped: %d," + "First_MB_Bit_Offset: %d" + "\n", + (data[7] >> 7) & 0x1, + (data[7] >> 0) & 0x7); + + for (i = 8; i < 16; i++) + instr_out(data, offset, i, "dword %d\n", i); + } else { + instr_out(data, offset, 1, "phantom slice\n"); + + for (i = 2; i < 6; i++) + instr_out(data, offset, i, "dword %d\n", i); + + instr_out(data, offset, 6, + "Slice_Start_Mb_Num: %d" + "\n", + (data[6] >> 0) & 0x7fff); + + for (i = 7; i < 16; i++) + instr_out(data, offset, i, "dword %d\n", i); + + } +} + +static void +dump_g4x_avc_bsd_object(unsigned int *data, unsigned int offset, int *failures) +{ + +} + +static void +dump_avc_bsd_object(unsigned int *data, unsigned int offset, unsigned int device, int *failures) +{ + if (IS_IRONLAKE(device)) + dump_ironlake_avc_bsd_object(data, offset, failures); + else + dump_g4x_avc_bsd_object(data, offset, failures); +} + +static int +dump_bsd_avc(unsigned int *data, unsigned int offset, int count, unsigned int device, int *failures) +{ + unsigned int subopcode; + int length, i; + + struct { + unsigned int subopcode; + int min_len; + int max_len; + char *name; + void (*detail)(unsigned int *data, unsigned int offset, unsigned int device, int *failures); + } avc_commands[] = { + { 0x00, 0x06, 0x06, "AVC_BSD_IMG_STATE", dump_avc_bsd_img_state }, + { 0x01, 0x02, 0x3a, "AVC_BSD_QM_STATE", dump_avc_bsd_qm_state }, + { 0x02, 0x02, 0xd2, "AVC_BSD_SLICE_STATE", NULL }, + { 0x03, 0x4a, 0x4a, "AVC_BSD_BUF_BASE_STATE", dump_avc_bsd_buf_base_state }, + { 0x04, 0x03, 0x03, "BSD_IND_OBJ_BASE_ADDR", dump_bsd_ind_obj_base_addr }, + { 0x08, 0x08, 0x10, "AVC_BSD_OBJECT", dump_avc_bsd_object }, + }; + + subopcode = ((data[0] & MASK_GFXPIPE_SUBOPCODE) >> SHIFT_GFXPIPE_SUBOPCODE); + + for (i = 0; i < sizeof(avc_commands) / sizeof(avc_commands[0]); i++) { + if (subopcode == avc_commands[i].subopcode) { + unsigned int index; + + length = (data[0] & MASK_GFXPIPE_LENGTH) >> SHIFT_GFXPIPE_LENGTH; + length += 2; + instr_out(data, offset, 0, "%s\n", avc_commands[i].name); + + if (length < avc_commands[i].min_len || + length > avc_commands[i].max_len) { + fprintf(gout, "Bad length(%d) in %s [%d, %d]\n", + length, avc_commands[i].name, + avc_commands[i].min_len, + avc_commands[i].max_len); + } + + if (length - 1 >= count) + BUFFER_FAIL(count, length, avc_commands[i].name); + + if (avc_commands[i].detail) + avc_commands[i].detail(data, offset, device, failures); + else { + for (index = 1; index < length; index++) + instr_out(data, offset, index, "dword %d\n", index); + } + + return length; + } + } + + instr_out(data, offset, 0, "UNKNOWN AVC COMMAND\n"); + (*failures)++; + return 1; +} + +static int +dump_gfxpipe_bsd(unsigned int *data, unsigned int offset, int count, unsigned int device, int *failures) +{ + int length; + + switch ((data[0] & MASK_GFXPIPE_OPCODE) >> SHIFT_GFXPIPE_OPCODE) { + case OPCODE_BSD_AVC: + length = dump_bsd_avc(data, offset, count, device, failures); + break; + + default: + length = 1; + (*failures)++; + instr_out(data, offset, 0, "UNKNOWN BSD OPCODE\n"); + break; + } + + return length; +} + +static int +dump_gfxpipe(unsigned int *data, unsigned int offset, int count, unsigned int device, int *failures) +{ + int length; + + switch ((data[0] & MASK_GFXPIPE_SUBTYPE) >> SHIFT_GFXPIPE_SUBTYPE) { + case GFXPIPE_3D: + length = dump_gfxpipe_3d(data, offset, count, device, failures); + break; + + case GFXPIPE_BSD: + length = dump_gfxpipe_bsd(data, offset, count, device, failures); + break; + + default: + length = 1; + (*failures)++; + instr_out(data, offset, 0, "UNKNOWN GFXPIPE COMMAND\n"); + break; + } + + return length; +} + +int intel_batchbuffer_dump(unsigned int *data, unsigned int offset, int count, unsigned int device) +{ + int index = 0; + int failures = 0; + + gout = fopen("/tmp/bsd_command_dump.txt", "w+"); + + while (index < count) { + switch ((data[index] & MASK_CMD_TYPE) >> SHIFT_CMD_TYPE) { + case CMD_TYPE_MI: + index += dump_mi(data + index, offset + index * 4, + count - index, device, &failures); + break; + + case CMD_TYPE_GFXPIPE: + index += dump_gfxpipe(data + index, offset + index * 4, + count - index, device, &failures); + break; + + default: + instr_out(data, offset, index, "UNKNOWN COMMAND\n"); + failures++; + index++; + break; + } + + fflush(gout); + } + + fclose(gout); + + return failures; +} diff --git a/intel_batchbuffer_dump.h b/intel_batchbuffer_dump.h new file mode 100644 index 0000000..ad096a9 --- /dev/null +++ b/intel_batchbuffer_dump.h @@ -0,0 +1,49 @@ +#ifndef _INTEL_BATCHBUFFER_DUMP_H_ +#define _INTEL_BATCHBUFFER_DUMP_H_ + +#define MASK_CMD_TYPE 0xE0000000 + +#define SHIFT_CMD_TYPE 29 + +#define CMD_TYPE_GFXPIPE 3 +#define CMD_TYPE_BLT 2 +#define CMD_TYPE_MI 0 + + +/* GFXPIPE */ +#define MASK_GFXPIPE_SUBTYPE 0x18000000 +#define MASK_GFXPIPE_OPCODE 0x07000000 +#define MASK_GFXPIPE_SUBOPCODE 0x00FF0000 +#define MASK_GFXPIPE_LENGTH 0x0000FFFF + +#define SHIFT_GFXPIPE_SUBTYPE 27 +#define SHIFT_GFXPIPE_OPCODE 24 +#define SHIFT_GFXPIPE_SUBOPCODE 16 +#define SHIFT_GFXPIPE_LENGTH 0 + +/* 3D */ +#define GFXPIPE_3D 3 + +/* BSD */ +#define GFXPIPE_BSD 2 + +#define OPCODE_BSD_AVC 4 + +#define SUBOPCODE_BSD_IMG 0 +#define SUBOPCODE_BSD_QM 1 +#define SUBOPCODE_BSD_SLICE 2 +#define SUBOPCODE_BSD_BUF_BASE 3 +#define SUBOPCODE_BSD_IND_OBJ 4 +#define SUBOPCODE_BSD_OBJECT 8 + +/* MI */ +#define MASK_MI_OPCODE 0x1F800000 + +#define SHIFT_MI_OPCODE 23 + +#define OPCODE_MI_FLUSH 0x04 +#define OPCODE_MI_BATCH_BUFFER_END 0x0A + +int intel_batchbuffer_dump(unsigned int *data, unsigned int offset, int count, unsigned int device); + +#endif /* _INTEL_BATCHBUFFER_DUMP_H_ */ diff --git a/intel_driver.h b/intel_driver.h index 9ab95b7..7fcb760 100644 --- a/intel_driver.h +++ b/intel_driver.h @@ -17,7 +17,7 @@ #define INLINE #endif -#define BATCH_SIZE 0x10000 +#define BATCH_SIZE 0x100000 #define BATCH_RESERVED 0x10 #define CMD_MI (0x0 << 29) @@ -26,6 +26,7 @@ #define MI_NOOP (CMD_MI | 0) #define MI_BATCH_BUFFER_END (CMD_MI | (0xA << 23)) +#define MI_BATCH_BUFFER_START (CMD_MI | (0x31 << 23)) #define MI_FLUSH (CMD_MI | (0x4 << 23)) #define STATE_INSTRUCTION_CACHE_INVALIDATE (0x1 << 0) @@ -85,6 +86,7 @@ struct intel_driver_data int locked; struct intel_batchbuffer *batch; + struct intel_batchbuffer *batch_bcs; dri_bufmgr *bufmgr; }; @@ -118,8 +120,8 @@ struct intel_region #define PCI_CHIP_G45_G 0x2E22 #define PCI_CHIP_G41_G 0x2E32 -#define PCI_CHIP_IGDNG_D_G 0x0042 -#define PCI_CHIP_IGDNG_M_G 0x0046 +#define PCI_CHIP_IRONLAKE_D_G 0x0042 +#define PCI_CHIP_IRONLAKE_M_G 0x0046 #define IS_G45(devid) (devid == PCI_CHIP_IGD_E_G || \ devid == PCI_CHIP_Q45_G || \ @@ -128,8 +130,8 @@ struct intel_region #define IS_GM45(devid) (devid == PCI_CHIP_GM45_GM) #define IS_G4X(devid) (IS_G45(devid) || IS_GM45(devid)) -#define IS_IGDNG_D(devid) (devid == PCI_CHIP_IGDNG_D_G) -#define IS_IGDNG_M(devid) (devid == PCI_CHIP_IGDNG_M_G) -#define IS_IGDNG(devid) (IS_IGDNG_D(devid) || IS_IGDNG_M(devid)) +#define IS_IRONLAKE_D(devid) (devid == PCI_CHIP_IRONLAKE_D_G) +#define IS_IRONLAKE_M(devid) (devid == PCI_CHIP_IRONLAKE_M_G) +#define IS_IRONLAKE(devid) (IS_IRONLAKE_D(devid) || IS_IRONLAKE_M(devid)) #endif /* _INTEL_DRIVER_H_ */ diff --git a/object_heap.c b/object_heap.c index 1900f8c..e867139 100644 --- a/object_heap.c +++ b/object_heap.c @@ -50,7 +50,6 @@ static int object_heap_expand( object_heap_p heap ) { return -1; /* Out of memory */ } - memset(new_heap_index + heap->heap_size*heap->object_size, 0, heap->heap_increment * new_heap_size); heap->heap_index = new_heap_index; next_free = heap->next_free; for(i = new_heap_size; i-- > heap->heap_size; ) diff --git a/shaders/Makefile.am b/shaders/Makefile.am index 24f1aa8..2fd019b 100644 --- a/shaders/Makefile.am +++ b/shaders/Makefile.am @@ -1 +1 @@ -SUBDIRS = mpeg2 render +SUBDIRS = h264 mpeg2 render diff --git a/shaders/h264/Makefile.am b/shaders/h264/Makefile.am new file mode 100644 index 0000000..d6d106b --- /dev/null +++ b/shaders/h264/Makefile.am @@ -0,0 +1 @@ +SUBDIRS = mc diff --git a/shaders/h264/ildb/AVC_ILDB.inc b/shaders/h264/ildb/AVC_ILDB.inc new file mode 100644 index 0000000..734717f --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB.inc @@ -0,0 +1,718 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__AVC_ILDB_HEADER__) // Make sure this file is only included once
+#define __AVC_ILDB_HEADER__
+
+// Module name: AVC_ILDB.inc
+
+#undef ORIX
+#undef ORIY
+
+//========== Root thread input parameters ==================================================
+#define RootParam r1 // :w
+#define MBsCntX r1.0 // :w, MB count per row
+#define MBsCntY r1.1 // :w, MB count per col
+//#define PicType r1.2 // :w, Picture type
+#define MaxThreads r1.3 // :w, Max Thread limit
+#define EntrySignature r1.4 // :w, Debug flag
+#define BitFields r1.5 // :uw
+#define MbaffFlag BIT0 // :w, mbaff flag, bit 0 in BitFields
+#define BotFieldFlag BIT1 // :w, bottom field flag, bit 1 in BitFields
+#define CntlDataExpFlag BIT2 // :w, Control Data Expansion Flag, bit 2 in BitFields
+#define RampConst r1.12 // 8 :ub, Ramp constant, r1.12 - r1.19:ub
+#define StepToNextMB r1.20 // :b, 2 bytes
+#define Minus2Minus1 r1.22 // :b, 2 bytes
+// next one starts at r1.11:w
+
+#define TopFieldFlag 0xFFFD // :w, top field flag, used to set bit1 to 0.
+
+
+//========== Root Locals =============================================================
+
+// Variables in root kernel for launching child therad
+#define ChildParam r2.0 // :w
+//Not used #define URBOffset r2.3 // :w, Each row occupies 4 URB entries. All children in the same row use the same set of URB entries
+#define CurCol r2.10 // :w, current col
+#define CurColB r2.20 // :b, current col
+#define CurRow r2.11 // :w, current row
+#define CurRowB r2.22 // :b, current row
+#define LastCol r2.12 // :w, last col
+#define LastRow r2.13 // :w, last row
+
+// Root local constants during spawning process
+#define Col_Boundary r3.0 // :w,
+#define Row_Boundary r3.1 // :w,
+//#define TotalBlocks r3.2 // :w, Total blocks in the frame
+#define URB_EntriesPerMB_2 r3.3 // :w, = URB entries per MB, but in differnt form
+#define URBOffsetUVBase r3.4 // :w, UV Base offset in URB
+
+#define Temp1_D r3.6 // :d:
+#define Temp1_W r3.12 // :w, Temp1
+#define Temp1_B r3.24 // :b, = Temp1_W
+#define Temp2_W r3.13 // :w, Temp2
+#define Temp2_B r3.26 // :b, = Temp2_W
+
+// Root local variables
+#define JumpTable r4 // :d, jump table
+#define JUMPTABLE_BASE 4*32
+#define JumpAddr a0.7
+
+#define TopRowForScan r5.0 // :w, track the top row for scan. All rows above this row is deblocked already.
+
+
+// Child Thread R0 Header Field
+#define MRF0 m0
+#define CT_R0Hdr m1
+
+/*
+.declare GatewayAperture Base=r50.0 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+#define GatewayApertureB 1600 // r50 byte offset from r0.0
+
+// Chroma root thread updates luma root's ThreadLimit at r10.0:w via gateway
+#define ThreadLimit r62.0 // :w, thread limit //r56.0
+#define THREAD_LIMIT_OFFSET 0x01800000 // Offset from r50 to r56 = 12*32 = 384 = 0x0180. 0x180 << 16 = 0x01800000
+ //#define THREAD_LIMIT_OFFSET 0x00C00000 // Offset from r50 to r56 = 6*32 = 192 = 0x00C0. 0xC0 << 16 = 0x00C00000
+*/
+
+// Gateway size is 16 GRF. 68 rows of MBs takes 9 GRFs (r6 - r14)
+// For CTG: Expended to support 1280 rows of pixel (80 rows of MBs). It requires 10 GRFs (r6 - r15)
+.declare GatewayAperture Base=r6.0 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+#define GatewayApertureB 192 // r0.0 byte offset from r0.0
+
+// Chroma root thread updates luma root's ThreadLimit at r10.0:w via gateway
+#define ThreadLimit r18.0 // :w, thread limit
+#define THREAD_LIMIT_OFFSET 0x01800000 // Offset from r50 to r56 = 12*32 = 384 = 0x0180. 0x180 << 16 = 0x01800000
+#define TotalBlocks r18.1 // :w, Total blocks in the frame
+
+// Root local variables
+#define ChildThreadsID r19.0 // :w, Child thread ID, unique to each child
+#define OutstandingThreads r20.0 // :w, Outstanding threads
+#define ProcessedMBs r20.1 // :w, # of MBs processed
+
+#define URBOffset r21.0 // :w, Each row occupies 4 URB entries. All children in the same row use the same set of URB entries
+
+//=================================================================================
+
+#define ScoreBd_Size 128 //96 // size of Status[] or ProcCol[]
+
+#define ScoreBd_Idx 2
+//#define Saved_Col 0
+
+#define StatusAddr a0.4 // :w, point to r50
+//=================================================================================
+
+
+// Gateway payload
+#define GatewayPayload r48.0 // :ud
+#define GatewayPayloadKey r48.8 // :uw
+#define DispatchID r48.20 // :ub
+#define RegBase_GatewaySize r48.5 // :ud, used in open a gateway
+#define Offset_Length r48.5 // :ud, used in forwardmsg back to root
+#define EUID_TID r48.9 // :uw, used in forwardmsg back to root
+
+// Gateway response
+#define GatewayResponse r49.0 // :ud, one GRF
+
+#define URBWriteMsgDesc a0.0 // Used in URB write, :ud
+#define URBWriteMsgDescLow a0.0 // Used in URB write, :uw
+#define URBWriteMsgDescHigh a0.1 // Used in URB write, :uw
+
+.declare WritebackResponse Base=r50 ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 1 GRF for write backs
+
+
+/////////////////////////////////////////////////////////////////////////////////////////////
+// IDesc Order Offset
+//
+// 0) luma root 0 from luma root
+// 1) luma child 16 from luma root
+// 2) chroma root 32 from luma root
+// 3) chroma child 16 from chroma root
+
+// 4) luma field root 0 from luma field root
+// 5) luma field child 16 from luma field root
+// 6) chroma field root 32 from luma field root
+// 7) chroma field child 16 from chroma field root
+
+// 8) luma Mbaff root 0 from luma Mbaff root
+// 9) luma Mbaff child 16 from luma Mbaff root
+// 10) chroma Mbaff root 32 from luma Mbaff root
+// 11) chroma Mbaff child 16 from chroma Mbaff root
+
+// IDesc offset within non-mbaff or mbaff mode
+#define CHROMA_ROOT_OFFSET 32 // Offset from luma root to chroma root
+#define CHILD_OFFSET 16 // Offset from luma root to luma child,
+ // and from chroma root to chroma child
+/////////////////////////////////////////////////////////////////////////////////////////////
+
+
+//========== End of Root Variables ======================================================
+
+
+//========== Child thread input parameters ==============================================
+//#define MBsCntX r1.0 // :w, MB count per row (same as root)
+//#define MBsCntY r1.1 // :w, MB count per col (same as root)
+//#define PicTypeC r1.2 // :w, Picture type same as root thread (same as root)
+#define URBOffsetC r1.3 // :w,
+#define EntrySignatureC r1.4 // :w, Debug field (same as root)
+//#define BitFields r1.5 // :w (same as root)
+//#define MbaffFlag BIT0 // :w, mbaff flag, bit 0 in BitFields
+//#define BotFieldFlag BIT1 // :w, bottom field flag, bit 1 in BitFields
+//#define CntlDataExpFlag BIT2 // :w, Control Data Expansion Flag, bit 2 in BitFields
+#define RampConstC r1.12 // 8 :ub, Ramp constant, r1.12 - r1.19:ub.
+#define ORIX r1.10 // :w, carry over from root r1 in MB count
+#define ORIY r1.11 // :w, carry over from root r1 in MB count
+#define LastColC r1.12 // :w, last col
+#define LastRowC r1.13 // :w, last row
+
+.declare GatewayApertureC Base=r1.0 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+#define GatewayApertureCBase 32 // r1 byte offset from r0.0
+
+
+//========== Child Variables ============================================================
+
+// Mbaff Alpha, Beta, Tc0 vectors for an edge
+.declare Mbaff_ALPHA Base=r14.0 ElementSize=2 SrcRegion=REGION(8,1) Type=uw // r14
+.declare Mbaff_BETA Base=r15.0 ElementSize=2 SrcRegion=REGION(8,1) Type=uw // r15
+.declare Mbaff_TC0 Base=r16.0 ElementSize=2 SrcRegion=REGION(8,1) Type=uw // r16
+.declare RRampW Base=r17.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w // r17
+
+.declare Mbaff_ALPHA2 Base=r45.0 ElementSize=2 SrcRegion=REGION(8,1) Type=uw // alpha2 = (alpha >> 2) + 2
+
+
+#define ORIX_CUR r46.0 // :w, current block origin X in bytes
+#define ORIY_CUR r46.1 // :w, current block origin Y in bytes
+#define ORIX_LEFT r46.2 // :w, left block origin X in bytes
+#define ORIY_LEFT r46.3 // :w, left block origin Y in bytes
+#define ORIX_TOP r46.4 // :w, top block origin X in bytes
+#define ORIY_TOP r46.5 // :w, top block origin Y in bytes
+//#define FilterSampleFlag r46.6 // :uw,
+#define CTemp0_W r46.7 // :w, child Temp0
+
+#define alpha r46.8 // :w, Scaler version for non Mbaff
+#define beta r46.9 // :w, Scaler version for non Mbaff
+#define tc0 r46.20 // 4 :ub, r46.20 ~ r46.23, Scaler version for non Mbaff
+#define MaskA r46.12 // :uw
+#define MaskB r46.13 // :uw
+
+// Child control flags
+#define DualFieldMode r47.0 // Cur MB is frame based, above MB is field based in mbaff mode
+ // :uw, 0 = not in dual field mode, 1 = in dual field mode, filter both top and bot fields
+#define GateWayOffsetC r47.1 // :w, Gateway offset for child writing into root space
+#define CntrlDataOffsetY r47.1 // :ud, MB control data data offset
+#define alpha2 r47.4 // :uw, alpha2 = (alpha >> 2) + 2
+
+#define VertEdgePattern r47.5 // :uw,
+
+#define CTemp1_W r47.6 // :w, child Temp1
+#define CTemp1_B r47.12 // :b, = child Temp1_W
+#define CTemp2_W r47.7 // :w, child Temp2
+#define CTemp2_B r47.14 // :b, = child Temp2_W
+
+// Used in child
+#define ECM_AddrReg a0.4 // Edge Control Map register
+#define P_AddrReg a0.6 // point to P samples in left or top MB
+#define Q_AddrReg a0.7 // point to Q samples in cur MB
+
+
+.declare RTempD Base=r26.0 ElementSize=4 SrcRegion=REGION(8,1) Type=d // r26-27
+.declare RTempB Base=r26.0 ElementSize=1 SrcRegion=REGION(8,4) Type=ub // r26-27
+.declare RTempW Base=r26.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w // r26-27
+#define LEFT_TEMP_D RTempD
+#define LEFT_TEMP_B RTempB
+#define LEFT_TEMP_W RTempW
+
+.declare TempRow0 Base=r26.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare TempRow0B Base=r26.0 ElementSize=1 SrcRegion=REGION(8,2) Type=ub
+.declare TempRow1 Base=r27.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare TempRow1B Base=r27.0 ElementSize=1 SrcRegion=REGION(8,2) Type=ub
+
+.declare CUR_TEMP_D Base=r28.0 ElementSize=4 SrcRegion=REGION(8,1) Type=d // 8 GRFs
+.declare CUR_TEMP_B Base=r28.0 ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+.declare CUR_TEMP_W Base=r28.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+#define FilterSampleFlag r28.0 // :uw,
+
+.declare A Base=r28.0 ElementSize=2 SrcRegion=REGION(16,1) Type=w
+.declare B Base=r29.0 ElementSize=2 SrcRegion=REGION(16,1) Type=w
+
+.declare TempRow3 Base=r30.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare TempRow3B Base=r30.0 ElementSize=1 SrcRegion=REGION(8,2) Type=ub
+
+.declare tc0_exp Base=r30.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare tc8 Base=r30.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+.declare tc_exp Base=r31.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare tx_exp_8 Base=r31.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+.declare q0_p0 Base=r32.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare ABS_q0_p0 Base=r33.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+.declare ap Base=r34.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare aq Base=r35.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+// These buffers have the src data for each edge to be beblocked.
+// They have modified pixels from previous edges.
+//
+// Y:
+// +----+----+----+----+----+----+----+----+
+// | p3 | p2 | P1 | p0 | q0 | q1 | q2 | q3 |
+// +----+----+----+----+----+----+----+----+
+//
+// p3 = r[P_AddrReg, 0]<16;16,1>
+// p2 = r[P_AddrReg, 16]<16;16,1>
+// p1 = r[P_AddrReg, 32]<16;16,1>
+// p0 = r[P_AddrReg, 48]<16;16,1>
+// q0 = r[Q_AddrReg, 0]<16;16,1>
+// q1 = r[Q_AddrReg, 16]<16;16,1>
+// q2 = r[Q_AddrReg, 32]<16;16,1>
+// q3 = r[Q_AddrReg, 48]<16;16,1>
+
+.declare p0123_W Base=r36.0 ElementSize=2 SrcRegion=REGION(16,1) Type=uw // r36, r37
+.declare q0123_W Base=r38.0 ElementSize=2 SrcRegion=REGION(16,1) Type=uw // r38, r39
+.declare p3 Base=r36.0 ElementSize=1 SrcRegion=REGION(8,1) Type=ub
+.declare p2 Base=r36.16 ElementSize=1 SrcRegion=REGION(8,1) Type=ub
+.declare p1 Base=r37.0 ElementSize=1 SrcRegion=REGION(8,1) Type=ub
+.declare p0 Base=r37.16 ElementSize=1 SrcRegion=REGION(8,1) Type=ub
+.declare q0 Base=r38.0 ElementSize=1 SrcRegion=REGION(8,1) Type=ub
+.declare q1 Base=r38.16 ElementSize=1 SrcRegion=REGION(8,1) Type=ub
+.declare q2 Base=r39.0 ElementSize=1 SrcRegion=REGION(8,1) Type=ub
+.declare q3 Base=r39.16 ElementSize=1 SrcRegion=REGION(8,1) Type=ub
+
+.declare TempRow2 Base=r38.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+// Temp space for mbaff dual field mode
+#define ABOVE_CUR_MB_BASE 40*GRFWIB // Byte offset to r40
+.declare ABOVE_CUR_MB_YW Base=r40 ElementSize=2 SrcRegion=REGION(8,1) Type=uw
+.declare ABOVE_CUR_MB_UW Base=r40 ElementSize=2 SrcRegion=REGION(8,1) Type=uw
+
+.declare P0_plus_P1 Base=r41.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare Q0_plus_Q1 Base=r42.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+.declare P2_plus_P3 Base=r43.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare Q2_plus_Q3 Base=r44.0 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+
+//////////////////////////////////////////////////////////////////////////////////////////
+// MB control data reference
+
+// Expanded control data is in r18 - r25
+.declare CNTRL_DATA_D Base=r18 ElementSize=4 SrcRegion=REGION(8,1) Type=ud // For read, 8 GRFs
+#define CNTRL_DATA_BASE 18*GRFWIB // Base offset to r18
+
+// Bit mask for extracting bits
+#define MbaffFrameFlag 0x01
+#define FieldModeCurrentMbFlag 0x02
+#define FieldModeLeftMbFlag 0x04
+#define FieldModeAboveMbFlag 0x08
+#define FilterInternal8x8EdgesFlag 0x10
+#define FilterInternal4x4EdgesFlag 0x20
+#define FilterLeftMbEdgeFlag 0x40
+#define FilterTopMbEdgeFlag 0x80
+
+#define DISABLE_ILDB_FLAG 0x01
+
+// Exact bit pattern for left and cur MB coding mode (frame vs. field)
+#define LEFT_FRAME_CUR_FRAME 0x00
+#define LEFT_FRAME_CUR_FIELD 0x02
+#define LEFT_FIELD_CUR_FRAME 0x04
+#define LEFT_FIELD_CUR_FIELD 0x06
+
+// Exact bit pattern for above and cur MB coding mode (frame vs. field)
+#define ABOVE_FRAME_CUR_FRAME 0x00
+#define ABOVE_FRAME_CUR_FIELD 0x02
+#define ABOVE_FIELD_CUR_FRAME 0x08
+#define ABOVE_FIELD_CUR_FIELD 0x0A
+
+
+
+//========== MB control data field offset in byte ==========
+
+#if !defined(_APPLE)
+
+// GRF0 - GRF1 holds original control data
+
+// GRF0
+#define HorizOrigin 0
+#define VertOrigin 1
+#define BitFlags 2 // Bit flags
+
+#define bbSinternalLeftVert 4 // Internal left vertical bS, 2 bits per bS for 4 Y pixels and 2 U/V pixels
+#define bbSinternalMidVert 5 // Internal mid vertical bS
+#define bbSinternalRightVert 6 // Internal right vertical bS
+#define bbSinternalTopHorz 7 // Internal top horizontal bS
+
+#define bbSinternalMidHorz 8 // Internal mid horizontal bS
+#define bbSinternalBotHorz 9 // Internal bottom horizontal bS
+#define wbSLeft0 10 // External left vertical bS (0), 4 bits per bS for 4 Y pixels and 2 U/V pixels, and byte 11
+
+#define wbSLeft1 12 // External left vertical bS (1), and byte 13
+#define wbSTop0 14 // External top horizontal bS (0), and byte 15
+
+#define wbSTop1 16 // Externaltop horizontal bS (1), and byte 17
+#define bIndexAinternal_Y 18 // Internal index A for Y
+#define bIndexBinternal_Y 19 // Internal index B for Y
+
+#define bIndexAleft0_Y 20 // Left index A for Y (0)
+#define bIndexBleft0_Y 21 // Left index B for Y (0)
+#define bIndexAleft1_Y 22 // Left index A for Y (1)
+#define bIndexBleft1_Y 23 // Left index B for Y (1)
+
+#define bIndexAtop0_Y 24 // Top index A for Y (0)
+#define bIndexBtop0_Y 25 // Top index B for Y (0)
+#define bIndexAtop1_Y 26 // Top index A for Y (1)
+#define bIndexBtop1_Y 27 // Top index B for Y (1)
+
+#define bIndexAinternal_Cb 28 // Internal index A for Cb
+#define bIndexBinternal_Cb 29 // Internal index B for Cb
+#define bIndexAleft0_Cb 30 // Left index A for Cb (0)
+#define bIndexBleft0_Cb 31 // Left index B for Cb (0)
+
+// GRF1
+#define bIndexAleft1_Cb 32 // Left index A for Cb (1)
+#define bIndexBleft1_Cb 33 // Left index B for Cb (1)
+#define bIndexAtop0_Cb 34 // Top index A for Cb (0)
+#define bIndexBtop0_Cb 35 // Top index B for Cb (0)
+
+#define bIndexAtop1_Cb 36 // Top index A for Cb (1)
+#define bIndexBtop1_Cb 37 // Top index B for Cb (1)
+#define bIndexAinternal_Cr 38 // Internal index A for Cr
+#define bIndexBinternal_Cr 39 // Internal index B for Cr
+
+#define bIndexAleft0_Cr 40 // Left index A for Cr (0)
+#define bIndexBleft0_Cr 41 // Left index B for Cr (0)
+#define bIndexAleft1_Cr 42 // Left index A for Cr (1)
+#define bIndexBleft1_Cr 43 // Left index B for Cr (1)
+
+#define bIndexAtop0_Cr 44 // Top index A for Cr (0)
+#define bIndexBtop0_Cr 45 // Top index B for Cr (0)
+#define bIndexAtop1_Cr 46 // Top index A for Cr (1)
+#define bIndexBtop1_Cr 47 // Top index B for Cr (1)
+
+#define ExtBitFlags 48 // Extended bit flags, such as disable ILDB bits
+
+// Offset 49 - 63 not used
+
+
+//===== GRF2 - GRF7 hold expanded control data =====
+
+// GRF2
+#define wEdgeCntlMap_IntLeftVert 64 // Derived from bbSinternalLeftVert, 1 bit per pixel
+#define wEdgeCntlMap_IntMidVert 66 // Derived from bbSinternalLeftVert
+
+#define wEdgeCntlMap_IntRightVert 68 // Derived from bbSinternalRightVert
+#define wEdgeCntlMap_IntTopHorz 70 // Derived from bbSinternalTopHorz, 1bit per pixel
+
+#define wEdgeCntlMap_IntMidHorz 72 // Derived from bbSinternalMidHorz
+#define wEdgeCntlMap_IntBotHorz 74 // Derived from bbSinternalBotHorz
+
+// Offset 76 - 79 not used
+
+#define wEdgeCntlMapA_ExtLeftVert0 80 // Derived from wbSLeft0, 1bit per pixel
+#define wEdgeCntlMapB_ExtLeftVert0 82 // Derived from wbSLeft0
+
+#define wEdgeCntlMapA_ExtTopHorz0 84 // Derived from wbSTop0, 1bit per pixel
+#define wEdgeCntlMapB_ExtTopHorz0 86 // Derived from wbSTop0
+
+#define wEdgeCntlMapA_ExtLeftVert1 88 // Derived from wbSLeft1, 1bit per pixel
+#define wEdgeCntlMapB_ExtLeftVert1 90 // Derived from wbSLeft1
+
+#define wEdgeCntlMapA_ExtTopHorz1 92 // Derived from wbSTop1, 1bit per pixel
+#define wEdgeCntlMapB_ExtTopHorz1 94 // Derived from wbSTop1
+
+
+// GRF3
+#define bTc0_v00_0_Y 96 // Derived from bSv00_0 and bIndexAleft0_Y, 4 pixels per tc0
+#define bTc0_v10_0_Y 97 // Derived from bSv10_0 and bIndexAleft0_Y
+#define bTc0_v20_0_Y 98 // Derived from bSv20_0 and bIndexAleft0_Y
+#define bTc0_v30_0_Y 99 // Derived from bSv30_0 and bIndexAleft0_Y
+
+#define bTc0_v01_Y 100 // Derived from bSv01 and bIndexAinternal_Y
+#define bTc0_v11_Y 101 // Derived from bSv11 and bIndexAinternal_Y
+#define bTc0_v21_Y 102 // Derived from bSv21 and bIndexAinternal_Y
+#define bTc0_v31_Y 103 // Derived from bSv31 and bIndexAinternal_Y
+
+#define bTc0_v02_Y 104 // Derived from bSv02 and bIndexAinternal_Y
+#define bTc0_v12_Y 105 // Derived from bSv12 and bIndexAinternal_Y
+#define bTc0_v22_Y 106 // Derived from bSv22 and bIndexAinternal_Y
+#define bTc0_v32_Y 107 // Derived from bSv32 and bIndexAinternal_Y
+
+#define bTc0_v03_Y 108 // Derived from bSv03 and bIndexAinternal_Y
+#define bTc0_v13_Y 109 // Derived from bSv13 and bIndexAinternal_Y
+#define bTc0_v23_Y 110 // Derived from bSv23 and bIndexAinternal_Y
+#define bTc0_v33_Y 111 // Derived from bSv33 and bIndexAinternal_Y
+
+#define bTc0_h00_0_Y 112 // Derived from bSh00_0 and bIndexAleft0_Y
+#define bTc0_h01_0_Y 113 // Derived from bSh01_0 and bIndexAleft0_Y
+#define bTc0_h02_0_Y 114 // Derived from bSh02_0 and bIndexAleft0_Y
+#define bTc0_h03_0_Y 115 // Derived from bSh03_0 and bIndexAleft0_Y
+
+#define bTc0_h10_Y 116 // Derived from bSh10 and bIndexAinternal_Y
+#define bTc0_h11_Y 117 // Derived from bSh11 and bIndexAinternal_Y
+#define bTc0_h12_Y 118 // Derived from bSh12 and bIndexAinternal_Y
+#define bTc0_h13_Y 119 // Derived from bSh13 and bIndexAinternal_Y
+
+#define bTc0_h20_Y 120 // Derived from bSh20 and bIndexAinternal_Y
+#define bTc0_h21_Y 121 // Derived from bSh21 and bIndexAinternal_Y
+#define bTc0_h22_Y 122 // Derived from bSh22 and bIndexAinternal_Y
+#define bTc0_h23_Y 123 // Derived from bSh23 and bIndexAinternal_Y
+
+#define bTc0_h30_Y 124 // Derived from bSh30 and bIndexAinternal_Y
+#define bTc0_h31_Y 125 // Derived from bSh31 and bIndexAinternal_Y
+#define bTc0_h32_Y 126 // Derived from bSh32 and bIndexAinternal_Y
+#define bTc0_h33_Y 127 // Derived from bSh33 and bIndexAinternal_Y
+
+// GRF4
+#define bAlphaLeft0_Y 128 // Derived from bIndexAleft0_Y
+#define bBetaLeft0_Y 129 // Derived from bIndexBleft0_Y
+#define bAlphaTop0_Y 130 // Derived from bIndexAtop0_Y
+#define bBetaTop0_Y 131 // Derived from bIndexBtop0_Y
+
+#define bAlphaInternal_Y 132 // Derived from bIndexAinternal_Y
+#define bBetaInternal_Y 133 // Derived from bIndexBinternal_Y
+// Offset 134 - 135 not used
+
+// Offset 136 - 143 not used
+#define bAlphaLeft1_Y 144 // Derived from bIndexAleft1_Y Used in Mbaff mode only
+#define bBetaLeft1_Y 145 // Derived from bIndexBleft1_Y Used in Mbaff mode only
+#define bAlphaTop1_Y 146 // Derived from bIndexAtop1_Y Used in Mbaff mode only
+#define bBetaTop1_Y 147 // Derived from bIndexBtop1_Y Used in Mbaff mode only
+
+// Offset 148 - 151 not used
+#define bTc0_v00_1_Y 152 // Derived from bSv00_1 and bIndexAleft1_Y Used in Mbaff mode only
+#define bTc0_v10_1_Y 153 // Derived from bSv10_1 and bIndexAleft1_Y Used in Mbaff mode only
+#define bTc0_v20_1_Y 154 // Derived from bSv20_1 and bIndexAleft1_Y Used in Mbaff mode only
+#define bTc0_v30_1_Y 155 // Derived from bSv30_1 and bIndexAleft1_Y Used in Mbaff mode only
+
+#define bTc0_h00_1_Y 156 // Derived from bSh00_1 and bIndexAleft1_Y Used in Mbaff mode only
+#define bTc0_h01_1_Y 157 // Derived from bSh01_1 and bIndexAleft1_Y Used in Mbaff mode only
+#define bTc0_h02_1_Y 158 // Derived from bSh02_1 and bIndexAleft1_Y Used in Mbaff mode only
+#define bTc0_h03_1_Y 159 // Derived from bSh03_1 and bIndexAleft1_Y Used in Mbaff mode only
+
+
+// GRF5
+#define bTc0_v00_0_Cb 160 // Derived from bSv00_0 and bIndexAleft0_Cb, 2 pixels per tc0 Left0
+#define bTc0_v10_0_Cb 161 // Derived from bSv10_0 and bIndexAleft0_Cb
+#define bTc0_v20_0_Cb 162 // Derived from bSv20_0 and bIndexAleft0_Cb
+#define bTc0_v30_0_Cb 163 // Derived from bSv30_0 and bIndexAleft0_Cb
+
+#define bTc0_v02_Cb 164 // Derived from bSv02 and bIndexAinternal_Cb MidVert
+#define bTc0_v12_Cb 165 // Derived from bSv12 and bIndexAinternal_Cb
+#define bTc0_v22_Cb 166 // Derived from bSv22 and bIndexAinternal_Cb
+#define bTc0_v32_Cb 167 // Derived from bSv32 and bIndexAinternal_Cb
+
+#define bTc0_h00_0_Cb 168 // Derived from bSh00_0 and bIndexAleft0_Cb Top0
+#define bTc0_h01_0_Cb 169 // Derived from bSh01_0 and bIndexAleft0_Cb
+#define bTc0_h02_0_Cb 170 // Derived from bSh02_0 and bIndexAleft0_Cb
+#define bTc0_h03_0_Cb 171 // Derived from bSh03_0 and bIndexAleft0_Cb
+
+#define bTc0_h20_Cb 172 // Derived from bSh20 and bIndexAinternal_Cb MidHorz
+#define bTc0_h21_Cb 173 // Derived from bSh21 and bIndexAinternal_Cb
+#define bTc0_h22_Cb 174 // Derived from bSh22 and bIndexAinternal_Cb
+#define bTc0_h23_Cb 175 // Derived from bSh23 and bIndexAinternal_Cb
+
+#define bTc0_v00_0_Cr 176 // Derived from bSv00_0 and bIndexAleft0_Cr, 2 pixels per tc0 Left0
+#define bTc0_v10_0_Cr 177 // Derived from bSv10_0 and bIndexAleft0_Cr
+#define bTc0_v20_0_Cr 178 // Derived from bSv20_0 and bIndexAleft0_Cr
+#define bTc0_v30_0_Cr 179 // Derived from bSv30_0 and bIndexAleft0_Cr
+
+#define bTc0_v02_Cr 180 // Derived from bSv02 and bIndexAinternal_Cr Mid Vert
+#define bTc0_v12_Cr 181 // Derived from bSv12 and bIndexAinternal_Cr
+#define bTc0_v22_Cr 182 // Derived from bSv22 and bIndexAinternal_Cr
+#define bTc0_v32_Cr 183 // Derived from bSv32 and bIndexAinternal_Cr
+
+#define bTc0_h00_0_Cr 184 // Derived from bSh00_0 and bIndexAleft0_Cr, 2 pixels per tc0 Top0
+#define bTc0_h01_0_Cr 185 // Derived from bSh01_0 and bIndexAleft0_Cr
+#define bTc0_h02_0_Cr 186 // Derived from bSh02_0 and bIndexAleft0_Cr
+#define bTc0_h03_0_Cr 187 // Derived from bSh03_0 and bIndexAleft0_Cr
+
+#define bTc0_h20_Cr 188 // Derived from bSh20 and bIndexAinternal_Cr Mid Horz
+#define bTc0_h21_Cr 189 // Derived from bSh21 and bIndexAinternal_Cr
+#define bTc0_h22_Cr 190 // Derived from bSh22 and bIndexAinternal_Cr
+#define bTc0_h23_Cr 191 // Derived from bSh23 and bIndexAinternal_Cr
+
+// GRF6
+#define bAlphaLeft0_Cb 192 // Derived from bIndexAleft0_Cb
+#define bBetaLeft0_Cb 193 // Derived from bIndexBleft0_Cb
+#define bAlphaTop0_Cb 194 // Derived from bIndexAtop0_Cb
+#define bBetaTop0_Cb 195 // Derived from bIndexBtop0_Cb
+
+#define bAlphaInternal_Cb 196 // Derived from bIndexAinternal_Cb
+#define bBetaInternal_Cb 197 // Derived from bIndexBinternal_Cb
+// Offset 198 - 199 not used
+
+#define bAlphaLeft0_Cr 200 // Derived from bIndexAleft0_Cr
+#define bBetaLeft0_Cr 201 // Derived from bIndexBleft0_Cr
+#define bAlphaTop0_Cr 202 // Derived from bIndexAtop0_Cr
+#define bBetaTop0_Cr 203 // Derived from bIndexBtop0_Cr
+
+#define bAlphaInternal_Cr 204 // Derived from bIndexAinternal_Cr
+#define bBetaInternal_Cr 205 // Derived from bIndexBinternal_Cr
+// Offset 206 - 223 not used
+
+// GRF7
+#define bAlphaLeft1_Cb 224 // Derived from bIndexAleft1_Cb Used in Mbaff mode only
+#define bBetaLeft1_Cb 225 // Derived from bIndexBleft1_Cb Used in Mbaff mode only
+#define bAlphaTop1_Cb 226 // Derived from bIndexAtop1_Cb Used in Mbaff mode only
+#define bBetaTop1_Cb 227 // Derived from bIndexBtop1_Cb Used in Mbaff mode only
+
+// Offset 228 - 231 not used
+
+#define bTc0_v00_1_Cb 232 // Derived from bSv00_1 and bIndexAleft1_Cb Used in Mbaff mode only
+#define bTc0_v10_1_Cb 233 // Derived from bSv10_1 and bIndexAleft1_Cb Used in Mbaff mode only
+#define bTc0_v20_1_Cb 234 // Derived from bSv20_1 and bIndexAleft1_Cb Used in Mbaff mode only
+#define bTc0_v30_1_Cb 235 // Derived from bSv30_1 and bIndexAleft1_Cb Used in Mbaff mode only
+
+#define bTc0_h00_1_Cb 236 // Derived from bSh00_1 and bIndexAleft1_Cb Used in Mbaff mode only
+#define bTc0_h01_1_Cb 237 // Derived from bSh01_1 and bIndexAleft1_Cb Used in Mbaff mode only
+#define bTc0_h02_1_Cb 238 // Derived from bSh02_1 and bIndexAleft1_Cb Used in Mbaff mode only
+#define bTc0_h03_1_Cb 239 // Derived from bSh03_1 and bIndexAleft1_Cb Used in Mbaff mode only
+
+#define bAlphaLeft1_Cr 240 // Derived from bIndexAleft1_Cr Used in Mbaff mode only
+#define bBetaLeft1_Cr 241 // Derived from bIndexBleft1_Cr Used in Mbaff mode only
+#define bAlphaTop1_Cr 242 // Derived from bIndexAtop1_Cr Used in Mbaff mode only
+#define bBetaTop1_Cr 243 // Derived from bIndexBtop1_Cr Used in Mbaff mode only
+
+// Offset 244 - 247 not used
+
+#define bTc0_v00_1_Cr 248 // Derived from bSv00_1 and bIndexAleft1_Cr Used in Mbaff mode only
+#define bTc0_v10_1_Cr 249 // Derived from bSv10_1 and bIndexAleft1_Cr Used in Mbaff mode only
+#define bTc0_v20_1_Cr 250 // Derived from bSv20_1 and bIndexAleft1_Cr Used in Mbaff mode only
+#define bTc0_v30_1_Cr 251 // Derived from bSv30_1 and bIndexAleft1_Cr Used in Mbaff mode only
+
+#define bTc0_h00_1_Cr 252 // Derived from bSh00_1 and bIndexAleft1_Cr Used in Mbaff mode only
+#define bTc0_h01_1_Cr 253 // Derived from bSh01_1 and bIndexAleft1_Cr Used in Mbaff mode only
+#define bTc0_h02_1_Cr 254 // Derived from bSh02_1 and bIndexAleft1_Cr Used in Mbaff mode only
+#define bTc0_h03_1_Cr 255 // Derived from bSh03_1 and bIndexAleft1_Cr Used in Mbaff mode only
+
+
+#else // _APPLE is defined
+
+//******** Crestline for Apple, progressive only, 88 bytes **********
+
+// GRF0
+#define HorizOrigin 0
+#define VertOrigin 1
+#define BitFlags 2 // Bit flags
+
+#define wEdgeCntlMap_IntLeftVert 4 // Derived from bbSinternalLeftVert, 1 bit per pixel
+#define wEdgeCntlMap_IntMidVert 6 // Derived from bbSinternalLeftVert
+#define wEdgeCntlMap_IntRightVert 8 // Derived from bbSinternalRightVert
+#define wEdgeCntlMap_IntTopHorz 10 // Derived from bbSinternalTopHorz, 1bit per pixel
+#define wEdgeCntlMap_IntMidHorz 12 // Derived from bbSinternalMidHorz
+#define wEdgeCntlMap_IntBotHorz 14 // Derived from bbSinternalBotHorz
+#define wEdgeCntlMapA_ExtLeftVert0 16 // Derived from wbSLeft0, 1bit per pixel
+#define wEdgeCntlMapB_ExtLeftVert0 18 // Derived from wbSLeft0
+#define wEdgeCntlMapA_ExtTopHorz0 20 // Derived from wbSTop0, 1bit per pixel
+#define wEdgeCntlMapB_ExtTopHorz0 22 // Derived from wbSTop0
+
+#define bAlphaLeft0_Y 24 // Derived from bIndexAleft0_Y
+#define bBetaLeft0_Y 25 // Derived from bIndexBleft0_Y
+#define bAlphaTop0_Y 26 // Derived from bIndexAtop0_Y
+#define bBetaTop0_Y 27 // Derived from bIndexBtop0_Y
+#define bAlphaInternal_Y 28 // Derived from bIndexAinternal_Y
+#define bBetaInternal_Y 29 // Derived from bIndexBinternal_Y
+
+// GRF1
+#define bTc0_v00_0_Y 32 // Derived from bSv00_0 and bIndexAleft0_Y, 4 pixels per tc0
+#define bTc0_v10_0_Y 33 // Derived from bSv10_0 and bIndexAleft0_Y
+#define bTc0_v20_0_Y 34 // Derived from bSv20_0 and bIndexAleft0_Y
+#define bTc0_v30_0_Y 35 // Derived from bSv30_0 and bIndexAleft0_Y
+#define bTc0_v01_Y 36 // Derived from bSv01 and bIndexAinternal_Y
+#define bTc0_v11_Y 37 // Derived from bSv11 and bIndexAinternal_Y
+#define bTc0_v21_Y 38 // Derived from bSv21 and bIndexAinternal_Y
+#define bTc0_v31_Y 39 // Derived from bSv31 and bIndexAinternal_Y
+#define bTc0_v02_Y 40 // Derived from bSv02 and bIndexAinternal_Y
+#define bTc0_v12_Y 41 // Derived from bSv12 and bIndexAinternal_Y
+#define bTc0_v22_Y 42 // Derived from bSv22 and bIndexAinternal_Y
+#define bTc0_v32_Y 43 // Derived from bSv32 and bIndexAinternal_Y
+#define bTc0_v03_Y 44 // Derived from bSv03 and bIndexAinternal_Y
+#define bTc0_v13_Y 45 // Derived from bSv13 and bIndexAinternal_Y
+#define bTc0_v23_Y 46 // Derived from bSv23 and bIndexAinternal_Y
+#define bTc0_v33_Y 47 // Derived from bSv33 and bIndexAinternal_Y
+
+#define bTc0_h00_0_Y 48 // Derived from bSh00_0 and bIndexAleft0_Y
+#define bTc0_h01_0_Y 49 // Derived from bSh01_0 and bIndexAleft0_Y
+#define bTc0_h02_0_Y 50 // Derived from bSh02_0 and bIndexAleft0_Y
+#define bTc0_h03_0_Y 51 // Derived from bSh03_0 and bIndexAleft0_Y
+#define bTc0_h10_Y 52 // Derived from bSh10 and bIndexAinternal_Y
+#define bTc0_h11_Y 53 // Derived from bSh11 and bIndexAinternal_Y
+#define bTc0_h12_Y 54 // Derived from bSh12 and bIndexAinternal_Y
+#define bTc0_h13_Y 55 // Derived from bSh13 and bIndexAinternal_Y
+#define bTc0_h20_Y 56 // Derived from bSh20 and bIndexAinternal_Y
+#define bTc0_h21_Y 57 // Derived from bSh21 and bIndexAinternal_Y
+#define bTc0_h22_Y 58 // Derived from bSh22 and bIndexAinternal_Y
+#define bTc0_h23_Y 59 // Derived from bSh23 and bIndexAinternal_Y
+#define bTc0_h30_Y 60 // Derived from bSh30 and bIndexAinternal_Y
+#define bTc0_h31_Y 61 // Derived from bSh31 and bIndexAinternal_Y
+#define bTc0_h32_Y 62 // Derived from bSh32 and bIndexAinternal_Y
+#define bTc0_h33_Y 63 // Derived from bSh33 and bIndexAinternal_Y
+
+// GRF2,
+#define bTc0_v00_0_Cb 64 // Derived from bSv00_0 and bIndexAleft0_Cb, 2 pixels per tc0 Left0
+#define bTc0_v10_0_Cb 65 // Derived from bSv10_0 and bIndexAleft0_Cb
+#define bTc0_v20_0_Cb 66 // Derived from bSv20_0 and bIndexAleft0_Cb
+#define bTc0_v30_0_Cb 67 // Derived from bSv30_0 and bIndexAleft0_Cb
+#define bTc0_v02_Cb 68 // Derived from bSv02 and bIndexAinternal_Cb MidVert
+#define bTc0_v12_Cb 69 // Derived from bSv12 and bIndexAinternal_Cb
+#define bTc0_v22_Cb 70 // Derived from bSv22 and bIndexAinternal_Cb
+#define bTc0_v32_Cb 71 // Derived from bSv32 and bIndexAinternal_Cb
+#define bTc0_h00_0_Cb 72 // Derived from bSh00_0 and bIndexAleft0_Cb Top0
+#define bTc0_h01_0_Cb 73 // Derived from bSh01_0 and bIndexAleft0_Cb
+#define bTc0_h02_0_Cb 74 // Derived from bSh02_0 and bIndexAleft0_Cb
+#define bTc0_h03_0_Cb 75 // Derived from bSh03_0 and bIndexAleft0_Cb
+#define bTc0_h20_Cb 76 // Derived from bSh20 and bIndexAinternal_Cb MidHorz
+#define bTc0_h21_Cb 77 // Derived from bSh21 and bIndexAinternal_Cb
+#define bTc0_h22_Cb 78 // Derived from bSh22 and bIndexAinternal_Cb
+#define bTc0_h23_Cb 79 // Derived from bSh23 and bIndexAinternal_Cb
+
+#define bAlphaLeft0_Cb 80 // Derived from bIndexAleft0_Cb
+#define bBetaLeft0_Cb 81 // Derived from bIndexBleft0_Cb
+#define bAlphaTop0_Cb 82 // Derived from bIndexAtop0_Cb
+#define bBetaTop0_Cb 83 // Derived from bIndexBtop0_Cb
+#define bAlphaInternal_Cb 84 // Derived from bIndexAinternal_Cb
+#define bBetaInternal_Cb 85 // Derived from bIndexBinternal_Cb
+
+#define ExtBitFlags 86 // Extended bit flags, such as disable ILDB bits
+
+// Shared between Cb and Cr
+#define bTc0_v00_0_Cr bTc0_v00_0_Cb
+#define bTc0_v10_0_Cr bTc0_v10_0_Cb
+#define bTc0_v20_0_Cr bTc0_v20_0_Cb
+#define bTc0_v30_0_Cr bTc0_v30_0_Cb
+#define bTc0_v02_Cr bTc0_v02_Cb
+#define bTc0_v12_Cr bTc0_v12_Cb
+#define bTc0_v22_Cr bTc0_v22_Cb
+#define bTc0_v32_Cr bTc0_v32_Cb
+#define bTc0_h00_0_Cr bTc0_h00_0_Cb
+#define bTc0_h01_0_Cr bTc0_h01_0_Cb
+#define bTc0_h02_0_Cr bTc0_h02_0_Cb
+#define bTc0_h03_0_Cr bTc0_h03_0_Cb
+#define bTc0_h20_Cr bTc0_h20_Cb
+#define bTc0_h21_Cr bTc0_h21_Cb
+#define bTc0_h22_Cr bTc0_h22_Cb
+#define bTc0_h23_Cr bTc0_h23_Cb
+
+#define bAlphaLeft0_Cr bAlphaLeft0_Cb
+#define bBetaLeft0_Cr bBetaLeft0_Cb
+#define bAlphaTop0_Cr bAlphaTop0_Cb
+#define bBetaTop0_Cr bBetaTop0_Cb
+#define bAlphaInternal_Cr bAlphaInternal_Cb
+#define bBetaInternal_Cr bBetaInternal_Cb
+
+
+#endif
+
+
+//========== End of Child Variables ===============================================================
+
+#if !defined(COMBINED_KERNEL)
+#define ILDB_LABEL(x) x // No symbol extension for standalone kernels
+#endif
+
+#endif // !defined(__AVC_ILDB_HEADER__)
diff --git a/shaders/h264/ildb/AVC_ILDB_Child_Field_UV.asm b/shaders/h264/ildb/AVC_ILDB_Child_Field_UV.asm new file mode 100644 index 0000000..b0986b5 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Child_Field_UV.asm @@ -0,0 +1,9 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#include "AVC_ILDB_Child_UV.asm"
diff --git a/shaders/h264/ildb/AVC_ILDB_Child_Field_Y.asm b/shaders/h264/ildb/AVC_ILDB_Child_Field_Y.asm new file mode 100644 index 0000000..fafd6c0 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Child_Field_Y.asm @@ -0,0 +1,9 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#include "AVC_ILDB_Child_Y.asm"
diff --git a/shaders/h264/ildb/AVC_ILDB_Child_Mbaff_UV.asm b/shaders/h264/ildb/AVC_ILDB_Child_Mbaff_UV.asm new file mode 100644 index 0000000..64dd802 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Child_Mbaff_UV.asm @@ -0,0 +1,173 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// AVC Child Kernel (Vertical and horizontal de-block a 4:2:0 MB UV comp)
+//
+// First de-block vertical edges from left to right.
+// Second de-block horizontal edge from top to bottom.
+//
+// For 4:2:0, chroma is always de-blocked at 8x8.
+// NV12 format allows to filter U and V together.
+//
+//////////////////////////////////////////////////////////////////////////////////////////////////////////////
+#define AVC_ILDB
+
+.kernel AVC_ILDB_CHILD_MBAFF_UV
+#if defined(COMBINED_KERNEL)
+ILDB_LABEL(AVC_ILDB_CHILD_UV):
+#endif
+
+#include "setupVPKernel.asm"
+#include "AVC_ILDB.inc"
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xE997:w
+#endif
+
+ // Setup temp buf used by load and save code
+ #define BUF_B RTempB
+ #define BUF_W RTempW
+ #define BUF_D RTempD
+
+ // Init local variables
+ mul (4) ORIX_CUR<2>:w ORIX<0;1,0>:w 16:w { NoDDClr } // Expand X addr to bytes, repeat 4 times
+ mul (4) ORIY_CUR<2>:w ORIY<0;1,0>:w 32:w { NoDDChk } // Expand Y addr to bytes, repeat 4 times
+
+ mov (2) f0.0<1>:w 0:w
+
+ mov (1) GateWayOffsetC:uw ORIY:uw // Use row # as Gateway offset
+
+ //=== Null Kernel ===============================================================
+// jmpi ILDB_LABEL(POST_ILDB_UV)
+ //===============================================================================
+
+ //====================================================================================
+ // Assuming the MB control data is laid out in scan line order in a rectangle with width = 16 bytes.
+ // Control data has dimension of X x Y = 16 x N bytes, where N = W x H / 16
+ // Each MB has 256 bytes of control data
+
+ // For CRESTLINE, 256 bytes are stored in memory and fetched into GRF.
+ // MB_offset = MBsCntX * CurRow + CurCol
+ // Byte_offset = MB_offset * (256 << Mbaff_flag), Mbaff_flag = 0 or 1.
+ // Base address of a control data block = (x, y) = (0, y'=y/x), region width is 16 bytes
+ // where y' = Byte_offset / 16 = MB_offset * (16 << Mbaff_flag)
+ // MBCntrlDataOffsetY holds y'.
+
+ // For BearLake-C, 64 bytes are stored in memory and dataport expands to 256 bytes. Need to use a special read command on BL-C.
+ // MB_offset = MBsCntX * CurRow + CurCol
+ // Byte_offset = MB_offset * (64 << Mbaff_flag), Mbaff_flag = 0 or 1.
+ // MBCntrlDataOffsetY holds globel byte offset.
+
+#if !defined(DEV_CL)
+ mul (1) CntrlDataOffsetY:ud MBsCntX:w ORIY:w
+ add (1) CntrlDataOffsetY:ud CntrlDataOffsetY:ud ORIX:w
+ mul (1) CntrlDataOffsetY:ud CntrlDataOffsetY:ud 128:uw
+#endif
+ //====================================================================================
+
+ add (1) ORIX_LEFT:w ORIX_LEFT:w -4:w
+ add (1) ORIY_TOP:w ORIY_TOP:w -4:w
+
+ //=========== Process Top MB ============
+ and (1) BitFields:w BitFields:w TopFieldFlag:w // Reset BotFieldFlag
+
+ // Build a ramp from 0 to 15
+ mov (16) RRampW(0)<1> RampConstC<0;8,1>:ub
+ add (8) RRampW(0,8)<1> RRampW(0,8) 8:w // RRampW = ramp 15-0
+
+ILDB_LABEL(RE_ENTRY_UV): // for bootom field
+
+ // Load current MB control data
+#if defined(DEV_CL)
+ #include "load_ILDB_Cntrl_Data_64DW.asm" // Crestline
+#else
+ #include "load_ILDB_Cntrl_Data_16DW.asm" // Cantiga and beyond
+#endif
+
+ // Init addr register for vertical control data
+ mov (1) ECM_AddrReg<1>:w CNTRL_DATA_BASE:w // Init ECM_AddrReg
+
+ // Use free cycles here
+ // Check loaded control data
+ and.z.f0.1 (16) null<1>:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]<16;16,1>:uw 0xFFFF:uw // Skip ILDB?
+ and.nz.f0.0 (1) null:w r[ECM_AddrReg, ExtBitFlags]:ub DISABLE_ILDB_FLAG:w // Skip ILDB?
+
+ // Set DualFieldMode for all data read, write and deblocking
+ and (1) CTemp1_W:uw r[ECM_AddrReg, BitFlags]:ub FieldModeAboveMbFlag+FieldModeCurrentMbFlag:uw
+
+ // Get Vert Edge Pattern (frame vs. field MBs)
+ and (1) VertEdgePattern:uw r[ECM_AddrReg, BitFlags]:ub FieldModeLeftMbFlag+FieldModeCurrentMbFlag:uw
+
+ (f0.1.all16h) jmpi ILDB_LABEL(SKIP_ILDB_UV) // Skip ILDB
+ (f0.0) jmpi ILDB_LABEL(SKIP_ILDB_UV) // Skip ILDB
+
+ // Set DualFieldMode for all data read, write and deblocking
+// and (1) CTemp1_W:uw r[ECM_AddrReg, BitFlags]:ub FieldModeAboveMbFlag+FieldModeCurrentMbFlag:uw
+ cmp.z.f0.0 (1) null:w CTemp1_W:uw ABOVE_FIELD_CUR_FRAME:w
+ and (1) DualFieldMode:w f0.0:w 0x0001:w
+
+ #include "load_Cur_UV_8x8T_Mbaff.asm" // Load transposed data 8x8
+ #include "load_Left_UV_2x8T_Mbaff.asm" // Load left MB (2x8) UV data from memory if exists
+
+ #include "Transpose_Cur_UV_8x8.asm"
+ #include "Transpose_Left_UV_2x8.asm"
+
+
+ //---------- Perform vertical ILDB filting on UV ----------
+ #include "AVC_ILDB_Filter_Mbaff_UV_v.asm"
+ //---------------------------------------------------------
+
+ #include "save_Left_UV_8x2T_Mbaff.asm" // Write left MB (2x8) Y data to memory if exists
+ #include "load_Top_UV_8x2_Mbaff.asm" // Load top MB (8x2) Y data from memory if exists
+
+ #include "Transpose_Cur_UV_8x8.asm" // Transpose a MB for horizontal edge de-blocking
+
+ //---------- Perform horizontal ILDB filting on UV ----------
+ #include "AVC_ILDB_Filter_Mbaff_UV_h.asm"
+ //-----------------------------------------------------------
+
+ #include "save_Cur_UV_8x8_Mbaff.asm" // Write 8x8
+ #include "save_Top_UV_8x2_Mbaff.asm" // Write top MB (8x2) if not the top row
+
+ //-----------------------------------------------------------
+ILDB_LABEL(SKIP_ILDB_UV):
+
+ and.z.f0.0 (1) null:w BitFields:w BotFieldFlag:w
+
+ //=========== Process Bottom MB ============
+ or (1) BitFields:w BitFields:w BotFieldFlag:w // Set BotFieldFlag to 1
+ (f0.0) jmpi ILDB_LABEL(RE_ENTRY_UV) // Loop back for bottom deblocking
+
+ // Fall through to finish
+
+ //=========== Check write commit of the last write ============
+ mov (8) WritebackResponse(0)<1> WritebackResponse(0)
+
+ILDB_LABEL(POST_ILDB_UV):
+
+ // Send notification thru Gateway to root thread, update chroma Status[CurRow]
+ #include "AVC_ILDB_ForwardMsg.asm"
+
+#if !defined(GW_DCN) // For non-ILK chipsets
+ //child send EOT : Request type = 1
+ END_CHILD_THREAD
+#endif // !defined(DEV_ILK)
+
+ // The thread finishs here
+ //------------------------------------------------------------------------------
+
+ ////////////////////////////////////////////////////////////////////////////////
+ // Include other subrutines being called
+ #include "AVC_ILDB_Chroma_Core_Mbaff.asm"
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Child_Mbaff_Y.asm b/shaders/h264/ildb/AVC_ILDB_Child_Mbaff_Y.asm new file mode 100644 index 0000000..b4b1937 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Child_Mbaff_Y.asm @@ -0,0 +1,188 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// AVC Child Kernel (Vertical and horizontal de-block a 4:2:0 MB Y comp)
+//
+// First, de-block vertical edges from left to right.
+// Second, de-block horizontal edge from top to bottom.
+//
+// ***** MBAFF Mode *****
+// This version deblocks top MB first, followed by bottom MB.
+//
+// Need variable CurMB to indicate top MB or bottom MB (CurMB = 0 or 1).
+// We can use BotFieldFlag in BitFields to represent it.
+//
+// Usage:
+// 1) Access control data for top
+// CntrlDataOffsetY + CurMB * Control data block size (64 DWs for CL, 16 DWs for BLC)
+//
+// 2) Load frame/field video data based on flags: FieldModeCurrentMbFlag, FieldModeLeftMbFlag, FieldModeaboveMbFlag,
+//
+// E.g.
+// if (pCntlData->BitField & FieldModeCurrentMbFlag)
+// cur_y = ORIX_CUR.y + CurMB * 1; // Add field vertical offset for bot field MB .
+// else
+// cur_y = ORIX_CUR.y + CurMB * MB_Rows_Y; // Add bottom MB vertical offset for bot MB
+//
+//
+//////////////////////////////////////////////////////////////////////////////////////////////////////////////
+#define AVC_ILDB
+
+.kernel AVC_ILDB_CHILD_MBAFF_Y
+#if defined(COMBINED_KERNEL)
+ILDB_LABEL(AVC_ILDB_CHILD_Y):
+#endif
+
+#include "setupVPKernel.asm"
+#include "AVC_ILDB.inc"
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xE998:w
+#endif
+
+ // Setup temp buf used by load and save code
+ #define BUF_B RTempB
+ #define BUF_D RTempD
+
+ // Init local variables
+ // These coordinates are in progressive fashion
+ mul (4) ORIX_CUR<2>:w ORIX<0;1,0>:w 16:w { NoDDClr } // Expand X addr to bytes, repeat 4 times
+ mul (4) ORIY_CUR<2>:w ORIY<0;1,0>:w 32:w { NoDDChk } // Expand Y addr to bytes, repeat 4 times
+
+ mov (2) f0.0<1>:w 0:w
+
+ mov (1) GateWayOffsetC:uw ORIY:uw // Use row # as Gateway offset
+
+ //=== Null Kernel ===============================================================
+// jmpi POST_ILDB
+ //===============================================================================
+
+ //====================================================================================
+ // Assuming the MB control data is laid out in scan line order in a rectangle with width = 16 bytes.
+ // Control data has dimension of X x Y = 16 x N bytes, where N = W x H / 16
+ // Each MB has 256 bytes of control data
+
+ // For CRESTLINE, 256 bytes are stored in memory and fetched into GRF.
+ // MB_offset = MBsCntX * CurRow + CurCol
+ // Byte_offset = MB_offset * (256 << Mbaff_flag), Mbaff_flag = 0 or 1.
+ // Base address of a control data block = (x, y) = (0, y'=y/x), region width is 16 bytes
+ // where y' = Byte_offset / 16 = MB_offset * (16 << Mbaff_flag)
+ // MBCntrlDataOffsetY holds y'.
+
+ // For BearLake-C, 64 bytes are stored in memory and dataport expands to 256 bytes. Need to use a special read command on BL-C.
+ // MB_offset = MBsCntX * CurRow + CurCol
+ // Byte_offset = MB_offset * (64 << Mbaff_flag), Mbaff_flag = 0 or 1.
+ // MBCntrlDataOffsetY holds globel byte offset.
+
+#if !defined(DEV_CL)
+ mul (1) CntrlDataOffsetY:ud MBsCntX:w ORIY:w
+ add (1) CntrlDataOffsetY:ud CntrlDataOffsetY:ud ORIX:w
+ mul (1) CntrlDataOffsetY:ud CntrlDataOffsetY:ud 128:uw
+#endif
+
+ //====================================================================================
+
+ add (1) ORIX_LEFT:w ORIX_LEFT:w -4:w
+ add (1) ORIY_TOP:w ORIY_TOP:w -4:w
+
+
+ //=========== Process Top MB ============
+ and (1) BitFields:w BitFields:w TopFieldFlag:w // Reset BotFieldFlag
+
+RE_ENTRY: // for bootom field
+
+ // Load current MB control data
+#if defined(DEV_CL)
+ #include "load_ILDB_Cntrl_Data_64DW.asm" // Crestline
+#else
+ #include "load_ILDB_Cntrl_Data_16DW.asm" // Cantiga and beyond
+#endif
+
+ // Init addr register for vertical control data
+ mov (1) ECM_AddrReg<1>:w CNTRL_DATA_BASE:w // Init edge control map AddrReg
+
+ // Check loaded control data
+ and.z.f0.1 (16) null<1>:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]<16;16,1>:uw 0xFFFF:uw // Skip ILDB?
+ and.nz.f0.0 (1) null:w r[ECM_AddrReg, ExtBitFlags]:ub DISABLE_ILDB_FLAG:w // Skip ILDB?
+
+ // Use free cycles here
+ // Set DualFieldMode for all data read, write and deblocking
+ and (1) CTemp1_W:uw r[ECM_AddrReg, BitFlags]:ub FieldModeAboveMbFlag+FieldModeCurrentMbFlag:uw
+
+ // Get Vert Edge Pattern (frame vs. field MBs)
+ and (1) VertEdgePattern:uw r[ECM_AddrReg, BitFlags]:ub FieldModeLeftMbFlag+FieldModeCurrentMbFlag:uw
+
+ (f0.1.all16h) jmpi SKIP_ILDB // Skip ILDB
+ (f0.0) jmpi SKIP_ILDB // Skip ILDB
+
+ // Set DualFieldMode for all data read, write and deblocking
+// and (1) CTemp1_W:uw r[ECM_AddrReg, BitFlags]:ub FieldModeAboveMbFlag+FieldModeCurrentMbFlag:uw
+ cmp.z.f0.0 (1) null:w CTemp1_W:uw ABOVE_FIELD_CUR_FRAME:w
+ and (1) DualFieldMode:w f0.0:w 0x0001:w
+
+ // Load current MB // DDD1
+ #include "load_Cur_Y_16x16T_Mbaff.asm" // Load cur Y, 16x16, transpose
+ #include "load_Left_Y_4x16T_Mbaff.asm" // Load left MB (4x16) Y data from memory if exists
+
+ #include "Transpose_Cur_Y_16x16.asm"
+ #include "Transpose_Left_Y_4x16.asm"
+
+ //---------- Perform vertical ILDB filting on Y----------
+ #include "AVC_ILDB_Filter_Mbaff_Y_v.asm"
+ //-------------------------------------------------------
+
+ #include "save_Left_Y_16x4T_Mbaff.asm" // Write left MB (4x16) Y data to memory if exists
+ #include "load_Top_Y_16x4_Mbaff.asm" // Load top MB (16x4) Y data from memory if exists
+ #include "Transpose_Cur_Y_16x16.asm" // Transpose a MB for horizontal edge de-blocking
+
+ //---------- Perform horizontal ILDB filting on Y ----------
+ #include "AVC_ILDB_Filter_Mbaff_Y_h.asm"
+ //----------------------------------------------------------
+
+ #include "save_Cur_Y_16x16_Mbaff.asm" // Write cur MB (16x16)
+ #include "save_Top_Y_16x4_Mbaff.asm" // Write top MB (16x4) if not the top row
+
+SKIP_ILDB:
+ //----------------------------------------------------------
+ and.z.f0.0 (1) null:w BitFields:w BotFieldFlag:w
+
+ //=========== Process Bottom MB ============
+ or (1) BitFields:w BitFields:w BotFieldFlag:w // Set BotFieldFlag to 1
+ (f0.0) jmpi RE_ENTRY // Loop back for bottom deblocking
+
+ // Fall through to finish
+
+ //=========== Check write commit of the last write ============
+ mov (8) WritebackResponse(0)<1> WritebackResponse(0)
+
+POST_ILDB:
+
+ //---------------------------------------------------------------------------
+ // Send notification thru Gateway to root thread, update luma Status[CurRow]
+
+ #include "AVC_ILDB_ForwardMsg.asm"
+
+#if !defined(GW_DCN) // For non-ILK chipsets
+ //child send EOT : Request type = 1
+ END_CHILD_THREAD
+#endif // !defined(DEV_ILK)
+
+ // The thread finishs here
+ //------------------------------------------------------------------------------
+
+
+ ////////////////////////////////////////////////////////////////////////////////
+ // Include other subrutines being called
+ #include "AVC_ILDB_Luma_Core_Mbaff.asm"
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Child_UV.asm b/shaders/h264/ildb/AVC_ILDB_Child_UV.asm new file mode 100644 index 0000000..df40ad4 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Child_UV.asm @@ -0,0 +1,186 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// AVC Child Kernel (Vertical and horizontal de-block a 4:2:0 MB UV comp)
+//
+// First de-block vertical edges from left to right.
+// Second de-block horizontal edge from top to bottom.
+//
+// For 4:2:0, chroma is always de-blocked at 8x8.
+// NV12 format allows to filter U and V together.
+//
+//////////////////////////////////////////////////////////////////////////////////////////////////////////////
+#define AVC_ILDB
+
+.kernel AVC_ILDB_CHILD_UV
+#if defined(COMBINED_KERNEL)
+ILDB_LABEL(AVC_ILDB_CHILD_UV):
+#endif
+
+#include "setupVPKernel.asm"
+#include "AVC_ILDB.inc"
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0x9997:w
+#endif
+
+ // Init local variables
+ shl (8) ORIX_CUR<1>:w ORIX<0;2,1>:w 4:w // Expand addr to bytes, repeat (x,y) 4 times
+
+ // Init addr register for vertical control data
+ mov (1) ECM_AddrReg<1>:w CNTRL_DATA_BASE:w // Init ECM_AddrReg
+
+ //=== Null Kernel ===============================================================
+// jmpi ILDB_LABEL(POST_ILDB_UV_UV)
+ //===============================================================================
+
+#if defined(DEV_CL)
+ mov (1) acc0.0:w 240:w
+#else
+ //====================================================================================
+ // For BearLake-C, 64 bytes are stored in memory and dataport expands to 256 bytes. Need to use a special read command on BL-C.
+ // MB_offset = MBsCntX * CurRow + CurCol
+ // MBCntrlDataOffsetY = globel_byte_offset = MB_offset * 64
+ mul (1) CntrlDataOffsetY:ud MBsCntX:w ORIY:w
+ add (1) CntrlDataOffsetY:ud CntrlDataOffsetY:ud ORIX:w
+
+ // Assign to MSGSRC.2:ud for memory access
+ // mul (1) CntrlDataOffsetY:ud CntrlDataOffsetY:ud 64:uw
+ mul (1) MSGSRC.2:ud CntrlDataOffsetY:ud 64:uw
+
+ mov (1) acc0.0:w 320:w
+#endif
+ mac (1) URBOffsetC:w ORIY:w 4:w // UV URB entries are right after Y entries
+
+
+ // Init local variables
+// shl (8) ORIX_CUR<1>:w ORIX<0;2,1>:w 4:w // Expand addr to bytes, repeat (x,y) 4 times
+ add (1) ORIX_LEFT:w ORIX_LEFT:w -4:w
+ add (1) ORIY_TOP:w ORIY_TOP:w -4:w
+
+ // Build a ramp from 0 to 15
+ mov (16) RRampW(0)<1> RampConstC<0;8,1>:ub
+ add (8) RRampW(0,8)<1> RRampW(0,8) 8:w // RRampW = ramp 15-0
+
+ // Load current MB control data
+#if defined(DEV_CL)
+ #if defined(_APPLE)
+ #include "load_ILDB_Cntrl_Data_22DW.asm" // Crestline for Apple, progressive only
+ #else
+ #include "load_ILDB_Cntrl_Data_64DW.asm" // Crestline
+ #endif
+#else
+ #include "load_ILDB_Cntrl_Data_16DW.asm" // Cantiga and beyond
+#endif
+
+ // Check loaded control data
+ #if defined(_APPLE)
+ and.z.f0.1 (8) null<1>:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]<8;8,1>:uw 0xFFFF:uw // Skip ILDB?
+ (f0.1) and.z.f0.1 (2) null<1>:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz0]<2;2,1>:uw 0xFFFF:uw // Skip ILDB?
+ #else
+ and.z.f0.1 (16) null<1>:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]<16;16,1>:uw 0xFFFF:uw // Skip ILDB?
+ #endif
+
+ and.nz.f0.0 (1) null:w r[ECM_AddrReg, ExtBitFlags]:ub DISABLE_ILDB_FLAG:w // Skip ILDB?
+
+ mov (1) GateWayOffsetC:uw ORIY:uw // Use row # as Gateway offset
+
+ #if defined(_APPLE)
+ (f0.1.all8h) jmpi ILDB_LABEL(READ_FOR_URB_UV) // Skip ILDB
+ #else
+ (f0.1.all16h) jmpi ILDB_LABEL(READ_FOR_URB_UV) // Skip ILDB
+ #endif
+
+ (f0.0) jmpi ILDB_LABEL(READ_FOR_URB_UV) // Skip ILDB
+
+
+
+ #include "load_Cur_UV_8x8T.asm" // Load transposed data 8x8
+// #include "load_Left_UV_2x8T.asm"
+ #include "load_Top_UV_8x2.asm" // Load top MB (8x2) Y data from memory if exists
+
+ #include "Transpose_Cur_UV_8x8.asm"
+// #include "Transpose_Left_UV_2x8.asm"
+
+
+ //---------- Perform vertical ILDB filting on UV ----------
+ #include "AVC_ILDB_Filter_UV_v.asm"
+ //---------------------------------------------------------
+
+ #include "save_Left_UV_8x2T.asm" // Write left MB (2x8) Y data to memory if exists
+ #include "Transpose_Cur_UV_8x8.asm" // Transpose a MB for horizontal edge de-blocking
+
+ //---------- Perform horizontal ILDB filting on UV ----------
+ #include "AVC_ILDB_Filter_UV_h.asm"
+ //-----------------------------------------------------------
+
+ #include "save_Cur_UV_8x8.asm" // Write 8x8
+ #include "save_Top_UV_8x2.asm" // Write top MB (8x2) if not the top row
+
+ //---------- Write right most 4 columns of cur MB to URB ----------
+ // Transpose the right most 2 cols 2x8 (word) in GRF to 8x2 in BUF_D. It is 2 left most cols in cur MB.
+ #include "Transpose_Cur_UV_2x8.asm"
+
+ILDB_LABEL(WRITE_URB_UV):
+ mov (8) m1<1>:ud LEFT_TEMP_D(1)<8;8,1> // Copy 1 GRF to 1 URB entry (U+V)
+
+ #include "writeURB_UV_Child.asm"
+ //-----------------------------------------------------------------
+
+ //=========== Check write commit of the last write ============
+ mov (8) WritebackResponse(0)<1> WritebackResponse(0)
+
+ILDB_LABEL(POST_ILDB_UV):
+ //---------------------------------
+
+ // Send notification thru Gateway to root thread, update chroma Status[CurRow]
+ #include "AVC_ILDB_ForwardMsg.asm"
+
+#if !defined(GW_DCN) // For non-ILK chipsets
+ //child send EOT : Request type = 1
+ END_CHILD_THREAD
+#endif // !defined(DEV_ILK)
+
+ // The thread finishs here
+ //------------------------------------------------------------------------------
+
+ILDB_LABEL(READ_FOR_URB_UV):
+ // Still need to prepare URB data for the right neighbor MB
+ #include "load_Cur_UV_Right_Most_2x8.asm" // Load cur MB ( right most 4x16) Y data from memory
+ #include "Transpose_Cur_UV_Right_Most_2x8.asm"
+// jmpi ILDB_LABEL(WRITE_URB_UV)
+
+ mov (8) m1<1>:ud LEFT_TEMP_D(1)<8;8,1> // Copy 1 GRF to 1 URB entry (U+V)
+
+ #include "writeURB_UV_Child.asm"
+ //-----------------------------------------------------------------
+
+ // Send notification thru Gateway to root thread, update chroma Status[CurRow]
+ #include "AVC_ILDB_ForwardMsg.asm"
+
+#if !defined(GW_DCN) // For non-ILK chipsets
+ //child send EOT : Request type = 1
+ END_CHILD_THREAD
+#endif // !defined(DEV_ILK)
+
+ // The thread finishs here
+ //------------------------------------------------------------------------------
+
+
+ ////////////////////////////////////////////////////////////////////////////////
+ // Include other subrutines being called
+// #include "AVC_ILDB_Luma_Core.asm"
+ #include "AVC_ILDB_Chroma_Core.asm"
+
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Child_Y.asm b/shaders/h264/ildb/AVC_ILDB_Child_Y.asm new file mode 100644 index 0000000..bef52d8 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Child_Y.asm @@ -0,0 +1,176 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////////////////////////
+// AVC Child Kernel (Vertical and horizontal de-block a 4:2:0 MB Y comp)
+//
+// First, de-block vertical edges from left to right.
+// Second, de-block horizontal edge from top to bottom.
+//
+// If transform_size_8x8_flag = 1, luma is de-blocked at 8x8. Otherwise, luma is de-blocked at 4x4.
+//
+//////////////////////////////////////////////////////////////////////////////////////////////////////////////
+#define AVC_ILDB
+
+.kernel AVC_ILDB_CHILD_Y
+#if defined(COMBINED_KERNEL)
+ILDB_LABEL(AVC_ILDB_CHILD_Y):
+#endif
+
+#include "setupVPKernel.asm"
+#include "AVC_ILDB.inc"
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0x9998:w
+#endif
+
+ // Init local variables
+ shl (8) ORIX_CUR<1>:w ORIX<0;2,1>:w 4:w // Expand addr to bytes, repeat (x,y) 4 times
+
+ // Init addr register for vertical control data
+ mov (1) ECM_AddrReg<1>:w CNTRL_DATA_BASE:w // Init edge control map AddrReg
+
+ //=== Null Kernel ===============================================================
+// jmpi ILDB_LABEL(POST_ILDB_Y)
+ //===============================================================================
+
+ mul (1) URBOffsetC:uw ORIY:uw 4:w
+
+#if !defined(DEV_CL)
+ //====================================================================================
+ // For BearLake-C, 64 bytes are stored in memory and dataport expands to 256 bytes. Need to use a special read command on BL-C.
+ // MB_offset = MBsCntX * CurRow + CurCol
+ // MBCntrlDataOffsetY = globel_byte_offset = MB_offset * 64
+ mul (1) CntrlDataOffsetY:ud MBsCntX:w ORIY:w
+ add (1) CntrlDataOffsetY:ud CntrlDataOffsetY:ud ORIX:w
+
+ // Assign to MSGSRC.2:ud for memory access
+ // mul (1) CntrlDataOffsetY:ud CntrlDataOffsetY:ud 64:uw
+ mul (1) MSGSRC.2:ud CntrlDataOffsetY:ud 64:uw
+
+#endif
+
+ // Load current MB control data
+#if defined(DEV_CL)
+ #if defined(_APPLE)
+ #include "load_ILDB_Cntrl_Data_22DW.asm" // Crestline for Apple, progressive only
+ #else
+ #include "load_ILDB_Cntrl_Data_64DW.asm" // Crestline
+ #endif
+#else
+ #include "load_ILDB_Cntrl_Data_16DW.asm" // Cantiga and beyond
+#endif
+
+ // Check loaded control data
+ #if defined(_APPLE)
+ and.z.f0.1 (8) null<1>:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]<8;8,1>:uw 0xFFFF:uw // Skip ILDB?
+ (f0.1) and.z.f0.1 (2) null<1>:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz0]<2;2,1>:uw 0xFFFF:uw // Skip ILDB?
+ #else
+ and.z.f0.1 (16) null<1>:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]<16;16,1>:uw 0xFFFF:uw // Skip ILDB?
+ #endif
+
+ and.nz.f0.0 (1) null:w r[ECM_AddrReg, ExtBitFlags]:ub DISABLE_ILDB_FLAG:w // Skip ILDB?
+
+ // Use free cycles here
+ add (1) ORIX_LEFT:w ORIX_LEFT:w -4:w
+// add (1) ORIY_TOP:w ORIY_TOP:w -4:w
+ mov (1) GateWayOffsetC:uw ORIY:uw // Use row # as Gateway offset
+
+ #if defined(_APPLE)
+ (f0.1.all8h) jmpi ILDB_LABEL(READ_FOR_URB_Y) // Skip ILDB
+ #else
+ (f0.1.all16h) jmpi ILDB_LABEL(READ_FOR_URB_Y) // Skip ILDB
+ #endif
+
+ (f0.0) jmpi ILDB_LABEL(READ_FOR_URB_Y) // Skip ILDB
+
+ add (1) ORIY_TOP:w ORIY_TOP:w -4:w
+
+ // Bettr performance is observed if boundary MBs are not checked and skipped.
+
+ #include "load_Cur_Y_16x16T.asm" // Load cur MB Y, 16x16, transpose
+// #include "load_Left_Y_4x16T.asm" // Load left MB (4x16) Y data from memory
+ #include "load_Top_Y_16x4.asm" // Load top MB (16x4) Y data from memory
+
+ #include "Transpose_Cur_Y_16x16.asm"
+// #include "Transpose_Left_Y_4x16.asm"
+
+ //---------- Perform vertical ILDB filting on Y ---------
+ #include "AVC_ILDB_Filter_Y_v.asm"
+ //-------------------------------------------------------
+
+ #include "save_Left_Y_16x4T.asm" // Write left MB (4x16) Y data to memory
+ #include "Transpose_Cur_Y_16x16.asm" // Transpose a MB for horizontal edge de-blocking
+
+ //---------- Perform horizontal ILDB filting on Y -------
+ #include "AVC_ILDB_Filter_Y_h.asm"
+ //-------------------------------------------------------
+
+ #include "save_Cur_Y_16x16.asm" // Write cur MB (16x16)
+ #include "save_Top_Y_16x4.asm" // Write top MB (16x4)
+
+ //---------- Write right most 4 columns of cur MB to URB ----------
+ // Transpose the right most 4 cols 4x16 in GRF to 16x4 in LEFT_TEMP_B. It is 4 left most cols in cur MB.
+ #include "Transpose_Cur_Y_4x16.asm"
+
+ILDB_LABEL(WRITE_URB_Y):
+ // Note: LEFT_TEMP_B(2) = TOP_TEMP_B(0), TOP_TEMP_B must be avail
+ mov (16) m1<1>:ud LEFT_TEMP_D(2)<8;8,1> // Copy 2 GRFs to 2 URB entries (Y)
+
+ #include "writeURB_Y_Child.asm"
+ //-----------------------------------------------------------------
+
+ //=========== Check write commit of the last write ============
+ mov (8) WritebackResponse(0)<1> WritebackResponse(0)
+
+ILDB_LABEL(POST_ILDB_Y):
+ // Send notification thru Gateway to root thread, update luma Status[CurRow]
+ #include "AVC_ILDB_ForwardMsg.asm"
+
+#if !defined(GW_DCN) // For non-ILK chipsets
+ //child send EOT : Request type = 1
+ END_CHILD_THREAD
+#endif // !defined(DEV_ILK)
+
+ // The thread finishs here
+ //------------------------------------------------------------------------------
+
+ILDB_LABEL(READ_FOR_URB_Y):
+ // Still need to prepare URB data for the right neighbor MB
+ #include "load_Cur_Y_Right_Most_4x16.asm" // Load cur MB ( right most 4x16) Y data from memory
+ #include "Transpose_Cur_Y_Right_Most_4x16.asm"
+// jmpi ILDB_LABEL(WRITE_URB_Y)
+
+ // Note: LEFT_TEMP_B(2) = TOP_TEMP_B(0), TOP_TEMP_B must be avail
+ mov (16) m1<1>:ud LEFT_TEMP_D(2)<8;8,1> // Copy 2 GRFs to 2 URB entries (Y)
+
+ #include "writeURB_Y_Child.asm"
+ //-----------------------------------------------------------------
+
+ // Send notification thru Gateway to root thread, update luma Status[CurRow]
+ #include "AVC_ILDB_ForwardMsg.asm"
+
+#if !defined(GW_DCN) // For non-ILK chipsets
+ //child send EOT : Request type = 1
+ END_CHILD_THREAD
+#endif // !defined(DEV_ILK)
+
+ // The thread finishs here
+ //------------------------------------------------------------------------------
+
+ ////////////////////////////////////////////////////////////////////////////////
+ // Include other subrutines being called
+ #include "AVC_ILDB_Luma_Core.asm"
+// #include "AVC_ILDB_Chroma_Core.asm"
+
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Chroma_Core.asm b/shaders/h264/ildb/AVC_ILDB_Chroma_Core.asm new file mode 100644 index 0000000..e33d022 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Chroma_Core.asm @@ -0,0 +1,165 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__AVC_ILDB_CHROMA_CORE__) // Make sure this file is only included once
+#define __AVC_ILDB_CHROMA_CORE__
+
+////////// AVC ILDB Chroma Core /////////////////////////////////////////////////////////////////////////////////
+//
+// This core performs AVC U or V ILDB filtering on one horizontal edge (8 pixels) of a MB.
+// If data is transposed, it can also de-block a vertical edge.
+//
+// Bafore calling this subroutine, caller needs to set the following parameters.
+//
+// - EdgeCntlMap1 // Edge control map A
+// - EdgeCntlMap2 // Edge control map B
+// - P_AddrReg // Src and dest address register for P pixels
+// - Q_AddrReg // Src and dest address register for Q pixels
+// - alpha // alpha corresponding to the edge to be filtered
+// - beta // beta corresponding to the edge to be filtered
+// - tc0 // tc0 corresponding to the edge to be filtered
+//
+// U or V:
+// +----+----+----+----+
+// | P1 | p0 | q0 | q1 |
+// +----+----+----+----+
+//
+// p1 = r[P_AddrReg, 0]<16;8,2>
+// p0 = r[P_AddrReg, 16]<16;8,2>
+// q0 = r[Q_AddrReg, 0]<16;8,2>
+// q1 = r[Q_AddrReg, 16]<16;8,2>
+//
+/////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+// The region is both src and dest
+// P0-P3 and Q0-Q3 should be only used if they have not been modified to new values
+#undef P1
+#undef P0
+#undef Q0
+#undef Q1
+
+#define P1 r[P_AddrReg, 0]<16;8,2>:ub
+#define P0 r[P_AddrReg, 16]<16;8,2>:ub
+#define Q0 r[Q_AddrReg, 0]<16;8,2>:ub
+#define Q1 r[Q_AddrReg, 16]<16;8,2>:ub
+
+// New region as dest
+#undef NewP0
+#undef NewQ0
+
+#define NewP0 r[P_AddrReg, 16]<2>:ub
+#define NewQ0 r[Q_AddrReg, 0]<2>:ub
+
+// Filter one chroma edge
+FILTER_UV:
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0x1112:w
+#endif
+ //---------- Derive filterSampleflag in AVC spec, equition (8-469) ----------
+ // bS is in MaskA
+
+ // Src copy of the p1, p0, q0, q1
+// mov (8) p1(0)<1> r[P_AddrReg, 0]<16;8,2>:ub
+// mov (8) p0(0)<1> r[P_AddrReg, 16]<16;8,2>:ub
+// mov (8) q0(0)<1> r[Q_AddrReg, 0]<16;8,2>:ub
+// mov (8) q1(0)<1> r[Q_AddrReg, 16]<16;8,2>:ub
+
+// mov (1) f0.0:uw MaskA:uw
+
+ add (8) q0_p0(0)<1> Q0 -P0 // q0-p0
+ add (8) TempRow0(0)<1> P1 -P0 // p1-p0
+ add (8) TempRow1(0)<1> Q1 -Q0 // q1-q0
+
+ // Build FilterSampleFlag
+ // abs(q0-p0) < alpha
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)q0_p0(0) alpha:w
+ // abs(p1-p0) < Beta
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)TempRow0(0) beta:w
+ // abs(q1-q0) < Beta
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)TempRow1(0) beta:w
+
+ //-----------------------------------------------------------------------------------------
+
+ // if
+ (f0.0) if (8) UV_ENDIF1
+ // For channels whose edge control map1 = 1 ---> perform de-blocking
+
+// mov (1) f0.1:w MaskB:w {NoMask} // Now check for which algorithm to apply
+
+ (f0.1) if (8) UV_ELSE2
+
+ // For channels whose edge control map2 = 1 ---> bS = 4 algorithm
+ // p0' = (2*p1 + p0 + q1 + 2) >> 2
+ // q0' = (2*q1 + q0 + p1 + 2) >> 2
+
+ // Optimized version:
+ // A = (p1 + q1 + 2)
+ // p0' = (p0 + p1 + A) >> 2
+ // q0' = (q0 + q1 + A) >> 2
+ //------------------------------------------------------------------------------------
+
+ // p0' = (2*p1 + p0 + q1 + 2) >> 2
+ add (8) acc0<1>:w Q1 2:w
+ mac (8) acc0<1>:w P1 2:w
+ add (8) acc0<1>:w acc0<8;8,1>:w P0
+ shr.sat (8) TempRow0B(0)<2> acc0<8;8,1>:w 2:w
+
+ // q0' = (2*q1 + q0 + p1 + 2) >> 2
+ add (8) acc0<1>:w P1 2:w
+ mac (8) acc0<1>:w Q1 2:w
+ add (8) acc0<1>:w acc0<8;8,1>:w Q0
+ shr.sat (8) TempRow1B(0)<2> acc0<8;8,1>:w 2:w
+
+ mov (8) NewP0 TempRow0B(0) // p0'
+ mov (8) NewQ0 TempRow1B(0) // q0'
+
+
+UV_ELSE2:
+ else (8) UV_ENDIF2
+ // For channels whose edge control map2 = 0 ---> bS < 4 algorithm
+
+ // Expand tc0 (tc0 has 4 bytes)
+// mov (8) tc0_exp(0)<1> tc0<1;2,0>:ub {NoMask} // tc0_exp = tc0, each tc0 is duplicated 2 times for 2 adjcent pixels
+ mov (8) acc0<1>:w tc0<1;2,0>:ub {NoMask} // tc0_exp = tc0, each tc0 is duplicated 2 times for 2 adjcent pixels
+
+ // tc_exp = tc0_exp + 1
+// add (8) tc_exp(0)<1> tc0_exp(0) 1:w
+ add (8) tc_exp(0)<1> acc0<8;8,1>:w 1:w
+
+ // delta = Clip3(-tc, tc, ((((q0 - p0)<<2) + (p1-q1) + 4) >> 3))
+ // 4 * (q0-p0) + p1 - q1 + 4
+ add (8) acc0<1>:w P1 4:w
+ mac (8) acc0<1>:w q0_p0(0) 4:w
+ add (8) acc0<1>:w acc0<8;8,1>:w -Q1
+ shr (8) TempRow0(0)<1> acc0<8;8,1>:w 3:w
+
+ // tc clip
+ cmp.g.f0.0 (8) null:w TempRow0(0) tc_exp(0) // Clip if > tc0
+ cmp.l.f0.1 (8) null:w TempRow0(0) -tc_exp(0) // Clip if < -tc0
+
+ (f0.0) mov (8) TempRow0(0)<1> tc_exp(0)
+ (f0.1) mov (8) TempRow0(0)<1> -tc_exp(0)
+
+ // p0' = Clip1(p0 + delta) = Clip3(0, 0xFF, p0 + delta)
+ add.sat (8) TempRow1B(0)<2> P0 TempRow0(0) // p0+delta
+
+ // q0' = Clip1(q0 - delta) = Clip3(0, 0xFF, q0 - delta)
+ add.sat (8) TempRow0B(0)<2> Q0 -TempRow0(0) // q0-delta
+
+ mov (8) NewP0 TempRow1B(0) // p0'
+ mov (8) NewQ0 TempRow0B(0) // q0'
+
+ endif
+UV_ENDIF2:
+UV_ENDIF1:
+ endif
+
+RETURN
+
+#endif // !defined(__AVC_ILDB_CHROMA_CORE__)
diff --git a/shaders/h264/ildb/AVC_ILDB_Chroma_Core_Mbaff.asm b/shaders/h264/ildb/AVC_ILDB_Chroma_Core_Mbaff.asm new file mode 100644 index 0000000..f567d95 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Chroma_Core_Mbaff.asm @@ -0,0 +1,146 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC ILDB Chroma Core Mbaff /////////////////////////////////////////////////////////////////////////////////
+//
+// This core performs AVC U or V ILDB filtering on one horizontal edge (8 pixels) of a MB.
+// If data is transposed, it can also de-block a vertical edge.
+//
+// Bafore calling this subroutine, caller needs to set the following parameters.
+//
+// - EdgeCntlMap1 // Edge control map A
+// - EdgeCntlMap2 // Edge control map B
+// - P_AddrReg // Src and dest address register for P pixels
+// - Q_AddrReg // Src and dest address register for Q pixels
+// - alpha // alpha corresponding to the edge to be filtered
+// - beta // beta corresponding to the edge to be filtered
+// - tc0 // tc0 corresponding to the edge to be filtered
+//
+// U or V:
+// +----+----+----+----+
+// | P1 | p0 | q0 | q1 |
+// +----+----+----+----+
+//
+// p1 = r[P_AddrReg, 0]<16;8,2>
+// p0 = r[P_AddrReg, 16]<16;8,2>
+// q0 = r[Q_AddrReg, 0]<16;8,2>
+// q1 = r[Q_AddrReg, 16]<16;8,2>
+//
+/////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+// The region is both src and dest
+// P0-P3 and Q0-Q3 should be only used if they have not been modified to new values
+#undef P1
+#undef P0
+#undef Q0
+#undef Q1
+
+#define P1 r[P_AddrReg, 0]<16;8,2>:ub
+#define P0 r[P_AddrReg, 16]<16;8,2>:ub
+#define Q0 r[Q_AddrReg, 0]<16;8,2>:ub
+#define Q1 r[Q_AddrReg, 16]<16;8,2>:ub
+
+// New region as dest
+#undef NewP0
+#undef NewQ0
+
+#define NewP0 r[P_AddrReg, 16]<2>:ub
+#define NewQ0 r[Q_AddrReg, 0]<2>:ub
+
+// Filter one chroma edge - mbaff
+FILTER_UV_MBAFF:
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0x1112:w
+#endif
+ //---------- Derive filterSampleflag in AVC spec, equition (8-469) ----------
+
+ //===== Assume f0.0 contains MaskA when entering this routine
+// mov (1) f0.0:uw MaskA:uw
+
+ add (8) q0_p0(0)<1> Q0 -P0 // q0-p0
+ add (8) TempRow0(0)<1> P1 -P0 // p1-p0
+ add (8) TempRow1(0)<1> Q1 -Q0 // q1-q0
+
+ // Build FilterSampleFlag
+ // abs(q0-p0) < alpha
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)q0_p0(0) Mbaff_ALPHA(0)
+ // abs(p1-p0) < Beta
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)TempRow0(0) Mbaff_BETA(0)
+ // abs(q1-q0) < Beta
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)TempRow1(0) Mbaff_BETA(0)
+
+ //-----------------------------------------------------------------------------------------
+
+ // if
+ (f0.0) if (8) MBAFF_UV_ENDIF1
+ // For channels whose edge control map1 = 1 ---> perform de-blocking
+
+// mov (1) f0.1:w MaskB:w {NoMask} // Now check for which algorithm to apply
+
+ (f0.1) if (8) MBAFF_UV_ELSE2
+
+ // For channels whose edge control map2 = 1 ---> bS = 4 algorithm
+ // p0' = (2*p1 + P0 + q1 + 2) >> 2
+ // q0' = (2*q1 + q0 + p1 + 2) >> 2
+ //------------------------------------------------------------------------------------
+
+ // p0' = (2*p1 + p0 + q1 + 2) >> 2
+ add (8) acc0<1>:w Q1 2:w
+ mac (8) acc0<1>:w P1 2:w
+ add (8) acc0<1>:w acc0<8;8,1>:w P0
+ shr.sat (8) TempRow0B(0)<2> acc0<8;8,1>:w 2:w
+
+ // q0' = (2*q1 + q0 + p1 + 2) >> 2
+ add (8) acc0<1>:w P1 2:w
+ mac (8) acc0<1>:w Q1 2:w
+ add (8) acc0<1>:w acc0<8;8,1>:w Q0
+ shr.sat (8) TempRow1B(0)<2> acc0<8;8,1>:w 2:w
+
+ mov (8) NewP0 TempRow0B(0) // p0'
+ mov (8) NewQ0 TempRow1B(0) // q0'
+
+MBAFF_UV_ELSE2:
+ else (8) MBAFF_UV_ENDIF2
+ // For channels whose edge control map2 = 0 ---> bS < 4 algorithm
+
+ // tc_exp = tc0_exp + 1
+ add (8) tc_exp(0)<1> Mbaff_TC0(0) 1:w
+
+ // delta = Clip3(-tc, tc, ((((q0 - p0)<<2) + (p1-q1) + 4) >> 3))
+ // 4 * (q0-p0) + p1 - q1 + 4
+ add (8) acc0<1>:w P1 4:w
+ mac (8) acc0<1>:w q0_p0(0) 4:w
+ add (8) acc0<1>:w acc0<8;8,1>:w -Q1
+ shr (8) TempRow0(0)<1> acc0<8;8,1>:w 3:w
+
+ // tc clip
+ cmp.g.f0.0 (8) null:w TempRow0(0) tc_exp(0) // Clip if > tc0
+ cmp.l.f0.1 (8) null:w TempRow0(0) -tc_exp(0) // Clip if < -tc0
+
+ (f0.0) mov (8) TempRow0(0)<1> tc_exp(0)
+ (f0.1) mov (8) TempRow0(0)<1> -tc_exp(0)
+
+ // p0' = Clip1(p0 + delta) = Clip3(0, 0xFF, p0 + delta)
+ add.sat (8) TempRow1B(0)<2> P0 TempRow0(0) // p0+delta
+
+ // q0' = Clip1(q0 - delta) = Clip3(0, 0xFF, q0 - delta)
+ add.sat (8) TempRow0B(0)<2> Q0 -TempRow0(0) // q0-delta
+
+ mov (8) NewP0 TempRow1B(0) // p0'
+ mov (8) NewQ0 TempRow0B(0) // q0'
+
+ endif
+MBAFF_UV_ENDIF2:
+MBAFF_UV_ENDIF1:
+ endif
+
+RETURN
+
+
+
diff --git a/shaders/h264/ildb/AVC_ILDB_CloseGateway.asm b/shaders/h264/ildb/AVC_ILDB_CloseGateway.asm new file mode 100644 index 0000000..e522ce4 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_CloseGateway.asm @@ -0,0 +1,22 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//----- Close a Message Gateway -----
+
+#if defined(_DEBUG)
+ mov (1) EntrySignature:b 0x4444:w
+#endif
+
+// Message descriptor
+// bit 31 EOD
+// 27:24 FFID = 0x0011 for msg gateway
+// 23:20 msg length = 1 MRF
+// 19:16 Response length = 0
+// 1:0 SubFuncID = 01 for CloseGateway
+// Message descriptor: 0 000 0011 0001 0000 + 0 0 000000000000 01 ==> 0000 0011 0001 0000 0000 0000 0000 0001
+send (8) null:ud m7 r0.0:ud MSG_GW CGWMSGDSC
diff --git a/shaders/h264/ildb/AVC_ILDB_Dep_Check.asm b/shaders/h264/ildb/AVC_ILDB_Dep_Check.asm new file mode 100644 index 0000000..70f8a55 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Dep_Check.asm @@ -0,0 +1,186 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//---------- Check dependency and spawn all MBs ----------
+
+// Launch the 1st round of child threads for Vertical ILDB
+#if defined(_DEBUG)
+ mov (1) EntrySignature:w 0x3333:w
+#endif
+
+//=====================================================================
+// Jump Table 1
+ // 0 0 ---> Goto ALL_SPAWNED
+ // 0 1 ---> Goto ALL_SPAWNED
+ // 1 0 ---> Goto SLEEP_ENTRY
+ // 1 1 ---> Goto POST_SLEEP
+ mov (2) JumpTable.0<1>:d 0:d { NoDDClr }
+#if defined(CHROMA_ROOT)
+ mov (1) JumpTable.2:d SLEEP_ENTRY_UV_ILDB_FRAME_IP-ALL_SPAWNED_UV_ILDB_FRAME_IP:d { NoDDClr, NoDDChk }
+ mov (1) JumpTable.3:d POST_SLEEP_UV_ILDB_FRAME_IP-ALL_SPAWNED_UV_ILDB_FRAME_IP:d { NoDDChk }
+#else
+ mov (1) JumpTable.2:d SLEEP_ENTRY_Y_ILDB_FRAME_IP-ALL_SPAWNED_Y_ILDB_FRAME_IP:d { NoDDClr, NoDDChk }
+ mov (1) JumpTable.3:d POST_SLEEP_Y_ILDB_FRAME_IP-ALL_SPAWNED_Y_ILDB_FRAME_IP:d { NoDDChk }
+#endif
+//=====================================================================
+
+ mov (2) f0.0<1>:w 0:w
+
+ // Get m0 most of fields ready for URB write
+ mov (8) MRF0<1>:ud MSGSRC.0<8;8,1>:ud
+
+ // Add child kernel offset
+ add (1) CT_R0Hdr.2:ud r0.2:ud CHILD_OFFSET:w
+
+ // Init
+ mov (1) Col_Boundary:w 2:w
+ mov (1) Row_Boundary:w LastRow:w
+ mov (1) TopRowForScan:w 0:w
+ mov (2) OutstandingThreads<1>:w 0:w
+
+ // Init Scoreboard (idle = 0x00FF, busy = 0x0000)
+ // Low word is saved col. High word is busy/idle status
+ mov (16) GatewayAperture(0)<1> 0x00FF00FF:ud // Init r6-r7
+ mov (16) GatewayAperture(2)<1> 0x00FF00FF:ud // Init r8-r9
+ mov (16) GatewayAperture(4)<1> 0x00FF00FF:ud // Init r10-r11
+ mov (16) GatewayAperture(6)<1> 0x00FF00FF:ud // Init r12-r13
+ mov (16) GatewayAperture(8)<1> 0x00FF00FF:ud // Init r14-r15
+
+ mul (1) StatusAddr:w CurRow:w 4:w // dword to bytes offset conversion
+
+ //=====================================================================
+
+//SPAWN_LOOP:
+ //===== OutstandingThreads < ThreadLimit ? ============================
+ cmp.l.f0.1 (1) null:w OutstandingThreads:w ThreadLimit:w // Check the thread limit
+#if defined(CHROMA_ROOT)
+ (f0.1) jmpi ILDB_LABEL(POST_SLEEP_UV)
+#else // LUMA_ROOT
+ (f0.1) jmpi ILDB_LABEL(POST_SLEEP_Y)
+#endif
+
+#if defined(CHROMA_ROOT)
+ILDB_LABEL(SLEEP_ENTRY_UV):
+#else // LUMA_ROOT
+ILDB_LABEL(SLEEP_ENTRY_Y):
+#endif
+ //===== Goto Sleep ====================================================
+ // Either reached max thread limit or no child thread can be spawned due to dependency.
+ add (1) OutstandingThreads:w OutstandingThreads:w -1:w // Do this before wait is faster
+ wait n0.0:d
+
+#if defined(CHROMA_ROOT)
+ILDB_LABEL(POST_SLEEP_UV):
+#else // LUMA_ROOT
+ILDB_LABEL(POST_SLEEP_Y):
+#endif
+ //===== Luma Status[CurRow] == busy ? =====
+ cmp.z.f0.0 (1) null:uw r[StatusAddr, GatewayApertureB+ScoreBd_Idx]:uw 0:uw // Check west neighbor
+ cmp.g.f0.1 (1) null:w CurCol:w LastCol:w // Check if the curCol > LastCol
+
+#if defined(CHROMA_ROOT)
+ mov (16) acc0.0<1>:w URBOffsetUVBase<0;1,0>:w // Add offset to UV base (MBsCntY * URB_EBTRIES_PER_MB)
+ mac (1) URBOffset:w CurRow:w 4:w // 4 entries per row
+#else
+ mul (1) URBOffset:w CurRow:w 4:w // 4 entries per row
+#endif
+
+#if defined(CHROMA_ROOT)
+ (f0.0) jmpi ILDB_LABEL(SLEEP_ENTRY_UV) // Current row has a child thread running, can not spawn a new child thread, go back to sleep
+ (f0.1) jmpi ILDB_LABEL(NEXT_MB_UV) // skip MB if the curCol > LastCol
+#else // LUMA_ROOT
+ (f0.0) jmpi ILDB_LABEL(SLEEP_ENTRY_Y) // Current row has a child thread running, can not spawn a new child thread, go back to sleep
+ (f0.1) jmpi ILDB_LABEL(NEXT_MB_Y) // skip MB if the curCol > LastCol
+#endif
+
+ //========== Spwan a child thread ========================================
+ // Save cur col and set Status[CurRow] to busy
+ mov (2) r[StatusAddr, GatewayApertureB]<1>:uw CurColB<2;2,1>:ub // Store the new col
+
+ // Increase OutstandingThreads and ProcessedMBs by 1
+ add (2) OutstandingThreads<1>:w OutstandingThreads<2;2,1>:w 1:w
+
+ #include "AVC_ILDB_SpawnChild.asm"
+
+ //===== Find next MB ===================================================
+#if defined(CHROMA_ROOT)
+ILDB_LABEL(NEXT_MB_UV):
+#else // LUMA_ROOT
+ILDB_LABEL(NEXT_MB_Y):
+#endif
+ // Check pic boundary, results are in f0.0 bit0 and bit1
+ cmp.ge.f0.0 (2) null<1>:w CurCol<2;2,1>:w Col_Boundary<2;2,1>:w
+
+ // Update TopRowForScan if the curCol = LastCol
+ (f0.1) add (1) TopRowForScan:w CurRow:w 1:w
+
+// cmp.l.f0.1 (1) null<1>:w ProcessedMBs:w TotalBlocks:w // Processed all blocks ?
+ // 2 sets compare
+ // ProcessedMBs:w < TotalBlocks:w OutstandingThreads:w < ThreadLimit:wProcessedMBs:w
+ // 0 0 ---> Goto ALL_SPAWNED
+ // 0 1 ---> Goto ALL_SPAWNED
+ // 1 0 ---> Goto SLEEP_ENTRY
+ // 1 1 ---> Goto POST_SLEEP
+ cmp.l.f0.1 (2) null<1>:w OutstandingThreads<2;2,1>:w ThreadLimit<2;2,1>:w
+
+ // Just do it in stalled cycles
+ mov (1) acc0.0:w 4:w
+ mac (1) StatusAddr:w CurRow:w 4:w // dword to bytes offset conversion
+ add (2) CurCol<1>:w CurCol<2;2,1>:w StepToNextMB<2;2,1>:b // CurCol -= 2 and CurRow += 1
+
+ // Set f0.0 if turning around is needed, assuming bit 15 - 2 are zeros for correct comparison.
+ cmp.nz.f0.0 (1) null<1>:w f0.0:w 0x01:w
+
+ mul (1) JumpAddr:w f0.1:w 4:w // byte offet in dword count
+
+ // The next MB is at the row TopRowForScan
+ (f0.0) mul (1) StatusAddr:w TopRowForScan:w 4:w // dword to bytes offset conversion
+ (f0.0) mov (1) CurRow:w TopRowForScan:w { NoDDClr } // Restart from the top row that has MBs not deblocked yet.
+ (f0.0) add (1) CurCol:w r[StatusAddr, GatewayApertureB]:uw 1:w { NoDDChk }
+
+ //===== Processed all blocks ? =========================================
+ // (f0.1) jmpi SPAWN_LOOP
+
+ jmpi r[JumpAddr, JUMPTABLE_BASE]:d
+//JUMP_BASE:
+
+ //======================================================================
+
+ // All MB are spawned at this point, check for outstanding thread count
+#if defined(CHROMA_ROOT)
+ILDB_LABEL(ALL_SPAWNED_UV):
+#else // LUMA_ROOT
+ILDB_LABEL(ALL_SPAWNED_Y):
+#endif
+ cmp.e.f0.1 (1) null:w OutstandingThreads:w 0:w // Check before goto sleep
+#if defined(CHROMA_ROOT)
+ (f0.1) jmpi ILDB_LABEL(ALL_DONE_UV)
+#else // LUMA_ROOT
+ (f0.1) jmpi ILDB_LABEL(ALL_DONE_Y)
+#endif
+
+ wait n0.0:d // Wake up by a finished child thread
+ add (1) OutstandingThreads:w OutstandingThreads:w -1:w
+
+#if defined(CHROMA_ROOT)
+ // One thread is free and give it to luma thread limit --- Increase luma thread limit by one.
+ #include "AVC_ILDB_LumaThrdLimit.asm"
+#endif
+
+#if defined(CHROMA_ROOT)
+ jmpi ILDB_LABEL(ALL_SPAWNED_UV) // Waked up and goto dependency check
+#else // LUMA_ROOT
+ jmpi ILDB_LABEL(ALL_SPAWNED_Y) // Waked up and goto dependency check
+#endif
+
+ // All child threads are finsihed at this point
+#if defined(CHROMA_ROOT)
+ILDB_LABEL(ALL_DONE_UV):
+#else // LUMA_ROOT
+ILDB_LABEL(ALL_DONE_Y):
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_UV_h.asm b/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_UV_h.asm new file mode 100644 index 0000000..ff807d5 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_UV_h.asm @@ -0,0 +1,223 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC ILDB filter horizontal Mbaff UV ///////////////////////////////////////////////////////
+//
+// This filter code prepares the src data and control data for ILDB filtering on all horizontal edges of UV.
+//
+// It sssumes the data for horizontal de-blocking is already transposed.
+//
+// Chroma:
+//
+// +-------+-------+ H0 Edge
+// | | |
+// | | |
+// | | |
+// +-------+-------+ H1 Edge
+// | | |
+// | | |
+// | | |
+// +-------+-------+
+//
+/////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xBBBC:w
+#endif
+
+//=============== Chroma deblocking ================
+
+//---------- Deblock UV external top edge ----------
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterTopMbEdgeFlag:w // Check for FilterTopMbEdgeFlag
+
+ mov (1) f0.1:w DualFieldMode:w // Check for dual field mode
+
+ // Get Luma maskA and maskB
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz0]<0;1,0>:uw RRampW(0)
+ shr (16) TempRow1(0)<1> r[ECM_AddrReg, wEdgeCntlMapB_ExtTopHorz0]<0;1,0>:uw RRampW(0)
+
+ (f0.0) jmpi H0_UV_DONE // Skip H0 UV edge
+
+ (f0.1) jmpi DUAL_FIELD_UV
+
+ // Non dual field mode
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+ and.nz.f0.1 (8) null:w TempRow1(0)<16;8,2> 1:w
+
+ // Ext U
+ // p1 = Prev MB U row 0
+ // p0 = Prev MB U row 1
+ // q0 = Cur MB U row 0
+ // q1 = Cur MB U row 1
+ mov (1) P_AddrReg:w PREV_MB_U_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_U_BASE:w { NoDDChk }
+
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaTop0_Cb]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaTop0_Cb]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h00_0_Cb]<1;2,0>:ub
+
+ // Store UV MaskA and MaskB
+ mov (2) MaskA<1>:uw f0.0<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1)
+
+ // Ext V
+ mov (1) P_AddrReg:w PREV_MB_V_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_V_BASE:w { NoDDChk }
+
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaTop0_Cr]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaTop0_Cr]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h00_0_Cr]<1;2,0>:ub
+
+ // Set UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1)
+
+ jmpi H0_UV_DONE
+
+DUAL_FIELD_UV:
+ // Dual field mode, FieldModeCurrentMbFlag=0 && FieldModeAboveMbFlag=1
+
+ //===== Ext U, Top field
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+ and.nz.f0.1 (8) null:w TempRow1(0)<16;8,2> 1:w
+
+ mov (1) P_AddrReg:w ABOVE_CUR_MB_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w ABOVE_CUR_MB_BASE+32:w { NoDDChk }
+
+ mov (16) ABOVE_CUR_MB_UW(0)<1> PREV_MB_UW(0, 0)<16;8,1> // Copy p1, p0
+ mov (16) ABOVE_CUR_MB_UW(1)<1> SRC_UW(0, 0)<16;8,1> // Copy q1, q0
+
+ //===== Ext U, top field
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaTop0_Cb]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaTop0_Cb]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h00_0_Cb]<1;2,0>:ub
+
+ // Store UV MaskA and MaskB
+ mov (2) MaskA<1>:uw f0.0<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1) // Ext U, top field
+
+ //===== Ext V, top field
+ mov (1) P_AddrReg:w ABOVE_CUR_MB_BASE+1:w { NoDDClr }
+ mov (1) Q_AddrReg:w ABOVE_CUR_MB_BASE+33:w { NoDDChk }
+
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaTop0_Cr]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaTop0_Cr]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h00_0_Cr]<1;2,0>:ub
+
+ // Set UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1) // Ext U, top field
+
+ // Prefetch for bottom field
+ // Get bot field Luma maskA and maskB
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz1]<0;1,0>:uw RRampW(0)
+ shr (16) TempRow1(0)<1> r[ECM_AddrReg, wEdgeCntlMapB_ExtTopHorz1]<0;1,0>:uw RRampW(0)
+
+ // Save deblocked top field rows
+ mov (8) PREV_MB_UW(1, 0)<1> ABOVE_CUR_MB_UW(0, 8) // Copy p0
+ mov (8) SRC_UW(0, 0)<1> ABOVE_CUR_MB_UW(1, 0) // Copy q0
+ //==========================================================================
+
+ //===== Ext U, Bot field
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+ and.nz.f0.1 (8) null:w TempRow1(0)<16;8,2> 1:w
+
+ mov (1) P_AddrReg:w ABOVE_CUR_MB_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w ABOVE_CUR_MB_BASE+32:w { NoDDChk }
+
+ mov (16) ABOVE_CUR_MB_UW(0)<1> PREV_MB_UW(0, 8)<16;8,1> // Copy p1, p0
+ mov (16) ABOVE_CUR_MB_UW(1)<1> SRC_UW(0, 8)<16;8,1> // Copy q1, q0
+
+ //===== Ext U, bottom field
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaTop1_Cb]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaTop1_Cb]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h00_1_Cb]<1;2,0>:ub
+
+ // Store UV MaskA and MaskB
+ mov (2) MaskA<1>:uw f0.0<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1) // Ext U, bottom field
+
+ //===== Ext V, bot field
+ mov (1) P_AddrReg:w ABOVE_CUR_MB_BASE+1:w { NoDDClr }
+ mov (1) Q_AddrReg:w ABOVE_CUR_MB_BASE+33:w { NoDDChk }
+
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaTop1_Cr]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaTop1_Cr]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h00_1_Cr]<1;2,0>:ub
+
+ // Set UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1) // Ext V, bottom field
+
+ // Save deblocked bot field rows
+ mov (8) PREV_MB_UW(1, 8)<1> ABOVE_CUR_MB_UW(0, 8) // Copy p0
+ mov (8) SRC_UW(0, 8)<1> ABOVE_CUR_MB_UW(1, 0) // Copy q0
+ //========================================
+
+H0_UV_DONE:
+
+//---------- Deblock U internal horz middle edge ----------
+
+ //***** Need to take every other bit to form U maskA in core
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMap_IntMidHorz]<0;1,0>:uw RRampW(0)
+
+ // p1 = Cur MB U row 2
+ // p0 = Cur MB U row 3
+ // q0 = Cur MB U row 4
+ // q1 = Cur MB U row 5
+ mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDChk }
+
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaInternal_Cb]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaInternal_Cb]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h20_Cb]<1;2,0>:ub
+
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+
+ // Store UV MaskA and MaskB
+ mov (1) f0.1:uw 0:w
+ mov (1) MaskB:uw 0:w { NoDDClr }
+ mov (1) MaskA:uw f0.0:uw { NoDDChk }
+
+ CALL(FILTER_UV_MBAFF, 1)
+//-----------------------------------------------
+
+
+//---------- Deblock V internal horz middle edge ----------
+
+ // p1 = Cur MB V row 2
+ // p0 = Cur MB V row 3
+ // q0 = Cur MB V row 4
+ // q1 = Cur MB V row 5
+ mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDChk }
+
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaInternal_Cr]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaInternal_Cr]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h20_Cr]<1;2,0>:ub
+
+ // Set UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1)
+//-----------------------------------------------
+
+
diff --git a/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_UV_v.asm b/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_UV_v.asm new file mode 100644 index 0000000..c0f2678 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_UV_v.asm @@ -0,0 +1,209 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC LDB filter vertical Mbaff UV ///////////////////////////////////////////////////////
+//
+// This filter code prepares the src data and control data for ILDB filtering on all vertical edges of UV.
+//
+// It sssumes the data for vertical de-blocking is already transposed.
+//
+// Chroma:
+//
+// +-------+-------+
+// | | |
+// | | |
+// | | |
+// +-------+-------+
+// | | |
+// | | |
+// | | |
+// +-------+-------+
+//
+// V0 V1
+// Edge Edge
+//
+/////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xBBBC:w
+#endif
+
+//=============== Chroma deblocking ================
+
+//---------- Deblock U external left edge ----------
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterLeftMbEdgeFlag:w // Check for FilterLeftMbEdgeFlag
+
+ cmp.z.f0.1 (1) null:w VertEdgePattern:uw LEFT_FIELD_CUR_FRAME:w
+
+ // Get Luma maskA and maskB
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMapA_ExtLeftVert0]<0;1,0>:uw RRampW(0)
+ shr (16) TempRow1(0)<1> r[ECM_AddrReg, wEdgeCntlMapB_ExtLeftVert0]<0;1,0>:uw RRampW(0)
+
+ (f0.0) jmpi BYPASS_V0_UV // Do not deblock Left ext edge
+
+ cmp.z.f0.0 (1) null:w VertEdgePattern:uw LEFT_FRAME_CUR_FIELD:w
+
+ (-f0.1) jmpi V0_U_NEXT1 // Jump if not LEFT_FIELD_CUR_FRAME
+
+ //----- For LEFT_FIELD_CUR_FRAME
+
+ // Extract UV MaskA and MaskB from every other 2 bits of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<4;2,1> 1:w
+ and.nz.f0.1 (8) null:w TempRow1(0)<4;2,1> 1:w
+
+ // For FieldModeLeftMbFlag=1 && FieldModeCurrentMbFlag=0
+ mov (4) Mbaff_ALPHA(0,0)<2> r[ECM_AddrReg, bAlphaLeft0_Cb]<0;1,0>:ub { NoDDClr }
+ mov (4) Mbaff_ALPHA(0,1)<2> r[ECM_AddrReg, bAlphaLeft1_Cb]<0;1,0>:ub { NoDDChk }
+ mov (4) Mbaff_BETA(0,0)<2> r[ECM_AddrReg, bBetaLeft0_Cb]<0;1,0>:ub { NoDDClr }
+ mov (4) Mbaff_BETA(0,1)<2> r[ECM_AddrReg, bBetaLeft1_Cb]<0;1,0>:ub { NoDDChk }
+ mov (4) Mbaff_TC0(0,0)<2> r[ECM_AddrReg, bTc0_v00_0_Cb]<4;4,1>:ub { NoDDClr }
+ mov (4) Mbaff_TC0(0,1)<2> r[ECM_AddrReg, bTc0_v00_1_Cb]<4;4,1>:ub { NoDDChk }
+
+ jmpi V0_U_NEXT3
+
+V0_U_NEXT1:
+
+ (-f0.0) jmpi V0_U_NEXT2 // Jump if not LEFT_FRAME_CUR_FIELD
+
+ //----- For LEFT_FRAME_CUR_FIELD
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+ and.nz.f0.1 (8) null:w TempRow1(0)<16;8,2> 1:w
+
+ // For FieldModeLeftMbFlag=0 && FieldModeCurrentMbFlag=1
+ mov (4) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Cb]<0;1,0>:ub { NoDDClr }
+ mov (4) Mbaff_ALPHA(0,4)<1> r[ECM_AddrReg, bAlphaLeft1_Cb]<0;1,0>:ub { NoDDChk }
+ mov (4) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaLeft0_Cb]<0;1,0>:ub { NoDDClr }
+ mov (4) Mbaff_BETA(0,4)<1> r[ECM_AddrReg, bBetaLeft1_Cb]<0;1,0>:ub { NoDDChk }
+ mov (4) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v00_0_Cb]<4;4,1>:ub { NoDDClr }
+ mov (4) Mbaff_TC0(0,4)<1> r[ECM_AddrReg, bTc0_v00_1_Cb]<4;4,1>:ub { NoDDChk }
+
+ jmpi V0_U_NEXT3
+
+V0_U_NEXT2:
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+ and.nz.f0.1 (8) null:w TempRow1(0)<16;8,2> 1:w
+
+ // Both are frames or fields
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Cb]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaLeft0_Cb]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v00_0_Cb]<1;2,0>:ub
+
+V0_U_NEXT3:
+
+ // p1 = Prev MB U row 0
+ // p0 = Prev MB U row 1
+ // q0 = Cur MB U row 0
+ // q1 = Cur MB U row 1
+ mov (1) P_AddrReg:w PREV_MB_U_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_U_BASE:w { NoDDChk }
+
+ // Store UV MaskA and MaskB
+ mov (2) MaskA<1>:uw f0.0<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1)
+//-----------------------------------------------
+
+//---------- Deblock V external left edge ----------
+
+ // No change to MaskA and MaskB
+
+ cmp.z.f0.0 (4) null:w VertEdgePattern:uw LEFT_FIELD_CUR_FRAME:w
+ cmp.z.f0.1 (4) null:w VertEdgePattern:uw LEFT_FRAME_CUR_FIELD:w
+
+ // both are frame or field
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Cr]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaLeft0_Cr]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v00_0_Cr]<1;2,0>:ub
+
+ // p1 = Prev MB V row 0
+ // p0 = Prev MB V row 1
+ // q0 = Cur MB V row 0
+ // q1 = Cur MB V row 1
+ mov (1) P_AddrReg:w PREV_MB_V_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_V_BASE:w { NoDDChk }
+
+ // For FieldModeLeftMbFlag=1 && FieldModeCurrentMbFlag=0
+ (f0.0) mov (4) Mbaff_ALPHA(0,0)<2> r[ECM_AddrReg, bAlphaLeft0_Cr]<0;1,0>:ub { NoDDClr }
+ (f0.0) mov (4) Mbaff_ALPHA(0,1)<2> r[ECM_AddrReg, bAlphaLeft1_Cr]<0;1,0>:ub { NoDDChk }
+ (f0.0) mov (4) Mbaff_BETA(0,0)<2> r[ECM_AddrReg, bBetaLeft0_Cr]<0;1,0>:ub { NoDDClr }
+ (f0.0) mov (4) Mbaff_BETA(0,1)<2> r[ECM_AddrReg, bBetaLeft1_Cr]<0;1,0>:ub { NoDDChk }
+ (f0.0) mov (4) Mbaff_TC0(0,0)<2> r[ECM_AddrReg, bTc0_v00_0_Cr]<4;4,1>:ub { NoDDClr }
+ (f0.0) mov (4) Mbaff_TC0(0,1)<2> r[ECM_AddrReg, bTc0_v00_1_Cr]<4;4,1>:ub { NoDDChk }
+
+ // For FieldModeLeftMbFlag=0 && FieldModeCurrentMbFlag=1
+ (f0.1) mov (4) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Cr]<0;1,0>:ub { NoDDClr }
+ (f0.1) mov (4) Mbaff_ALPHA(0,4)<1> r[ECM_AddrReg, bAlphaLeft1_Cr]<0;1,0>:ub { NoDDChk }
+ (f0.1) mov (4) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaLeft0_Cr]<0;1,0>:ub { NoDDClr }
+ (f0.1) mov (4) Mbaff_BETA(0,4)<1> r[ECM_AddrReg, bBetaLeft1_Cr]<0;1,0>:ub { NoDDChk }
+ (f0.1) mov (4) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v00_0_Cr]<4;4,1>:ub { NoDDClr }
+ (f0.1) mov (4) Mbaff_TC0(0,4)<1> r[ECM_AddrReg, bTc0_v00_1_Cr]<4;4,1>:ub { NoDDChk }
+
+ // Set UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV_MBAFF, 1)
+//-----------------------------------------------
+
+BYPASS_V0_UV:
+ // Set EdgeCntlMap2 = 0, so it always uses bS < 4 algorithm.
+ // Same alpha and beta for all internal vert and horiz edges
+
+//---------- Deblock U internal vert middle edge ----------
+
+ //***** Need to take every other bit to form U or V maskA
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMap_IntMidVert]<0;1,0>:uw RRampW(0)
+
+ // p1 = Cur MB U row 2
+ // p0 = Cur MB U row 3
+ // q0 = Cur MB U row 4
+ // q1 = Cur MB U row 5
+ mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDClr } // Skip 2 U rows and 2 V rows
+ mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDChk }
+
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaInternal_Cb]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaInternal_Cb]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v02_Cb]<1;2,0>:ub
+
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+
+ // Store MaskA and MaskB
+ mov (1) f0.1:uw 0:w
+ mov (1) MaskB:uw 0:w { NoDDClr }
+ mov (1) MaskA:uw f0.0:uw { NoDDChk }
+
+ CALL(FILTER_UV_MBAFF, 1)
+
+//-----------------------------------------------
+
+
+//---------- Deblock V internal vert middle edge ----------
+
+ // P1 = Cur MB V row 2
+ // P0 = Cur MB V row 3
+ // Q0 = Cur MB V row 4
+ // Q1 = Cur MB V row 5
+ mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDClr } // Skip 2 U rows and 2 V rows
+ mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDChk }
+
+ // Put MaskA into f0.0
+ // Put MaskB into f0.1
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaInternal_Cr]<0;1,0>:ub
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaInternal_Cr]<0;1,0>:ub
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v02_Cr]<1;2,0>:ub
+
+ CALL(FILTER_UV_MBAFF, 1)
+
+//-----------------------------------------------
+
diff --git a/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_Y_h.asm b/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_Y_h.asm new file mode 100644 index 0000000..a98b024 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_Y_h.asm @@ -0,0 +1,234 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC ILDB filter horizontal Mbaff Y ///////////////////////////////////////////////////////
+//
+// This filter code prepares the src data and control data for ILDB filtering on all horizontal edges of Y.
+//
+// It sssumes the data for horizontal de-blocking is already transposed.
+//
+// Luma:
+//
+// +-------+-------+-------+-------+ H0 Edge
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+ H1 Edge
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+ H2 Edge
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+ H3 Edge
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+//
+/////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xBBBB:w
+#endif
+
+
+//========== Luma deblocking ==========
+
+
+//---------- Deblock Y external top edge (H0) ----------
+
+ // Bypass deblocking if it is the top edge of the picture.
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterTopMbEdgeFlag:w // Check for FilterTopMbEdgeFlag
+ mov (1) f0.1:w DualFieldMode:w // Check for dual field mode
+
+ // Non dual field mode
+
+ // Get (alpha >> 2) + 2
+ shr (16) Mbaff_ALPHA2(0,0)<1> r[ECM_AddrReg, bAlphaTop0_Y]<0;1,0>:ub 2:w // alpha >> 2
+
+ mov (2) MaskA<1>:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz0]<2;2,1>:uw
+
+ // Ext Y
+ mov (16) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaTop0_Y]<0;1,0>:ub
+ mov (16) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaTop0_Y]<0;1,0>:ub
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h00_0_Y]<1;4,0>:ub
+
+ add (16) Mbaff_ALPHA2(0,0)<1> Mbaff_ALPHA2(0,0)<16;16,1> 2:w // alpha2 = (alpha >> 2) + 2
+
+ (f0.0) jmpi H0_Y_DONE // Skip Ext Y deblocking
+ (f0.1) jmpi DUAL_FIELD_Y
+
+ mov (1) P_AddrReg:w PREV_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_Y_BASE:w { NoDDChk }
+
+ CALL(FILTER_Y_MBAFF, 1) // Non dual field deblocking
+
+ jmpi H0_Y_DONE
+
+DUAL_FIELD_Y:
+ // Dual field mode, FieldModeCurrentMbFlag=0 && FieldModeAboveMbFlag=1
+
+ mov (1) P_AddrReg:w ABOVE_CUR_MB_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w ABOVE_CUR_MB_BASE+64:w { NoDDChk }
+
+ // Must use PREV_MB_YW. TOP_MB_YW is not big enough.
+ // Get top field rows
+ mov (16) ABOVE_CUR_MB_YW(0)<1> PREV_MB_YW(0, 0)<16;8,1> // Copy p3, p2
+ mov (16) ABOVE_CUR_MB_YW(1)<1> PREV_MB_YW(2, 0)<16;8,1> // Copy p1, p0
+ mov (16) ABOVE_CUR_MB_YW(2)<1> SRC_YW(0, 0)<16;8,1> // Copy q0, q1
+ mov (16) ABOVE_CUR_MB_YW(3)<1> SRC_YW(2, 0)<16;8,1> // Copy q2, q3
+
+ CALL(FILTER_Y_MBAFF, 1) // Ext Y, top field
+
+ // Save deblocked top field rows
+ mov (8) PREV_MB_YW(1, 0)<1> ABOVE_CUR_MB_YW(0, 8) // Copy p2
+ mov (8) PREV_MB_YW(2, 0)<1> ABOVE_CUR_MB_YW(1, 0) // Copy p1
+ mov (8) PREV_MB_YW(3, 0)<1> ABOVE_CUR_MB_YW(1, 8) // Copy p0
+ mov (8) SRC_YW(0, 0)<1> ABOVE_CUR_MB_YW(2, 0) // Copy q0
+ mov (8) SRC_YW(1, 0)<1> ABOVE_CUR_MB_YW(2, 8) // Copy q1
+ mov (8) SRC_YW(2, 0)<1> ABOVE_CUR_MB_YW(3, 0) // Copy q2
+
+ //==================================================================================
+ // Bottom field
+
+ // Get (alpha >> 2) + 2
+ shr (16) Mbaff_ALPHA2(0,0)<1> r[ECM_AddrReg, bAlphaTop1_Y]<0;1,0>:ub 2:w // alpha >> 2
+
+ mov (1) P_AddrReg:w ABOVE_CUR_MB_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w ABOVE_CUR_MB_BASE+64:w { NoDDChk }
+
+ // Get bot field rows
+ mov (16) ABOVE_CUR_MB_YW(0)<1> PREV_MB_YW(0, 8)<16;8,1> // Copy p3, p2
+ mov (16) ABOVE_CUR_MB_YW(1)<1> PREV_MB_YW(2, 8)<16;8,1> // Copy p1, p0
+ mov (16) ABOVE_CUR_MB_YW(2)<1> SRC_YW(0, 8)<16;8,1> // Copy q0, q1
+ mov (16) ABOVE_CUR_MB_YW(3)<1> SRC_YW(2, 8)<16;8,1> // Copy q2, q3
+
+ mov (2) MaskA<1>:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz1]<2;2,1>:uw
+
+ mov (16) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaTop1_Y]<0;1,0>:ub
+ mov (16) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaTop1_Y]<0;1,0>:ub
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h00_1_Y]<1;4,0>:ub
+
+ add (16) Mbaff_ALPHA2(0,0)<1> Mbaff_ALPHA2(0,0)<16;16,1> 2:w // alpha2 = (alpha >> 2) + 2
+
+ CALL(FILTER_Y_MBAFF, 1) // Ext Y, bot field
+
+ // Save deblocked top field rows
+ mov (8) PREV_MB_YW(1, 8)<1> ABOVE_CUR_MB_YW(0, 8) // Copy p2
+ mov (8) PREV_MB_YW(2, 8)<1> ABOVE_CUR_MB_YW(1, 0) // Copy p1
+ mov (8) PREV_MB_YW(3, 8)<1> ABOVE_CUR_MB_YW(1, 8) // Copy p0
+ mov (8) SRC_YW(0, 8)<1> ABOVE_CUR_MB_YW(2, 0) // Copy q0
+ mov (8) SRC_YW(1, 8)<1> ABOVE_CUR_MB_YW(2, 8) // Copy q1
+ mov (8) SRC_YW(2, 8)<1> ABOVE_CUR_MB_YW(3, 0) // Copy q2
+ //==================================================================================
+
+H0_Y_DONE:
+
+//BYPASS_H0_Y:
+//------------------------------------------------------------------
+ // Same alpha, alpha2, beta and MaskB for all internal edges
+
+ // Get (alpha >> 2) + 2
+ shr (16) Mbaff_ALPHA2(0,0)<1> r[ECM_AddrReg, bAlphaInternal_Y]<0;1,0>:ub 2:w // alpha >> 2
+
+ // alpha = bAlphaInternal_Y
+ // beta = bBetaInternal_Y
+ mov (16) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaInternal_Y]<0;1,0>:ub
+ mov (16) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaInternal_Y]<0;1,0>:ub
+
+ mov (1) MaskB:uw 0:w // Set MaskB = 0 for all 3 edges, so it always uses bS < 4 algorithm.
+
+ add (16) Mbaff_ALPHA2(0,0)<1> Mbaff_ALPHA2(0,0)<16;16,1> 2:w // alpha2 = (alpha >> 2) + 2
+
+//---------- Deblock Y internal top edge (H1) ----------
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+// (f0.0) jmpi BYPASS_H1_Y
+
+ // p3 = Cur MB Y row 0 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 1 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 2 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 3 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 4 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 5 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 6 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 7 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 4*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntTopHorz]:uw
+
+ // tc0 has bTc0_h13_Y + bTc0_h12_Y + bTc0_h11_Y + bTc0_h10_Y
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h10_Y]<1;4,0>:ub
+
+// CALL(FILTER_Y_MBAFF, 1)
+ PRED_CALL(-f0.0, FILTER_Y_MBAFF, 1)
+
+//BYPASS_H1_Y:
+//------------------------------------------------------------------
+
+
+//---------- Deblock Y internal mid horizontal edge (H2) ----------
+
+ // Bypass deblocking if FilterInternal8x8EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal8x8EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+// (f0.0) jmpi BYPASS_H2_Y
+
+ // p3 = Cur MB Y row 4 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 5 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 6 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 7 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 8 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 9 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 10 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 11 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w 4*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 8*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntMidHorz]:uw
+
+ // tc0 has bTc0_h23_Y + bTc0_h22_Y + bTc0_h21_Y + bTc0_h20_Y
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h20_Y]<1;4,0>:ub
+
+// CALL(FILTER_Y_MBAFF, 1)
+ PRED_CALL(-f0.0, FILTER_Y_MBAFF, 1)
+
+//BYPASS_H2_Y:
+//-----------------------------------------------
+
+
+//---------- Deblock Y internal bottom edge (H3) ----------
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+// (f0.0) jmpi BYPASS_H3_Y
+
+ // p3 = Cur MB Y row 8 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 9 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 10 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 11 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 12 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 13 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 14 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 15 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w 8*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 12*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntBotHorz]:uw
+
+ // tc0 has bTc0_h33_Y + bTc0_h32_Y + bTc0_h31_Y + bTc0_h30_Y
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_h30_Y]<1;4,0>:ub
+
+// CALL(FILTER_Y_MBAFF, 1)
+ PRED_CALL(-f0.0, FILTER_Y_MBAFF, 1)
+
+//BYPASS_H3_Y:
+//-----------------------------------------------
diff --git a/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_Y_v.asm b/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_Y_v.asm new file mode 100644 index 0000000..7846168 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Filter_Mbaff_Y_v.asm @@ -0,0 +1,269 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC ILDB filter vertical Mbaff Y ///////////////////////////////////////////////////////
+//
+// This filter code prepares the src data and control data for ILDB filtering on all vertical edges of Y.
+//
+// It sssumes the data for vertical de-blocking is already transposed.
+//
+// Luma:
+//
+// +-------+-------+-------+-------+
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+//
+// V0 V1 V2 V3
+// Edge Edge Edge Edge
+//
+/////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xBBBB:w
+#endif
+
+
+//========== Luma deblocking ==========
+
+
+//---------- Deblock Y external left edge (V0) ----------
+
+ cmp.z.f0.0 (8) null:w VertEdgePattern:uw LEFT_FIELD_CUR_FRAME:w
+ cmp.z.f0.1 (8) null:w VertEdgePattern:uw LEFT_FRAME_CUR_FIELD:w
+
+ // Intial set for both are frame or field
+ mov (16) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Y]<0;1,0>:ub
+ mov (16) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaLeft0_Y]<0;1,0>:ub
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v00_0_Y]<1;4,0>:ub
+
+ // For FieldModeCurrentMbFlag=1 && FieldModeLeftMbFlag=0
+ (f0.0) mov (8) Mbaff_ALPHA(0,0)<2> r[ECM_AddrReg, bAlphaLeft0_Y]<0;1,0>:ub { NoDDClr }
+ (f0.0) mov (8) Mbaff_ALPHA(0,1)<2> r[ECM_AddrReg, bAlphaLeft1_Y]<0;1,0>:ub { NoDDChk }
+ (f0.0) mov (8) Mbaff_BETA(0,0)<2> r[ECM_AddrReg, bBetaLeft0_Y]<0;1,0>:ub { NoDDClr }
+ (f0.0) mov (8) Mbaff_BETA(0,1)<2> r[ECM_AddrReg, bBetaLeft1_Y]<0;1,0>:ub { NoDDChk }
+ (f0.0) mov (8) Mbaff_TC0(0,0)<2> r[ECM_AddrReg, bTc0_v00_0_Y]<1;2,0>:ub { NoDDClr }
+ (f0.0) mov (8) Mbaff_TC0(0,1)<2> r[ECM_AddrReg, bTc0_v00_1_Y]<1;2,0>:ub { NoDDChk }
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterLeftMbEdgeFlag:w // Check for FilterLeftMbEdgeFlag
+
+ // For FieldModeCurrentMbFlag=0 && FieldModeLeftMbFlag=1
+ (f0.1) mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Y]<0;1,0>:ub { NoDDClr }
+ (f0.1) mov (8) Mbaff_ALPHA(0,8)<1> r[ECM_AddrReg, bAlphaLeft1_Y]<0;1,0>:ub { NoDDChk }
+ (f0.1) mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaLeft0_Y]<0;1,0>:ub { NoDDClr }
+ (f0.1) mov (8) Mbaff_BETA(0,8)<1> r[ECM_AddrReg, bBetaLeft1_Y]<0;1,0>:ub { NoDDChk }
+ (f0.1) mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v00_0_Y]<1;2,0>:ub { NoDDClr }
+ (f0.1) mov (8) Mbaff_TC0(0,8)<1> r[ECM_AddrReg, bTc0_v00_1_Y]<1;2,0>:ub { NoDDChk }
+
+ // Get (alpha >> 2) + 2
+ shr (16) Mbaff_ALPHA2(0,0)<1> Mbaff_ALPHA(0) 2:w // alpha >> 2
+
+ // p3 = Prev MB Y row 0 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Prev MB Y row 1 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Prev MB Y row 2 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Prev MB Y row 3 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 0 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 1 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 2 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 3 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w PREV_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_Y_BASE:w { NoDDChk }
+
+ // Set MaskA and MaskB
+ mov (2) MaskA<1>:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtLeftVert0]<2;2,1>:uw
+
+ add (16) Mbaff_ALPHA2(0,0)<1> Mbaff_ALPHA2(0,0)<16;16,1> 2:w // alpha2 = (alpha >> 2) + 2
+
+// CALL(FILTER_Y_MBAFF, 1)
+ PRED_CALL(-f0.0, FILTER_Y_MBAFF, 1)
+
+//BYPASS_V0_Y:
+//------------------------------------------------------------------
+
+
+/*
+//---------- Deblock Y external left edge (V0) ----------
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterLeftMbEdgeFlag:w // Check for FilterLeftMbEdgeFlag
+ (f0.0) jmpi ILDB_LABEL(BYPASS_EXT_LEFT_EDGE_Y)
+
+ // Get vertical border edge control data
+
+// mov (1) f0.0 0:w
+ and (1) CTemp1_W:uw r[ECM_AddrReg, BitFlags]:ub FieldModeLeftMbFlag+FieldModeCurrentMbFlag:uw
+ cmp.z.f0.0 (1) null:w CTemp1_W:uw LEFT_FIELD_CUR_FRAME:w
+ (-f0.0) jmpi LEFT_EDGE_Y_NEXT1
+
+ // For FieldModeCurrentMbFlag=1 && FieldModeLeftMbFlag=0
+ mov (8) Mbaff_ALPHA(0,0)<2> r[ECM_AddrReg, bAlphaLeft0_Y]<0;1,0>:ub { NoDDClr }
+ mov (8) Mbaff_ALPHA(0,1)<2> r[ECM_AddrReg, bAlphaLeft1_Y]<0;1,0>:ub { NoDDChk }
+ mov (8) Mbaff_BETA(0,0)<2> r[ECM_AddrReg, bBetaLeft0_Y]<0;1,0>:ub { NoDDClr }
+ mov (8) Mbaff_BETA(0,1)<2> r[ECM_AddrReg, bBetaLeft1_Y]<0;1,0>:ub { NoDDChk }
+ mov (8) Mbaff_TC0(0,0)<2> r[ECM_AddrReg, bTc0_v00_0_Y]<1;2,0>:ub { NoDDClr }
+ mov (8) Mbaff_TC0(0,1)<2> r[ECM_AddrReg, bTc0_v00_1_Y]<1;2,0>:ub { NoDDChk }
+
+ jmpi LEFT_EDGE_Y_ALPHA_BETA_TC0_SELECTED
+
+LEFT_EDGE_Y_NEXT1:
+ cmp.z.f0.0 (1) null:w CTemp1_W:uw LEFT_FRAME_CUR_FIELD:w
+ (-f0.0) jmpi LEFT_EDGE_Y_NEXT2
+
+
+ // For FieldModeCurrentMbFlag=0 && FieldModeLeftMbFlag=1
+ mov (8) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Y]<0;1,0>:ub { NoDDClr }
+ mov (8) Mbaff_ALPHA(0,8)<1> r[ECM_AddrReg, bAlphaLeft1_Y]<0;1,0>:ub { NoDDChk }
+ mov (8) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaLeft0_Y]<0;1,0>:ub { NoDDClr }
+ mov (8) Mbaff_BETA(0,8)<1> r[ECM_AddrReg, bBetaLeft1_Y]<0;1,0>:ub { NoDDChk }
+ mov (8) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v00_0_Y]<1;2,0>:ub { NoDDClr }
+ mov (8) Mbaff_TC0(0,8)<1> r[ECM_AddrReg, bTc0_v00_1_Y]<1;2,0>:ub { NoDDChk }
+
+ jmpi LEFT_EDGE_Y_ALPHA_BETA_TC0_SELECTED
+
+LEFT_EDGE_Y_NEXT2:
+ // both are frame or field
+ mov (16) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Y]<0;1,0>:ub
+ mov (16) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaLeft0_Y]<0;1,0>:ub
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v00_0_Y]<1;4,0>:ub
+
+LEFT_EDGE_Y_ALPHA_BETA_TC0_SELECTED:
+
+ mov (2) MaskA<1>:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtLeftVert0]<2;2,1>:uw
+
+ // p3 = Prev MB Y row 0 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Prev MB Y row 1 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Prev MB Y row 2 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Prev MB Y row 3 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 0 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 1 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 2 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 3 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w PREV_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_Y_BASE:w { NoDDChk }
+
+ // Get (alpha >> 2) + 2
+ shr (16) Mbaff_ALPHA2(0,0)<1> r[ECM_AddrReg, bAlphaLeft0_Y]<0;1,0>:ub 2:w // alpha >> 2
+ add (16) Mbaff_ALPHA2(0,0)<1> Mbaff_ALPHA2(0,0)<16;16,1> 2:w // alpha2 = (alpha >> 2) + 2
+
+ CALL(FILTER_Y_MBAFF, 1)
+
+ILDB_LABEL(BYPASS_EXT_LEFT_EDGE_Y):
+//------------------------------------------------------------------
+*/
+
+ // Same alpha, alpha2, beta and MaskB for all internal edges
+
+ // Get (alpha >> 2) + 2
+ shr (16) Mbaff_ALPHA2(0,0)<1> r[ECM_AddrReg, bAlphaInternal_Y]<0;1,0>:ub 2:w // alpha >> 2
+
+ // alpha = bAlphaInternal_Y
+ // beta = bBetaInternal_Y
+ mov (16) Mbaff_ALPHA(0,0)<1> r[ECM_AddrReg, bAlphaInternal_Y]<0;1,0>:ub
+ mov (16) Mbaff_BETA(0,0)<1> r[ECM_AddrReg, bBetaInternal_Y]<0;1,0>:ub
+
+ mov (1) MaskB:uw 0:w // Set MaskB = 0 for all 3 edges, so it always uses bS < 4 algorithm.
+
+ add (16) Mbaff_ALPHA2(0,0)<1> Mbaff_ALPHA2(0,0)<16;16,1> 2:w // alpha2 = (alpha >> 2) + 2
+
+//---------- Deblock Y internal left edge (V1) ----------
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+// (f0.0) jmpi BYPASS_V1_Y
+
+ // p3 = Cur MB Y row 0 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 1 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 2 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 3 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 4 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 5 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 6 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 7 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 4*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]:uw
+
+ // tc0 has bTc0_v31_Y + bTc0_v21_Y + bTc0_v11_Y + bTc0_v01_Y
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v01_Y]<1;4,0>:ub
+
+// CALL(FILTER_Y_MBAFF, 1)
+ PRED_CALL(-f0.0, FILTER_Y_MBAFF, 1)
+
+BYPASS_V1_Y:
+//------------------------------------------------------------------
+
+
+//---------- Deblock Y internal mid vert edge (V2) ----------
+
+ // Bypass deblocking if FilterInternal8x8EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal8x8EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+// (f0.0) jmpi BYPASS_V2_Y
+
+ // p3 = Cur MB Y row 4 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 5 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 6 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 7 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 8 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 9 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 10 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 11 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w 4*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 8*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntMidVert]:uw
+
+ // tc0 has bTc0_v32_Y + bTc0_v22_Y + bTc0_v12_Y + bTc0_v02_Y
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v02_Y]<1;4,0>:ub
+
+// CALL(FILTER_Y_MBAFF, 1)
+ PRED_CALL(-f0.0, FILTER_Y_MBAFF, 1)
+
+BYPASS_V2_Y:
+//-----------------------------------------------
+
+
+//---------- Deblock Y interal right edge (V3) ----------
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+// (f0.0) jmpi BYPASS_V3_Y
+
+ // p3 = Cur MB Y row 8 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 9 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 10 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 11 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 12 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 13 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 14 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 15 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w 8*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 12*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntRightVert]:uw
+
+ // tc0 has bTc0_v33_Y + bTc0_v23_Y + bTc0_v13_Y + bTc0_v03_Y
+ mov (16) Mbaff_TC0(0,0)<1> r[ECM_AddrReg, bTc0_v03_Y]<1;4,0>:ub
+
+// CALL(FILTER_Y_MBAFF, 1)
+ PRED_CALL(-f0.0, FILTER_Y_MBAFF, 1)
+
+BYPASS_V3_Y:
+//-----------------------------------------------
diff --git a/shaders/h264/ildb/AVC_ILDB_Filter_UV_h.asm b/shaders/h264/ildb/AVC_ILDB_Filter_UV_h.asm new file mode 100644 index 0000000..168df0f --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Filter_UV_h.asm @@ -0,0 +1,145 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC ILDB filter horizontal UV ///////////////////////////////////////////////////////
+//
+// This filter code prepares the src data and control data for ILDB filtering on all horizontal edges of UV.
+//
+// It sssumes the data for horizontal de-blocking is already transposed.
+//
+// Chroma:
+//
+// +-------+-------+ H0 Edge
+// | | |
+// | | |
+// | | |
+// +-------+-------+ H1 Edge
+// | | |
+// | | |
+// | | |
+// +-------+-------+
+//
+/////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xBBBC:w
+#endif
+
+//=============== Chroma deblocking ================
+
+//---------- Deblock U external top edge ----------
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterTopMbEdgeFlag:w // Check for FilterTopMbEdgeFlag
+// (f0.0) jmpi BYPASS_EXT_TOP_EDGE_UV
+
+ // Get horizontal border edge control data.
+
+ //***** Need to take every other bit to form U maskA and mask B
+ // Get Luma maskA and maskB
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz0]<0;1,0>:uw RRampW(0)
+ shr (16) TempRow1(0)<1> r[ECM_AddrReg, wEdgeCntlMapB_ExtTopHorz0]<0;1,0>:uw RRampW(0)
+
+ (f0.0) jmpi ILDB_LABEL(BYPASS_EXT_TOP_EDGE_UV)
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+ and.nz.f0.1 (8) null:w TempRow1(0)<16;8,2> 1:w
+
+//---------- Deblock U external edge ----------
+ // p1 = Prev MB U row 0
+ // p0 = Prev MB U row 1
+ // q0 = Cur MB U row 0
+ // q1 = Cur MB U row 1
+// mov (1) P_AddrReg:w PREV_MB_U_BASE:w { NoDDClr }
+ mov (1) P_AddrReg:w TOP_MB_U_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_U_BASE:w { NoDDChk }
+
+ // alpha = bAlphaTop0_Cb, beta = bBetaTop0_Cb
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaTop0_Cb]<2;2,1>:ub { NoDDClr }
+ // tc0 has bTc0_h03_0_Cb + bTc0_h02_0_Cb + bTc0_h01_0_Cb + bTc0_h00_0_Cb
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h00_0_Cb]<4;4,1>:ub { NoDDChk }
+
+ // UV MaskA and MaskB
+ mov (2) MaskA<1>:uw f0.0<2;2,1>:uw
+
+ CALL(FILTER_UV, 1)
+
+//---------- Deblock V external top edge ----------
+ // p1 = Prev MB V row 0
+ // p0 = Prev MB V row 1
+ // q0 = Cur MB V row 0
+ // q1 = Cur MB V row 1
+// mov (1) P_AddrReg:w PREV_MB_V_BASE:w { NoDDClr }
+ mov (1) P_AddrReg:w TOP_MB_V_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_V_BASE:w { NoDDChk }
+
+ // alpha = bAlphaTop0_Cr, beta = bBetaTop0_Cr
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaTop0_Cr]<2;2,1>:ub { NoDDClr }
+
+ // tc0 has bTc0_h03_0_Cr + bTc0_h02_0_Cr + bTc0_h01_0_Cr + bTc0_h00_0_Cr
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h00_0_Cr]<4;4,1>:ub { NoDDChk }
+
+ // UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV, 1)
+
+ILDB_LABEL(BYPASS_EXT_TOP_EDGE_UV):
+
+ // Set EdgeCntlMap2 = 0, so it always uses bS < 4 algorithm.
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+// and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+// (f0.0) jmpi BYPASS_4x4_DEBLOCK_H
+
+//---------- Deblock U internal horz middle edge ----------
+
+ //***** Need to take every other bit to form U maskA
+ // Get Luma maskA and maskB
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMap_IntMidHorz]<0;1,0>:uw RRampW(0)
+
+ // p1 = Cur MB U row 2
+ // p0 = Cur MB U row 3
+ // q0 = Cur MB U row 4
+ // q1 = Cur MB U row 5
+ mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDClr } // Skip 2 U rows and 2 V rows
+ mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDChk }
+
+ // alpha = bAlphaInternal_Cb, beta = bBetaInternal_Cb
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaInternal_Cb]<2;2,1>:ub { NoDDClr }
+ // tc0 has bTc0_h23_Cb + bTc0_h22_Cb + bTc0_h21_Cb + bTc0_h20_Cb
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h20_Cb]<4;4,1>:ub { NoDDChk }
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+
+ // UV MaskA and MaskB
+ mov (1) f0.1:uw 0:w
+ mov (1) MaskB:uw 0:w { NoDDClr }
+ mov (1) MaskA:uw f0.0:uw { NoDDChk }
+
+ CALL(FILTER_UV, 1)
+
+//---------- Deblock V internal horz middle edge ----------
+ // p1 = Cur MB V row 2
+ // p0 = Cur MB V row 3
+ // q0 = Cur MB V row 4
+ // q1 = Cur MB V row 5
+ mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDClr } // Skip 2 U rows and 2 V rows
+ mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDChk }
+
+ // alpha = bAlphaInternal_Cr, beta = bBetaInternal_Cr
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaInternal_Cr]<2;2,1>:ub { NoDDClr }
+ // tc0 has bTc0_h23_Cr + bTc0_h22_Cr + bTc0_h21_Cr + bTc0_h20_Cr
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h20_Cr]<4;4,1>:ub { NoDDChk }
+
+ // UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV, 1)
+
+//BYPASS_4x4_DEBLOCK_H:
diff --git a/shaders/h264/ildb/AVC_ILDB_Filter_UV_v.asm b/shaders/h264/ildb/AVC_ILDB_Filter_UV_v.asm new file mode 100644 index 0000000..8d331a0 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Filter_UV_v.asm @@ -0,0 +1,145 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC LDB filter vertical UV ///////////////////////////////////////////////////////
+//
+// This filter code prepares the src data and control data for ILDB filtering on all vertical edges of UV.
+//
+// It sssumes the data for vertical de-blocking is already transposed.
+//
+// Chroma:
+//
+// +-------+-------+
+// | | |
+// | | |
+// | | |
+// +-------+-------+
+// | | |
+// | | |
+// | | |
+// +-------+-------+
+//
+// V0 V1
+// Edge Edge
+//
+/////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xBBBC:w
+#endif
+
+//=============== Chroma deblocking ================
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterLeftMbEdgeFlag:w // Check for FilterLeftMbEdgeFlag
+// (f0.0) jmpi BYPASS_EXT_LEFT_EDGE_UV
+
+ // Get vertical border edge control data.
+
+ // Get Luma maskA and maskB
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMapA_ExtLeftVert0]<0;1,0>:uw RRampW(0)
+ shr (16) TempRow1(0)<1> r[ECM_AddrReg, wEdgeCntlMapB_ExtLeftVert0]<0;1,0>:uw RRampW(0)
+
+ (f0.0) jmpi ILDB_LABEL(BYPASS_EXT_LEFT_EDGE_UV)
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+ and.nz.f0.1 (8) null:w TempRow1(0)<16;8,2> 1:w
+
+//---------- Deblock U external edge ----------
+ // p1 = Prev MB U row 0
+ // p0 = Prev MB U row 1
+ // q0 = Cur MB U row 0
+ // q1 = Cur MB U row 1
+ mov (1) P_AddrReg:w PREV_MB_U_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_U_BASE:w { NoDDChk }
+
+ // alpha = bAlphaLeft0_Cb, beta = bBetaLeft0_Cb
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaLeft0_Cb]<2;2,1>:ub { NoDDClr }
+ // tc0 has bTc0_v30_0_Cb + bTc0_v20_0_Cb + bTc0_v10_0_Cb + bTc0_v00_0_Cb
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_v00_0_Cb]<4;4,1>:ub { NoDDChk }
+
+ // UV MaskA and MaskB
+ mov (2) MaskA<1>:uw f0.0<2;2,1>:uw
+
+ CALL(FILTER_UV, 1)
+
+//---------- Deblock V external edge ----------
+ // p1 = Prev MB V row 0
+ // p0 = Prev MB V row 1
+ // q0 = Cur MB V row 0
+ // q1 = Cur MB V row 1
+ mov (1) P_AddrReg:w PREV_MB_V_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_V_BASE:w { NoDDChk }
+
+ // for vert edge: alpha = bAlphaLeft0_Cr, beta = bBetaLeft0_Cr
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaLeft0_Cr]<2;2,1>:ub { NoDDClr }
+
+ // tc0 has bTc0_v30_0_Cr + bTc0_v20_0_Cr + bTc0_v10_0_Cr + bTc0_v00_0_Cr
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_v00_0_Cr]<4;4,1>:ub { NoDDChk }
+
+ // UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV, 1)
+
+
+ILDB_LABEL(BYPASS_EXT_LEFT_EDGE_UV):
+ // Set EdgeCntlMap2 = 0, so it always uses bS < 4 algorithm.
+ // Same alpha and beta for all internal vert and horiz edges
+
+
+ //***** Need to take every other bit to form U or V maskA
+ // Get Luma maskA and maskB
+ shr (16) TempRow0(0)<1> r[ECM_AddrReg, wEdgeCntlMap_IntMidVert]<0;1,0>:uw RRampW(0)
+
+//---------- Deblock U internal edge ----------
+ // p1 = Cur MB U row 2
+ // p0 = Cur MB U row 3
+ // q0 = Cur MB U row 4
+ // q1 = Cur MB U row 5
+ mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_U_BASE:w { NoDDChk }
+
+ // alpha = bAlphaInternal_Cb, beta = bBetaInternal_Cb
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaInternal_Cb]<2;2,1>:ub { NoDDClr }
+
+ // tc0 has bTc0_v32_Cb + bTc0_v22_Cb + bTc0_v12_Cb + bTc0_v02_Cb
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_v02_Cb]<4;4,1>:ub { NoDDChk }
+
+ // Extract UV MaskA and MaskB from every other bit of Y masks
+ and.nz.f0.0 (8) null:w TempRow0(0)<16;8,2> 1:w
+
+ // UV MaskA and MaskB
+ mov (1) f0.1:uw 0:w
+ mov (1) MaskB:uw 0:w { NoDDClr }
+ mov (1) MaskA:uw f0.0:uw { NoDDChk }
+
+ CALL(FILTER_UV, 1)
+
+
+//---------- Deblock V internal edge ----------
+ // P1 = Cur MB V row 2
+ // P0 = Cur MB V row 3
+ // Q0 = Cur MB V row 4
+ // Q1 = Cur MB V row 5
+ mov (1) P_AddrReg:w 4*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 8*UV_ROW_WIDTH+SRC_MB_V_BASE:w { NoDDChk }
+
+ // alpha = bAlphaInternal_Cr, beta = bBetaInternal_Cr
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaInternal_Cr]<2;2,1>:ub { NoDDClr }
+
+ // tc0 has bTc0_v32_Cr + bTc0_v22_Cr + bTc0_v12_Cr + bTc0_v02_Cr
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_v02_Cr]<4;4,1>:ub { NoDDChk }
+
+ // UV MaskA and MaskB
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ CALL(FILTER_UV, 1)
+
+
+//BYPASS_4x4_DEBLOCK_V:
diff --git a/shaders/h264/ildb/AVC_ILDB_Filter_Y_h.asm b/shaders/h264/ildb/AVC_ILDB_Filter_Y_h.asm new file mode 100644 index 0000000..45ab4df --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Filter_Y_h.asm @@ -0,0 +1,199 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC ILDB filter horizontal Y ///////////////////////////////////////////////////////
+//
+// This filter code prepares the src data and control data for ILDB filtering on all horizontal edges of Y.
+//
+// It sssumes the data for horizontal de-blocking is already transposed.
+//
+// Luma:
+//
+// +-------+-------+-------+-------+ H0 Edge
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+ H1 Edge
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+ H2 Edge
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+ H3 Edge
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+//
+/////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xBBBB:w
+#endif
+
+
+//========== Luma deblocking ==========
+
+
+//---------- Deblock Y external top edge (H0) ----------
+
+ // Bypass deblocking if it is the top edge of the picture.
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterTopMbEdgeFlag:w // Check for FilterTopMbEdgeFlag
+
+// and.z.f0.1 (1) null:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz0]:uw 0xFFFF:uw // MaskA = 0?
+
+ // Get (alpha >> 2) + 2
+ shr (1) alpha2:w r[ECM_AddrReg, bAlphaTop0_Y]:ub 2:w // alpha >> 2
+
+ // p3 = Prev MB Y row 0 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Prev MB Y row 1 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Prev MB Y row 2 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Prev MB Y row 3 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 0 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 1 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 2 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 3 = r[Q_AddrReg, 48]<16;16,1>
+// mov (1) P_AddrReg:w PREV_MB_Y_BASE:w { NoDDClr }
+ mov (1) P_AddrReg:w TOP_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_Y_BASE:w { NoDDChk }
+
+ // Get horizontal border edge control data
+ // alpha = bAlphaTop0_Y
+ // beta = bBetaTop0_Y
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaTop0_Y]<2;2,1>:ub { NoDDClr } // 2 channels for alpha and beta
+
+ mov (2) MaskA<1>:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtTopHorz0]<2;2,1>:uw { NoDDClr, NoDDChk }
+
+ // tc0 has bTc0_h03_0_Y | bTc0_h02_0_Y | bTc0_h01_0_Y | bTc0_h00_0_Y
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h00_0_Y]<4;4,1>:ub { NoDDChk }
+
+// (f0.0) jmpi BYPASS_EXT_TOP_EDGE_Y
+// (f0.0.anyv) jmpi BYPASS_EXT_TOP_EDGE_Y
+
+ add (1) alpha2:w alpha2:w 2:w // alpha2 = (alpha >> 2) + 2
+
+// CALL(FILTER_Y, 1)
+ PRED_CALL(-f0.0, FILTER_Y, 1)
+
+//BYPASS_EXT_TOP_EDGE_Y:
+//------------------------------------------------------------------
+ // Same alpha, alpha2, beta and MaskB for all internal edges
+
+ // Get (alpha >> 2) + 2
+ shr (1) alpha2:w r[ECM_AddrReg, bAlphaInternal_Y]:ub 2:w // alpha >> 2
+
+ // alpha = bAlphaInternal_Y
+ // beta = bBetaInternal_Y
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaInternal_Y]<2;2,1>:ub { NoDDClr }
+
+ // Set MaskB = 0 for all 3 int edges, so it always uses bS < 4 algorithm.
+ mov (1) MaskB:uw 0:w { NoDDChk }
+
+ add (1) alpha2:w alpha2:w 2:w // alpha2 = (alpha >> 2) + 2
+
+
+//---------- Deblock Y internal top edge (H1) ----------
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+
+ // p3 = Cur MB Y row 0 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 1 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 2 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 3 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 4 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 5 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 6 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 7 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 4*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntTopHorz]:uw { NoDDClr }
+
+ // tc0 has bTc0_h13_Y + bTc0_h12_Y + bTc0_h11_Y + bTc0_h10_Y
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h10_Y]<4;4,1>:ub { NoDDChk }
+
+// (f0.0) jmpi BYPASS_4x4_DEBLOCK_H
+// (f0.0.anyv) jmpi BYPASS_4x4_DEBLOCK_H
+
+// CALL(FILTER_Y, 1)
+ PRED_CALL(-f0.0, FILTER_Y, 1)
+
+//BYPASS_4x4_DEBLOCK_H:
+//------------------------------------------------------------------
+
+
+//---------- Deblock Y internal mid horizontal edge (H2) ----------
+
+ // Bypass deblocking if FilterInternal8x8EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal8x8EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+
+// and.z.f0.1 (1) null:uw r[ECM_AddrReg, wEdgeCntlMap_IntMidHorz]:uw 0xFFFF:uw // MaskA = 0?
+
+ // p3 = Cur MB Y row 4 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 5 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 6 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 7 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 8 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 9 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 10 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 11 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w 4*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 8*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntMidHorz]:uw { NoDDClr }
+// mov (1) MaskB:uw 0:w // Set MaskB = 0, so it always uses bS < 4 algorithm.
+
+ // tc0 has bTc0_h23_Y + bTc0_h22_Y + bTc0_h21_Y + bTc0_h20_Y
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h20_Y]<4;4,1>:ub { NoDDChk }
+
+// (f0.0) jmpi BYPASS_8x8_DEBLOCK_H
+// (f0.0.anyv) jmpi BYPASS_8x8_DEBLOCK_H
+
+// CALL(FILTER_Y, 1)
+ PRED_CALL(-f0.0, FILTER_Y, 1)
+
+//BYPASS_8x8_DEBLOCK_H:
+//-----------------------------------------------
+
+
+//---------- Deblock Y internal bottom edge (H3) ----------
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+
+// and.z.f0.1 (1) null:uw r[ECM_AddrReg, wEdgeCntlMap_IntBotHorz]:uw 0xFFFF:uw // MaskA = 0?
+
+ // p3 = Cur MB Y row 8 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 9 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 10 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 11 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 12 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 13 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 14 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 15 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w 8*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 12*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntBotHorz]:uw { NoDDClr }
+// mov (1) MaskB:uw 0:w // Set MaskB = 0, so it always uses bS < 4 algorithm.
+
+ // tc0 has bTc0_h33_Y + bTc0_h32_Y + bTc0_h31_Y + bTc0_h30_Y
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_h30_Y]<4;4,1>:ub { NoDDChk }
+
+// (f0.0) jmpi BYPASS_4x4_DEBLOCK_H2
+// (f0.0.anyv) jmpi BYPASS_4x4_DEBLOCK_H2
+
+// CALL(FILTER_Y, 1)
+ PRED_CALL(-f0.0, FILTER_Y, 1)
+
+//BYPASS_4x4_DEBLOCK_H2:
+//-----------------------------------------------
diff --git a/shaders/h264/ildb/AVC_ILDB_Filter_Y_v.asm b/shaders/h264/ildb/AVC_ILDB_Filter_Y_v.asm new file mode 100644 index 0000000..9d6bf0a --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Filter_Y_v.asm @@ -0,0 +1,203 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+////////// AVC ILDB filter vertical Y ///////////////////////////////////////////////////////
+//
+// This filter code prepares the src data and control data for ILDB filtering on all vertical edges of Y.
+//
+// It sssumes the data for vertical de-blocking is already transposed.
+//
+// Luma:
+//
+// +-------+-------+-------+-------+
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+// | | | | |
+// | | | | |
+// | | | | |
+// +-------+-------+-------+-------+
+//
+// V0 V1 V2 V3
+// Edge Edge Edge Edge
+//
+/////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xBBBB:w
+#endif
+
+
+//========== Luma deblocking ==========
+
+
+//---------- Deblock Y external left edge (V0) ----------
+
+ // Bypass deblocking if it is left edge of the picture.
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterLeftMbEdgeFlag:w // Check for FilterLeftMbEdgeFlag
+
+// and.z.f0.1 (1) null:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtLeftVert0]:uw 0xFFFF:uw // MaskA = 0?
+
+ // Get (alpha >> 2) + 2
+ shr (1) alpha2:w r[ECM_AddrReg, bAlphaLeft0_Y]:ub 2:w // alpha >> 2
+
+ // p3 = Prev MB Y row 0 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Prev MB Y row 1 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Prev MB Y row 2 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Prev MB Y row 3 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 0 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 1 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 2 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 3 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w PREV_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w SRC_MB_Y_BASE:w { NoDDChk }
+
+ // Get vertical border edge control data
+ // alpha = bAlphaLeft0_Y
+ // beta = bBetaLeft0_Y
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaLeft0_Y]<2;2,1>:ub { NoDDClr } // 2 channels for alpha and beta
+
+ mov (2) MaskA<1>:uw r[ECM_AddrReg, wEdgeCntlMapA_ExtLeftVert0]<2;2,1>:uw { NoDDClr, NoDDChk }
+
+ // tc0 has bTc0_v30_0_Y | bTc0_v20_0_Y | bTc0_v10_0_Y | bTc0_v00_0_Y
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_v00_0_Y]<4;4,1>:ub { NoDDChk }
+
+// (f0.0) jmpi BYPASS_EXT_LEFT_EDGE_Y
+// (f0.0.anyv) jmpi BYPASS_EXT_LEFT_EDGE_Y
+
+ add (1) alpha2:w alpha2:w 2:w // alpha2 = (alpha >> 2) + 2
+
+// CALL(FILTER_Y, 1)
+ PRED_CALL(-f0.0, FILTER_Y, 1)
+
+
+//BYPASS_EXT_LEFT_EDGE_Y:
+//------------------------------------------------------------------
+ // Same alpha, alpha2, beta and MaskB for all internal edges
+
+ // Get (alpha >> 2) + 2
+ shr (1) alpha2:w r[ECM_AddrReg, bAlphaInternal_Y]:ub 2:w // alpha >> 2
+
+ // alpha = bAlphaInternal_Y
+ // beta = bBetaInternal_Y
+ mov (2) alpha<1>:w r[ECM_AddrReg, bAlphaInternal_Y]<2;2,1>:ub { NoDDClr }
+
+ // Set MaskB = 0 for all 3 int edges, so it always uses bS < 4 algorithm.
+ mov (1) MaskB:uw 0:w { NoDDChk }
+
+ add (1) alpha2:w alpha2:w 2:w // alpha2 = (alpha >> 2) + 2
+
+
+//---------- Deblock Y internal left edge (V1) ----------
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+
+// and.z.f0.1 (1) null:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]:uw 0xFFFF:uw // MaskA = 0?
+
+ // p3 = Cur MB Y row 0 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 1 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 2 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 3 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 4 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 5 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 6 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 7 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 4*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntLeftVert]:uw { NoDDClr }
+
+ // tc0 has bTc0_v31_Y + bTc0_v21_Y + bTc0_v11_Y + bTc0_v01_Y
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_v01_Y]<4;4,1>:ub { NoDDChk }
+
+// (f0.0) jmpi BYPASS_4x4_DEBLOCK_V
+// (f0.0.anyv) jmpi BYPASS_4x4_DEBLOCK_V
+
+// CALL(FILTER_Y, 1)
+ PRED_CALL(-f0.0, FILTER_Y, 1)
+
+//BYPASS_4x4_DEBLOCK_V:
+//------------------------------------------------------------------
+
+
+//---------- Deblock Y internal mid vert edge (V2) ----------
+
+ // Bypass deblocking if FilterInternal8x8EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal8x8EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+
+// and.z.f0.1 (1) null:uw r[ECM_AddrReg, wEdgeCntlMap_IntMidVert]:uw 0xFFFF:uw // MaskA = 0?
+
+ // p3 = Cur MB Y row 4 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 5 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 6 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 7 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 8 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 9 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 10 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 11 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w 4*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 8*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntMidVert]:uw { NoDDClr }
+// mov (1) MaskB:uw 0:w // Set MaskB = 0, so it always uses bS < 4 algorithm.
+
+ // tc0 has bTc0_v32_Y + bTc0_v22_Y + bTc0_v12_Y + bTc0_v02_Y
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_v02_Y]<4;4,1>:ub { NoDDChk }
+
+// (f0.0) jmpi BYPASS_8x8_DEBLOCK_V
+// (f0.0.anyv) jmpi BYPASS_8x8_DEBLOCK_V
+
+// CALL(FILTER_Y, 1)
+ PRED_CALL(-f0.0, FILTER_Y, 1)
+
+//BYPASS_8x8_DEBLOCK_V:
+//-----------------------------------------------
+
+
+//---------- Deblock Y interal right edge (V3) ----------
+
+ // Bypass deblocking if FilterInternal4x4EdgesFlag = 0
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FilterInternal4x4EdgesFlag:w // Check for FilterInternal4x4EdgesFlag
+
+// and.z.f0.1 (1) null:uw r[ECM_AddrReg, wEdgeCntlMap_IntRightVert]:uw 0xFFFF:uw // MaskA = 0?
+
+ // p3 = Cur MB Y row 8 = r[P_AddrReg, 0]<16;16,1>
+ // p2 = Cur MB Y row 9 = r[P_AddrReg, 16]<16;16,1>
+ // p1 = Cur MB Y row 10 = r[P_AddrReg, 32]<16;16,1>
+ // p0 = Cur MB Y row 11 = r[P_AddrReg, 48]<16;16,1>
+ // q0 = Cur MB Y row 12 = r[Q_AddrReg, 0]<16;16,1>
+ // q1 = Cur MB Y row 13 = r[Q_AddrReg, 16]<16;16,1>
+ // q2 = Cur MB Y row 14 = r[Q_AddrReg, 32]<16;16,1>
+ // q3 = Cur MB Y row 15 = r[Q_AddrReg, 48]<16;16,1>
+ mov (1) P_AddrReg:w 8*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDClr }
+ mov (1) Q_AddrReg:w 12*Y_ROW_WIDTH+SRC_MB_Y_BASE:w { NoDDChk }
+
+ mov (1) MaskA:uw r[ECM_AddrReg, wEdgeCntlMap_IntRightVert]:uw { NoDDClr }
+// mov (1) MaskB:uw 0:w // Set MaskB = 0, so it always uses bS < 4 algorithm.
+
+ // tc0 has bTc0_v33_Y + bTc0_v23_Y + bTc0_v13_Y + bTc0_v03_Y
+ mov (4) tc0<1>:ub r[ECM_AddrReg, bTc0_v03_Y]<4;4,1>:ub { NoDDChk }
+
+// (f0.0) jmpi BYPASS_4x4_DEBLOCK_V2
+// (f0.0.anyv) jmpi BYPASS_4x4_DEBLOCK_V2
+
+// CALL(FILTER_Y, 1)
+ PRED_CALL(-f0.0, FILTER_Y, 1)
+
+//BYPASS_4x4_DEBLOCK_V2:
+//-----------------------------------------------
diff --git a/shaders/h264/ildb/AVC_ILDB_ForwardMsg.asm b/shaders/h264/ildb/AVC_ILDB_ForwardMsg.asm new file mode 100644 index 0000000..96fe828 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_ForwardMsg.asm @@ -0,0 +1,57 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//========== Forward message to root thread through gateway ==========
+// Each child thread write a byte into the root GRF r50 defiend in open Gataway.
+
+#if defined(_DEBUG)
+mov (1) EntrySignatureC:w 0x7777:w
+#endif
+
+// Init payload to r0
+mov (8) GatewayPayload<1>:ud 0:w //{ NoDDClr }
+
+// Forward a message:
+// Offset = x relative to r50 (defiend in open gataway), x = ORIX >> 4 [bit 28:16]
+// Need to shift left 16
+
+// shift 2 more bits for byte to word offset
+
+//shl (1) Offset_Length:ud GateWayOffsetC:w 16:w { NoDDClr, NoDDChk }
+shl (1) Offset_Length:ud GateWayOffsetC:w 18:w
+
+// 2 bytes offset
+add (1) Offset_Length:ud Offset_Length:ud 0x00020000:d { NoDDClr }
+
+// Length = 1 byte, [bit 10:8 = 000]
+//000 xxxxxxxxxxxxx 00000 000 00000000 ==> 000x xxxx xxxx xxxx 0000 0000 0000 0000
+
+//mov (1) DispatchID:ub r0.20:ub // Dispatch ID
+
+//Move in EUid and Thread ID that we received from the PARENT thread
+mov (1) EUID_TID:uw r0.6:uw { NoDDClr, NoDDChk }
+
+mov (1) GatewayPayloadKey:uw 0x1212:uw { NoDDClr, NoDDChk } // Key
+
+//mov (4) GatewayPayload<1>:ud 0:ud { NoDDClr, NoDDChk } // Init payload low 4 dword
+
+// Write back one byte (value = 0xFF) to root thread GRF to indicate this child thread is finished
+// All lower 4 bytes must be assigned to the same byte value.
+mov (4) GatewayPayload<1>:ub 0xFFFF:uw { NoDDChk }
+
+// msg descriptor bit 15 set to '1' for notification
+#ifdef GW_DCN
+// For ILK, EOT bit should also be set to terminate the thread. This is to fix a timing related HW issue.
+//
+send (8) null:ud m0 GatewayPayload<8;8,1>:ud MSG_GW_EOT FWDMSGDSC+NOTIFYMSG
+#else
+send (8) null:ud m0 GatewayPayload<8;8,1>:ud MSG_GW FWDMSGDSC+NOTIFYMSG
+#endif // GW_DCN
+
+//========== Forward Msg Done ========================================
+
diff --git a/shaders/h264/ildb/AVC_ILDB_LumaThrdLimit.asm b/shaders/h264/ildb/AVC_ILDB_LumaThrdLimit.asm new file mode 100644 index 0000000..714ee4d --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_LumaThrdLimit.asm @@ -0,0 +1,46 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//========== Forward message to root thread through gateway ==========
+
+// Chroma root kenrel updates luma thread limit.
+
+#if defined(_DEBUG)
+mov (1) EntrySignatureC:w 0x7788:w
+#endif
+
+// Init payload to r0
+mov (8) GatewayPayload<1>:ud 0:w { NoDDClr }
+
+// Forward a message:
+// Offset = x relative to r50 (defiend in open gataway), x = ORIX >> 4 [bit 28:16]
+// Need to shift left 16
+
+mov (1) Offset_Length:ud THREAD_LIMIT_OFFSET:ud { NoDDClr, NoDDChk }
+
+// Length = 1 byte, [bit 10:8 = 000]
+//000 xxxxxxxxxxxxx 00000 000 00000000 ==> 000x xxxx xxxx xxxx 0000 0000 0000 0000
+
+//mov (1) DispatchID:ub r0.20:ub // Dispatch ID
+
+// Copy EUid and Thread ID that we received from the PARENT thread
+mov (1) EUID_TID:uw r0.6:uw { NoDDClr, NoDDChk }
+
+mov (1) GatewayPayloadKey:uw 0x1212:uw { NoDDChk } // Key
+
+//mov (4) GatewayPayload<1>:ud 0:ud { NoDDClr, NoDDChk } // Init payload low 4 dword
+
+// Write back one byte (value = 0xFF) to root thread GRF to indicate this child thread is finished
+// All lower 4 bytes must be assigned to the same byte value.
+add (1) Temp1_W:w MaxThreads:uw -OutstandingThreads:uw
+mov (4) GatewayPayload<1>:ub Temp1_B:ub
+
+send (8) GatewayResponse:ud m0 GatewayPayload<8;8,1>:ud MSG_GW FWDMSGDSC
+
+//========== Forward Msg Done ========================================
+
diff --git a/shaders/h264/ildb/AVC_ILDB_Luma_Core.asm b/shaders/h264/ildb/AVC_ILDB_Luma_Core.asm new file mode 100644 index 0000000..edc8273 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Luma_Core.asm @@ -0,0 +1,419 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__AVC_ILDB_LUMA_CORE__) // Make sure this file is only included once
+#define __AVC_ILDB_LUMA_CORE__
+
+////////// AVC ILDB Luma Core /////////////////////////////////////////////////////////////////////////////////
+//
+// This core performs AVC LUMA ILDB filtering on one horizontal edge (16 pixels) of a MB.
+// If data is transposed, it can also de-block a vertical edge.
+//
+// Bafore calling this subroutine, caller needs to set the following parameters.
+//
+// - EdgeCntlMap1 // Edge control map A
+// - EdgeCntlMap2 // Edge control map B
+// - P_AddrReg // Src and dest address register for P pixels
+// - Q_AddrReg // Src and dest address register for Q pixels
+// - alpha // alpha corresponding to the edge to be filtered
+// - beta // beta corresponding to the edge to be filtered
+// - tc0 // tc0 corresponding to the edge to be filtered
+//
+//
+// +----+----+----+----+----+----+----+----+
+// | p3 | p2 | P1 | p0 | q0 | q1 | q2 | q3 |
+// +----+----+----+----+----+----+----+----+
+//
+// p3 = r[P_AddrReg, 0]<16;16,1>
+// p2 = r[P_AddrReg, 16]<16;16,1>
+// p1 = r[P_AddrReg, 32]<16;16,1>
+// p0 = r[P_AddrReg, 48]<16;16,1>
+// q0 = r[Q_AddrReg, 0]<16;16,1>
+// q1 = r[Q_AddrReg, 16]<16;16,1>
+// q2 = r[Q_AddrReg, 32]<16;16,1>
+// q3 = r[Q_AddrReg, 48]<16;16,1>
+//
+/////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+// The region is both src and dest
+// P0-P3 and Q0-Q3 should be only used if they have not been modified to new values
+#undef P3
+#undef P2
+#undef P1
+#undef P0
+#undef Q0
+#undef Q1
+#undef Q2
+#undef Q3
+
+#define P3 r[P_AddrReg, 0]<16;16,1>:ub
+#define P2 r[P_AddrReg, 16]<16;16,1>:ub
+#define P1 r[P_AddrReg, 32]<16;16,1>:ub
+#define P0 r[P_AddrReg, 48]<16;16,1>:ub
+#define Q0 r[Q_AddrReg, 0]<16;16,1>:ub
+#define Q1 r[Q_AddrReg, 16]<16;16,1>:ub
+#define Q2 r[Q_AddrReg, 32]<16;16,1>:ub
+#define Q3 r[Q_AddrReg, 48]<16;16,1>:ub
+
+// New region as dest
+#undef NewP2
+#undef NewP1
+#undef NewP0
+#undef NewQ0
+#undef NewQ1
+#undef NewQ2
+
+#define NewP2 r[P_AddrReg, 16]<1>:ub
+#define NewP1 r[P_AddrReg, 32]<1>:ub
+#define NewP0 r[P_AddrReg, 48]<1>:ub
+#define NewQ0 r[Q_AddrReg, 0]<1>:ub
+#define NewQ1 r[Q_AddrReg, 16]<1>:ub
+#define NewQ2 r[Q_AddrReg, 32]<1>:ub
+
+// Filter one luma edge
+FILTER_Y:
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0x1111:w
+#endif
+ //---------- Derive filterSampleflag in AVC spec, equition (8-469) ----------
+ // bS is in MaskA
+
+ // Src copy of the p3, p2, p1, p0, q0, q1, q2, q3
+// mov (16) p0123_W(0)<1> r[P_AddrReg]<16;16,1>:uw
+// mov (16) p0123_W(1)<1> r[P_AddrReg, 32]<16;16,1>:uw
+// mov (16) q0123_W(0)<1> r[Q_AddrReg]<16;16,1>:uw
+// mov (16) q0123_W(1)<1> r[Q_AddrReg, 32]<16;16,1>:uw
+
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ add (16) q0_p0(0)<1> Q0 -P0 // q0-p0
+ add (16) TempRow0(0)<1> P1 -P0 // p1-p0
+ add (16) TempRow1(0)<1> Q1 -Q0 // q1-q0
+
+ // Build FilterSampleFlag
+ // abs(q0-p0) < alpha
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)q0_p0(0) alpha:w
+ // abs(p1-p0) < Beta
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)TempRow0(0) beta:w
+ // abs(q1-q0) < Beta
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)TempRow1(0) beta:w
+
+ //-----------------------------------------------------------------------------------------
+
+ (f0.0) if (16) Y_ENDIF1
+ // For channels whose edge control map1 = 1 ---> perform de-blocking
+
+// mov (1) f0.1:uw MaskB:uw {NoMask} // Now check for which algorithm to apply
+
+ // (abs)ap = |p2-p0|
+ add (16) ap(0)<1> P2 -P0 // ap = p2-p0
+ // (abs)aq = |q2-q0|
+ add (16) aq(0)<1> Q2 -Q0 // aq = q2-q0
+
+ // Make a copy of unmodified p0 and p1 for use in q0'and q1' calculation
+ mov (16) p0123_W(1)<1> r[P_AddrReg, 32]<16;16,1>:uw {NoMask}
+
+ (f0.1) if (16) Y_ELSE2
+
+ // For channels whose edge control map2 = 1 ---> bS = 4 algorithm
+
+ // Compute q0', q1' and q2'
+ //-----------------------------------------------------------------------------
+ // bS = 4 Algorithm :
+ //
+ // gama = |p0-q0| < ((alpha >> 2) + 2)
+ // deltap = (ap<beta) && gama; // deep filter flag
+ // if (deltap) {
+ // p0' = ( p2 +2*p1 +2*p0 +2*q0 + q1 + 4) >> 3;
+ // p1' = ( p2 + p1 + p0 + q0 + 2) >> 2;
+ // p2' = (2*p3 +3*p2 + p1 + p0 + q0 + 4) >> 3;
+ // } else {
+ // p0' = ( 2*p1 + p0 + q1 + 2) >> 2;
+ // }
+ //-----------------------------------------------------------------------------
+
+ // gama = |p0-q0| < ((alpha >> 2) + 2) = |p0-q0| < alpha2
+ cmp.l.f0.1 (16) null:w (abs)q0_p0(0) alpha2:w
+
+ // Common P01 = p0 + p1
+ add (16) P0_plus_P1(0)<1> P0 P1
+
+ // Common Q01 = q0 + q1
+ add (16) Q0_plus_Q1(0)<1> Q0 Q1
+
+// mov (1) CTemp1_W:w f0.1:uw {NoMask}
+ mov (1) f0.0:uw f0.1:uw {NoMask}
+
+ // deltap = ((abs)ap < beta) && gama
+ (f0.1) cmp.l.f0.1 (16) null:w (abs)ap(0) beta<0;1,0>:w // (abs)ap < beta ?
+
+ // deltaq = ((abs)aq < beta) && gama
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)aq(0) beta<0;1,0>:w // (abs)aq < beta ?
+
+
+// mov (1) CTemp1_W:w f0.0:uw {NoMask} // gama = |p0-q0| < ((alpha >> 2) + 2) for each channel
+// and (1) f0.1:w f0.1:uw CTemp1_W:w {NoMask} // deltap = (ap<beta) && gama
+
+
+ (f0.1) if (16) Y_ELSE3 // for channels its deltap = true
+
+ add (16) P2_plus_P3(0)<1> P2 P3
+
+ // A = (p1 + p0) + q0 = P01 + q0
+ add (16) A(0)<1> P0_plus_P1(0) Q0 // A = P01 + q0
+
+ // Now acc0 = A
+
+ // B = p2 + (p1 + p0 + q0) + 4 = p2 + A + 4
+// add (16) acc0.0<1>:w P2 4:w // p2 + 4
+// add (16) B(0)<1> acc0.0<16;16,1>:w A(0) // B = p2 + A + 4
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w 4:w // p2 + 4
+ add (16) B(0)<1> acc0.0<16;16,1>:w P2 // B = p2 + A + 4
+
+ // Now acc0 = B
+
+ // p2' = (2*p3 +3*p2 + A + 4) >> 3 = (2*(p3+p2) + B) >> 3
+// mov (16) acc0.0<1>:w B(0)
+ mac (16) acc0.0<1>:w P2_plus_P3(0) 2:w
+ shr.sat (16) TempRow3B(0)<2> acc0.0<16;16,1>:w 3:w
+
+ // p1' = (p2 + A + 2) >> 2 = (B - 2) >> 2
+ add (16) acc0.0<1>:w B(0) -2:w
+ shr.sat (16) TempRow1B(0)<2> acc0.0<16;16,1>:w 2:w
+
+ // p0' = (p2 +2*A + q1 + 4) >> 3 = (B + A + q1) >> 3
+ add (16) acc0.0<1>:w Q1 A(0) // B + A
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w B(0) // B + A + q1
+ shr.sat (16) TempRow0B(0)<2> acc0.0<16;16,1>:w 3:w // (B + A + q1) >> 3
+
+ // p2' = (2*p3 +3*p2 + A + 4) >> 3 = (2*(p3+p2) + B) >> 3
+// mov (16) acc0.0<1>:w B(0)
+// mac (16) acc0.0<1>:w P2_plus_P3(0) 2:w
+// shr.sat (16) TempRow3B(0)<2> acc0.0<16;16,1>:w 3:w
+
+ mov (16) NewP2 TempRow3B(0) // p2'
+ mov (16) NewP1 TempRow1B(0) // p1'
+ mov (16) NewP0 TempRow0B(0) // p0'
+
+Y_ELSE3:
+ else (16) Y_ENDIF3 // for channels its deltap = false
+
+ // p0' = (2*p1 + p0 + q1 + 2) >> 2 = (p1 + P01 + q1 + 2) >> 2
+ add (16) acc0.0<1>:w P1 P0_plus_P1(0) // p1 + P01 (TempRow1(0) = P01)
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w Q1
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w 2:w // p1 + P01 + q1 + 2
+
+ shr.sat (16) TempRow0B(0)<2> acc0.0<16;16,1>:w 2:w // >> 2
+ mov (16) NewP0 TempRow0B(0) // p0'
+
+ endif
+Y_ENDIF3:
+ // Compute q0', q1' and q2'
+ //-----------------------------------------------------------------------------
+ // bS = 4 Algorithm (cont):
+ //
+ // deltaq = (aq<beta) && gama; // deep filter flag
+ // if (deltaq) {
+ // q0' = ( q2 +2*q1 +2*q0 +2*p0 + p1 + 4) >> 3;
+ // q1' = ( q2 + q1 + q0 + p0 + 2) >> 2;
+ // q2' = (2*q3 +3*q2 + q1 + q0 + p0 + 4) >> 3;
+ // } else {
+ // q0' = ( 2*q1 + q0 + p1 + 2) >> 2;
+ // }
+
+ // deltaq = ((abs)aq < beta) && gama
+// cmp.l.f0.1 (16) null:w (abs)aq(0) beta<0;1,0>:w // (abs)aq < beta ?
+
+ // Common Q01 = q0 + q1
+// add (16) Q0_plus_Q1(0)<1> Q0 Q1
+
+// and (1) f0.1:w f0.1:uw CTemp1_W:w {NoMask} // deltaq = ((abs)ap < beta) && gama
+
+ (f0.0) if (16) Y_ELSE4 // for channels its deltaq = true
+
+ add (16) Q2_plus_Q3(0)<1> Q2 Q3
+
+ // A = (q1 + q0) + p0 = Q01 + p0
+ add (16) A(0)<1> Q0_plus_Q1(0) p0(0) // A = q1+q0 + p0
+
+ // Acc0 = A
+
+ // B = q2 + q1 + q0 + p0 + 4 = q2 + A + 4
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w 4:w // q2 + 4
+ add (16) B(0)<1> acc0.0<16;16,1>:w Q2 // B = q2 + A + 4
+
+ // Acc0 = B
+
+ // q2' = (2*q3 +3*q2 + A + 4) >> 3 = (2*(q3+q2) + B) >> 3
+// mov (16) acc0.0<1>:w B(0)
+ mac (16) acc0.0<1>:w Q2_plus_Q3(0) 2:w
+ shr.sat (16) TempRow3B(0)<2> acc0.0<16;16,1>:w 3:w
+
+ // q1' = (q2 + A + 2) >> 2 = (B - 2) >> 2
+ add (16) acc0.0<1>:w B(0) -2:w
+ shr.sat (16) TempRow1B(0)<2> acc0.0<16;16,1>:w 2:w
+
+ // q0' = (q2 +2*A + p1 + 4) >> 3 = (B + A + p1) >> 3
+ add (16) acc0.0<1>:w p1(0) A(0)
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w B(0)
+ shr.sat (16) TempRow0B(0)<2> acc0.0<16;16,1>:w 3:w
+
+ mov (16) NewQ2 TempRow3B(0) // q2'
+ mov (16) NewQ1 TempRow1B(0) // q1'
+ mov (16) NewQ0 TempRow0B(0) // q0'
+
+Y_ELSE4:
+ else (16) Y_ENDIF4 // for channels its deltaq = false
+
+ // q0' = (2*q1 + q0 + p1 + 2) >> 2 = (q1 + Q01 + p1 + 2) >> 2
+ // Use original p1 values in p1(0)
+ add (16) acc0.0<1>:w p1(0) Q0_plus_Q1(0) // p1 + P01 (TempRow1(0) = P01)
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w Q1
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w 2:w // p1 + P01 + q1 + 2
+
+ shr.sat (16) TempRow0B(0)<2> acc0.0<16;16,1>:w 2:w // >> 2
+ mov (16) NewQ0 TempRow0B(0) // q0'
+
+ endif
+Y_ENDIF4:
+
+
+ // Done with bS = 4 algorithm
+
+Y_ELSE2:
+ else (16) Y_ENDIF2
+ // For channels whose edge control map2 = 0 ---> bS < 4 algorithm
+
+ //-----------------------------------------------------------------------------
+ // bS < 4 Algorithm :
+ // tc = tc0 + (|p2-p0|<Beta ? 1 : 0) + (|q2-q0|<Beta ? 1 : 0)
+ // delta = Clip3(-tc, tc, ((((q0-p0)<<2) + (p1-q1) + 4) >> 3))
+ // p0' = Clip1(p0 + delta) = Clip3(0, 0xFF, p0 + delta)
+ // q0' = Clip1(q0 - delta) = Clip3(0, 0xFF, q0 - delta)
+ // if (|p2-p0|<Beta)
+ // p1' = p1 + Clip3(-tc0, tc0, (p2 + ((p0+q0+1)>>1) - (p1<<1)) >> 1 )
+ // if (|q2-q0|<Beta)
+ // q1' = q1 + Clip3(-tc0, tc0, (q2 + ((p0+q0+1)>>1) - (q1<<1)) >> 1 )
+ //-----------------------------------------------------------------------------
+
+ // Expand tc0
+ mov (16) tc_exp(0)<1> tc0<1;4,0>:ub {NoMask}
+ mov (16) tc0_exp(0)<1> tc0<1;4,0>:ub {NoMask} // tc0_exp = tc0, each tc0 is duplicated 4 times for 4 adjcent 4 pixels
+
+ // tc_exp = tc0_exp + (|p2-p0|<Beta ? 1 : 0) + (|q2-q0|<Beta ? 1 : 0)
+// mov (16) tc_exp(0)<1> tc0_exp(0) // tc = tc0_exp first
+
+
+ cmp.l.f0.0 (16) null:w (abs)ap(0) beta:w // |p2-p0|< Beta ? ---> (abs)ap < Beta ?
+ cmp.l.f0.1 (16) null:w (abs)aq(0) beta:w // |q2-q0|< Beta ? ---> (abs)aq < Beta ?
+
+ //--- Use free cycles here ---
+ // delta = Clip3(-tc, tc, ((((q0-p0)<<2) + (p1-q1) + 4) >> 3))
+ // 4 * (q0-p0) + p1 - q1 + 4
+ add (16) acc0<1>:w P1 4:w // p1 + 4
+ mac (16) acc0<1>:w q0_p0(0) 4:w // 4 * (q0-p0) + p1 + 4
+ add (16) acc0<1>:w acc0<16;16,1>:w -Q1 // 4 * (q0-p0) + p1 - q1 + 4
+ shr (16) TempRow0(0)<1> acc0<16;16,1>:w 3:w
+
+ // Continue on getting tc_exp
+ (f0.0) add (16) tc_exp(0)<1> tc_exp(0) 1:w // tc0_exp + (|p2-p0|<Beta ? 1 : 0)
+ mov (2) CTemp1_W<1>:w f0.0<2;2,1>:w {NoMask} // Save |p2-p0|<Beta flag
+ (f0.1) add (16) tc_exp(0)<1> tc_exp(0) 1:w // tc_exp = tc0_exp + (|p2-p0|<Beta ? 1 : 0) + (|q2-q0|<Beta ? 1 : 0)
+
+
+ // Continue on cliping tc to get delta
+ cmp.g.f0.0 (16) null:w TempRow0(0) tc_exp(0) // Clip if delta' > tc
+ cmp.l.f0.1 (16) null:w TempRow0(0) -tc_exp(0) // Clip if delta' < -tc
+
+ //--- Use free cycles here ---
+ // common = (p0+q0+1) >> 1 ---> TempRow2(0)
+ // Same as avg of p0 and q0
+ avg (16) TempRow2(0)<1> P0 Q0
+
+ // Continue on cliping tc to get delta
+ (f0.0) mov (16) TempRow0(0)<1> tc_exp(0)
+ (f0.1) mov (16) TempRow0(0)<1> -tc_exp(0)
+
+ //--- Use free cycles here ---
+ mov (2) f0.0<1>:w CTemp1_W<2;2,1>:w {NoMask} // CTemp1_W = (|p2-p0|<Beta)
+ // CTemp2_W = (|q2-q0|<Beta)
+ //-----------------------------------------------------------------------
+
+ // p0' = Clip1(p0 + delta) = Clip3(0, 0xFF, p0 + delta)
+ // q0' = Clip1(q0 - delta) = Clip3(0, 0xFF, q0 - delta)
+ add.sat (16) TempRow1B(0)<2> P0 TempRow0(0) // p0+delta
+ add.sat (16) TempRow0B(0)<2> Q0 -TempRow0(0) // q0-delta
+ mov (16) NewP0 TempRow1B(0) // p0'
+ mov (16) NewQ0 TempRow0B(0) // q0'
+ //-----------------------------------------------------------------------
+
+ // Now compute p1' and q1'
+
+ // if (|p2-p0|<Beta)
+// mov (1) f0.0:w CTemp1_W:w {NoMask} // CTemp1_W = (|p2-p0|<Beta)
+ (f0.0) if (16) Y_ENDIF6
+
+ // p1' = p1 + Clip3(-tc0, tc0, adj)
+ // adj = (p2 + common - (p1<<1)) >> 1 = (p2 + common - (p1*2)) >> 1
+ add (16) acc0<1>:w P2 TempRow2(0) // TempRow2(0) = common = (p0+q0+1) >> 1
+ mac (16) acc0<1>:w P1 -2:w
+ shr (16) TempRow1(0)<1> acc0<16;16,1>:w 1:w
+
+ // tc clip to get tc_adj
+ cmp.g.f0.0 (16) null:w TempRow1(0) tc0_exp(0) // Clip if delta' > tc
+ cmp.l.f0.1 (16) null:w TempRow1(0) -tc0_exp(0) // Clip if delta' < -tc
+
+ (f0.0) mov (16) TempRow1(0)<1> tc0_exp(0)
+ (f0.1) mov (16) TempRow1(0)<1> -tc0_exp(0)
+
+ //--- Use free cycles here ---
+ mov (1) f0.1:w CTemp2_W:w {NoMask} // CTemp2_W = (|q2-q0|<Beta)
+
+ // p1' = p1 + tc_adj
+ add.sat (16) TempRow1B(0)<2> P1 TempRow1(0) // p1+tc_adj
+ mov (16) NewP1 TempRow1B(0) // p1'
+
+ //------------------------------------------------------------------------
+Y_ENDIF6:
+ endif
+
+ // if (|q2-q0|<Beta)
+// mov (1) f0.1:w CTemp2_W:w {NoMask} // CTemp2_W = (|q2-q0|<Beta)
+ (f0.1) if (16) Y_ENDIF7
+
+ // q1' = q1 + Clip3(-tc0, tc0, adj)
+ // adj = (q2 + common - (q1<<1)) >> 1
+ // same as q2 + common - (q1 * 2)
+ add (16) acc0<1>:w Q2 TempRow2(0)
+ mac (16) acc0<1>:w Q1 -2:w
+ shr (16) TempRow1(0)<1> acc0<16;16,1>:w 1:w
+
+ // tc clip to get tc_adj
+ cmp.g.f0.0 (16) null:w TempRow1(0) tc0_exp(0) // Clip if delta' > tc
+ cmp.l.f0.1 (16) null:w TempRow1(0) -tc0_exp(0) // Clip if delta' < -tc
+
+ (f0.0) mov (16) TempRow1(0)<1> tc0_exp(0)
+ (f0.1) mov (16) TempRow1(0)<1> -tc0_exp(0)
+
+ // q1' = q1 + tc_adj
+ add.sat (16) TempRow1B(0)<2> Q1 TempRow1(0) // q1+tc_adj
+ mov (16) NewQ1 TempRow1B(0) // q1'
+
+ //------------------------------------------------------------------------
+Y_ENDIF7:
+ endif
+
+ endif
+Y_ENDIF2:
+Y_ENDIF1:
+ endif
+
+RETURN
+
+#endif // !defined(__AVC_ILDB_LUMA_CORE__)
diff --git a/shaders/h264/ildb/AVC_ILDB_Luma_Core_Mbaff.asm b/shaders/h264/ildb/AVC_ILDB_Luma_Core_Mbaff.asm new file mode 100644 index 0000000..0c3f752 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Luma_Core_Mbaff.asm @@ -0,0 +1,391 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__AVC_ILDB_LUMA_CORE_MBAFF__) // Make sure this file is only included once
+#define __AVC_ILDB_LUMA_CORE_MBAFF__
+
+////////// AVC ILDB Luma Core Mbaff /////////////////////////////////////////////////////////////////////////////////
+//
+// This core performs AVC LUMA ILDB filtering on one horizontal edge (16 pixels) of a MB.
+// If data is transposed, it can also de-block a vertical edge.
+//
+// Bafore calling this subroutine, caller needs to set the following parameters.
+//
+// - EdgeCntlMap1 // Edge control map A
+// - EdgeCntlMap2 // Edge control map B
+// - P_AddrReg // Src and dest address register for P pixels
+// - Q_AddrReg // Src and dest address register for Q pixels
+// - alpha // alpha corresponding to the edge to be filtered
+// - beta // beta corresponding to the edge to be filtered
+// - tc0 // tc0 corresponding to the edge to be filtered
+//
+//
+// +----+----+----+----+----+----+----+----+
+// | p3 | p2 | P1 | p0 | q0 | q1 | q2 | q3 |
+// +----+----+----+----+----+----+----+----+
+//
+// p3 = r[P_AddrReg, 0]<16;16,1>
+// p2 = r[P_AddrReg, 16]<16;16,1>
+// p1 = r[P_AddrReg, 32]<16;16,1>
+// p0 = r[P_AddrReg, 48]<16;16,1>
+// q0 = r[Q_AddrReg, 0]<16;16,1>
+// q1 = r[Q_AddrReg, 16]<16;16,1>
+// q2 = r[Q_AddrReg, 32]<16;16,1>
+// q3 = r[Q_AddrReg, 48]<16;16,1>
+//
+/////////////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+// The region is both src and dest
+// P0-P3 and Q0-Q3 should be only used if they have not been modified to new values
+#undef P3
+#undef P2
+#undef P1
+#undef P0
+#undef Q0
+#undef Q1
+#undef Q2
+#undef Q3
+
+#define P3 r[P_AddrReg, 0]<16;16,1>:ub
+#define P2 r[P_AddrReg, 16]<16;16,1>:ub
+#define P1 r[P_AddrReg, 32]<16;16,1>:ub
+#define P0 r[P_AddrReg, 48]<16;16,1>:ub
+#define Q0 r[Q_AddrReg, 0]<16;16,1>:ub
+#define Q1 r[Q_AddrReg, 16]<16;16,1>:ub
+#define Q2 r[Q_AddrReg, 32]<16;16,1>:ub
+#define Q3 r[Q_AddrReg, 48]<16;16,1>:ub
+
+// New region as dest
+#undef NewP2
+#undef NewP1
+#undef NewP0
+#undef NewQ0
+#undef NewQ1
+#undef NewQ2
+
+#define NewP2 r[P_AddrReg, 16]<1>:ub
+#define NewP1 r[P_AddrReg, 32]<1>:ub
+#define NewP0 r[P_AddrReg, 48]<1>:ub
+#define NewQ0 r[Q_AddrReg, 0]<1>:ub
+#define NewQ1 r[Q_AddrReg, 16]<1>:ub
+#define NewQ2 r[Q_AddrReg, 32]<1>:ub
+
+
+
+// Filter one luma edge - mbaff
+FILTER_Y_MBAFF:
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0x1111:w
+#endif
+ //---------- Derive filterSampleflag in AVC spec, equition (8-469) ----------
+ // bS is in MaskA
+
+ // Src copy of the p3, p2, p1, p0, q0, q1, q2, q3
+// mov (16) p0123_W(0)<1> r[P_AddrReg]<16;16,1>:uw
+// mov (16) p0123_W(1)<1> r[P_AddrReg, 32]<16;16,1>:uw
+// mov (16) q0123_W(0)<1> r[Q_AddrReg]<16;16,1>:uw
+// mov (16) q0123_W(1)<1> r[Q_AddrReg, 32]<16;16,1>:uw
+
+ // Move MaskA and MaskB to flag regs
+ mov (2) f0.0<1>:uw MaskA<2;2,1>:uw
+
+ add (16) q0_p0(0)<1> Q0 -P0 // q0-p0
+ add (16) TempRow0(0)<1> P1 -P0 // p1-p0
+ add (16) TempRow1(0)<1> Q1 -Q0 // q1-q0
+
+ // abs(q0-p0) < alpha
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)q0_p0(0) Mbaff_ALPHA(0)
+ // abs(p1-p0) < Beta
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)TempRow0(0) Mbaff_BETA(0)
+ // abs(q1-q0) < Beta
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)TempRow1(0) Mbaff_BETA(0)
+
+ //-----------------------------------------------------------------------------------------
+
+ (f0.0) if (16) MBAFF_Y_ENDIF1
+ // For channels whose edge control map1 = 1 ---> perform de-blocking
+
+// mov (1) f0.1:uw MaskB:uw {NoMask} // Now check for which algorithm to apply
+
+ // (abs)ap = |p2-p0|
+ add (16) ap(0)<1> P2 -P0
+
+ // (abs)aq = |q2-q0|
+ add (16) aq(0)<1> Q2 -Q0
+
+ // Make a copy of unmodified p0 and p1 for use in q0'and q1' calculation
+ mov (16) p0123_W(1)<1> r[P_AddrReg, 32]<16;16,1>:uw {NoMask}
+
+ (f0.1) if (16) MBAFF_Y_ELSE2
+
+ // For channels whose edge control map2 = 1 ---> bS = 4 algorithm
+
+ // Compute q0', q1' and q2'
+ //-----------------------------------------------------------------------------
+ // bS = 4 Algorithm :
+ //
+ // gama = |p0-q0| < ((alpha >> 2) + 2)
+ // deltap = (ap<beta) && gama; // deep filter flag
+ // if (deltap) {
+ // p0' = ( p2 +2*p1 +2*p0 +2*q0 + q1 + 4) >> 3;
+ // p1' = ( p2 + p1 + p0 + q0 + 2) >> 2;
+ // p2' = (2*p3 +3*p2 + p1 + p0 + q0 + 4) >> 3;
+ // } else {
+ // p0' = ( 2*p1 + p0 + q1 + 2) >> 2;
+ // }
+ //-----------------------------------------------------------------------------
+
+ // gama = |p0-q0| < ((alpha >> 2) + 2) = |p0-q0| < alpha2
+ cmp.l.f0.1 (16) null:w (abs)q0_p0(0) Mbaff_ALPHA2(0)
+
+ // Common P01 = p0 + p1
+ add (16) P0_plus_P1(0)<1> P0 P1
+
+ // Common Q01 = q0 + q1
+ add (16) Q0_plus_Q1(0)<1> Q0 Q1
+
+ mov (1) f0.0:uw f0.1:uw {NoMask}
+
+ // deltap = ((abs)ap < beta) && gama
+ (f0.1) cmp.l.f0.1 (16) null:w (abs)ap(0) Mbaff_BETA(0) // (abs)ap < beta ?
+
+ // deltaq = ((abs)aq < beta) && gama
+ (f0.0) cmp.l.f0.0 (16) null:w (abs)aq(0) Mbaff_BETA(0) // (abs)aq < beta ?
+
+
+ (f0.1) if (16) MBAFF_Y_ELSE3 // for channels its deltap = true
+
+ add (16) P2_plus_P3(0)<1> P2 P3
+
+ // A = p1 + p0 + q0 = P01 + q0
+ add (16) A(0)<1> P0_plus_P1(0) Q0 // A = P01 + q0
+
+ // Now acc0 = A
+
+ // B = p2 + p1 + p0 + q0 + 4 = p2 + A + 4
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w 4:w // p2 + 4
+ add (16) B(0)<1> acc0.0<16;16,1>:w P2 // B = p2 + A + 4
+
+ // Now acc0 = B
+
+ // p2' = (2*p3 +3*p2 + A + 4) >> 3 = (2*(p3+p2) + B) >> 3
+ mac (16) acc0.0<1>:w P2_plus_P3(0) 2:w
+ shr.sat (16) TempRow3B(0)<2> acc0.0<16;16,1>:w 3:w
+
+ // p1' = (p2 + A + 2) >> 2 = (B - 2) >> 2
+ add (16) acc0.0<1>:w B(0) -2:w
+ shr.sat (16) TempRow1B(0)<2> acc0.0<16;16,1>:w 2:w
+
+ // p0' = (p2 +2*A + q1 + 4) >> 3 = (B + A + q1) >> 3
+ add (16) acc0.0<1>:w Q1 A(0) // B + A
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w B(0) // B + A + q1
+ shr.sat (16) TempRow0B(0)<2> acc0.0<16;16,1>:w 3:w // (B + A + q1) >> 3
+
+ mov (16) NewP2 TempRow3B(0) // p2'
+ mov (16) NewP1 TempRow1B(0) // p1'
+ mov (16) NewP0 TempRow0B(0) // p0'
+
+MBAFF_Y_ELSE3:
+ else (16) MBAFF_Y_ENDIF3 // for channels its deltap = false
+
+ // p0' = (2*p1 + p0 + q1 + 2) >> 2 = (p1 + P01 + q1 + 2) >> 2
+ add (16) acc0.0<1>:w P1 P0_plus_P1(0) // p1 + P01 (TempRow1(0) = P01)
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w Q1
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w 2:w // p1 + P01 + q1 + 2
+
+ shr.sat (16) TempRow0B(0)<2> acc0.0<16;16,1>:w 2:w // >> 2
+ mov (16) NewP0 TempRow0B(0) // p0'
+
+ endif
+
+MBAFF_Y_ENDIF3:
+ // Compute q0', q1' and q2'
+ //-----------------------------------------------------------------------------
+ // bS = 4 Algorithm (cont):
+ //
+ // deltaq = (aq<beta) && gama; // deep filter flag
+ // if (deltaq) {
+ // q0' = ( q2 +2*q1 +2*q0 +2*p0 + p1 + 4) >> 3;
+ // q1' = ( q2 + q1 + q0 + p0 + 2) >> 2;
+ // q2' = (2*q3 +3*q2 + q1 + q0 + p0 + 4) >> 3;
+ // } else {
+ // q0' = ( 2*q1 + q0 + p1 + 2) >> 2;
+ // }
+
+ (f0.0) if (16) MBAFF_Y_ELSE4 // for channels its deltaq = true
+
+ add (16) Q2_plus_Q3(0)<1> Q2 Q3
+
+ // A = q1 + q0 + p0 = Q01 + p0
+ add (16) A(0)<1> Q0_plus_Q1(0) p0(0) // A = q1+q0 + p0
+
+ // B = q2 + q1 + q0 + p0 + 4 = q2 + A + 4
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w 4:w // q2 + 4
+ add (16) B(0)<1> acc0.0<16;16,1>:w Q2 // B = q2 + A + 4
+
+ // Acc0 = B
+
+ // q2' = (2*q3 +3*q2 + A + 4) >> 3 = (2*(q3+q2) + B) >> 3
+ mac (16) acc0.0<1>:w Q2_plus_Q3(0) 2:w
+ shr.sat (16) TempRow3B(0)<2> acc0.0<16;16,1>:w 3:w
+
+ // q1' = (q2 + A + 2) >> 2 = (B - 2) >> 2
+ add (16) acc0.0<1>:w B(0) -2:w
+ shr.sat (16) TempRow1B(0)<2> acc0.0<16;16,1>:w 2:w
+
+ // q0' = (q2 +2*A + p1 + 4) >> 3 = (B + A + p1) >> 3
+ add (16) acc0.0<1>:w p1(0) A(0)
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w B(0)
+ shr.sat (16) TempRow0B(0)<2> acc0.0<16;16,1>:w 3:w
+
+ mov (16) NewQ2 TempRow3B(0) // q2'
+ mov (16) NewQ1 TempRow1B(0) // q1'
+ mov (16) NewQ0 TempRow0B(0) // q0'
+
+MBAFF_Y_ELSE4:
+ else (16) MBAFF_Y_ENDIF4 // for channels its deltaq = false
+
+ // q0' = (2*q1 + q0 + p1 + 2) >> 2 = (q1 + Q01 + p1 + 2) >> 2
+ // Use original p1 values in p1(0)
+ add (16) acc0.0<1>:w p1(0) Q0_plus_Q1(0) // p1 + P01 (TempRow1(0) = P01)
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w Q1
+ add (16) acc0.0<1>:w acc0.0<16;16,1>:w 2:w // p1 + P01 + q1 + 2
+
+ shr.sat (16) TempRow0B(0)<2> acc0.0<16;16,1>:w 2:w // >> 2
+ mov (16) NewQ0 TempRow0B(0) // q0'
+
+ endif
+MBAFF_Y_ENDIF4:
+
+
+ // Done with bS = 4 algorithm
+
+MBAFF_Y_ELSE2:
+ else (16) MBAFF_Y_ENDIF2
+ // For channels whose edge control map2 = 0 ---> bS < 4 algorithm
+
+ //-----------------------------------------------------------------------------
+ // bS < 4 Algorithm :
+ // tc = tc0 + (|p2-p0|<Beta ? 1 : 0) + (|q2-q0|<Beta ? 1 : 0)
+ // delta = Clip3(-tc, tc, ((((q0-p0)<<2) + (p1-q1) + 4) >> 3))
+ // p0' = Clip1(p0 + delta) = Clip3(0, 0xFF, p0 + delta)
+ // q0' = Clip1(q0 - delta) = Clip3(0, 0xFF, q0 - delta)
+ // if (|p2-p0|<Beta)
+ // p1' = p1 + Clip3(-tc0, tc0, (p2 + ((p0+q0+1)>>1) - (p1<<1)) >> 1 )
+ // if (|q2-q0|<Beta)
+ // q1' = q1 + Clip3(-tc0, tc0, (q2 + ((p0+q0+1)>>1) - (q1<<1)) >> 1 )
+ //-----------------------------------------------------------------------------
+
+ mov (16) tc_exp(0)<1> Mbaff_TC0(0) // tc = tc0_exp first
+
+ cmp.l.f0.0 (16) null:w (abs)ap(0) Mbaff_BETA(0) // |p2-p0|<Beta ?
+ cmp.l.f0.1 (16) null:w (abs)aq(0) Mbaff_BETA(0) // |q2-q0|<Beta ?
+
+ //--- Use free cycles here ---
+ // delta = Clip3(-tc, tc, ((((q0-p0)<<2) + (p1-q1) + 4) >> 3))
+ // 4 * (q0-p0) + p1 - q1 + 4
+ add (16) acc0<1>:w P1 4:w // p1 + 4
+ mac (16) acc0<1>:w q0_p0(0) 4:w // 4 * (q0-p0) + p1 + 4
+ add (16) acc0<1>:w acc0<16;16,1>:w -Q1 // 4 * (q0-p0) + p1 - q1 + 4
+ shr (16) TempRow0(0)<1> acc0<16;16,1>:w 3:w
+
+ // Continue on getting tc_exp
+ (f0.0) add (16) tc_exp(0)<1> tc_exp(0) 1:w // tc0_exp + (|p2-p0|<Beta ? 1 : 0)
+ mov (2) CTemp1_W<1>:w f0.0<2;2,1>:w {NoMask} // Save |p2-p0|<Beta flag
+ (f0.1) add (16) tc_exp(0)<1> tc_exp(0) 1:w // tc_exp = tc0_exp + (|p2-p0|<Beta ? 1 : 0) + (|q2-q0|<Beta ? 1 : 0)
+
+ // Continue on cliping tc to get delta
+ cmp.g.f0.0 (16) null:w TempRow0(0) tc_exp(0) // Clip if delta' > tc
+ cmp.l.f0.1 (16) null:w TempRow0(0) -tc_exp(0) // Clip if delta' < -tc
+
+ //--- Use free cycles here ---
+ // common = (p0+q0+1) >> 1 ---> TempRow2(0)
+ // Same as avg of p0 and q0
+ avg (16) TempRow2(0)<1> P0 Q0
+
+ // Continue on cliping tc to get delta
+ (f0.0) mov (16) TempRow0(0)<1> tc_exp(0)
+ (f0.1) mov (16) TempRow0(0)<1> -tc_exp(0)
+
+ //--- Use free cycles here ---
+ mov (2) f0.0<1>:w CTemp1_W<2;2,1>:w {NoMask} // CTemp1_W = (|p2-p0|<Beta)
+ // CTemp2_W = (|q2-q0|<Beta)
+
+ // p0' = Clip1(p0 + delta) = Clip3(0, 0xFF, p0 + delta)
+ // q0' = Clip1(q0 - delta) = Clip3(0, 0xFF, q0 - delta)
+ add.sat (16) TempRow1B(0)<2> P0 TempRow0(0) // p0+delta
+ add.sat (16) TempRow0B(0)<2> Q0 -TempRow0(0) // q0-delta
+
+ mov (16) NewP0 TempRow1B(0) // p0'
+ mov (16) NewQ0 TempRow0B(0) // q0'
+
+ //-----------------------------------------------------------------------
+
+ // Now compute p1' and q1'
+
+ // if (|p2-p0|<Beta)
+ (f0.0) if (16) MBAFF_Y_ENDIF6
+
+ // p1' = p1 + Clip3(-tc0, tc0, adj)
+ // adj = (p2 + common - (p1<<1)) >> 1 = (p2 + common - (p1*2)) >> 1
+ add (16) acc0<1>:w P2 TempRow2(0) // TempRow2(0) = common = (p0+q0+1) >> 1
+ mac (16) acc0<1>:w P1 -2:w
+ shr (16) TempRow1(0)<1> acc0<16;16,1>:w 1:w
+
+ // tc clip to get tc_adj
+ cmp.g.f0.0 (16) null:w TempRow1(0) Mbaff_TC0(0) // Clip if delta' > tc
+ cmp.l.f0.1 (16) null:w TempRow1(0) -Mbaff_TC0(0) // Clip if delta' < -tc
+
+ (f0.0) mov (16) TempRow1(0)<1> Mbaff_TC0(0)
+ (f0.1) mov (16) TempRow1(0)<1> -Mbaff_TC0(0)
+
+ //--- Use free cycles here ---
+ mov (1) f0.1:w CTemp2_W:w {NoMask} // CTemp2_W = (|q2-q0|<Beta)
+
+ // p1' = p1 + tc_adj
+ add.sat (16) TempRow1B(0)<2> P1 TempRow1(0) // p1+tc_adj
+ mov (16) NewP1 TempRow1B(0) // p1'
+ //------------------------------------------------------------------------
+
+MBAFF_Y_ENDIF6:
+ endif
+
+ // if (|q2-q0|<Beta)
+ (f0.1) if (16) MBAFF_Y_ENDIF7
+
+ // q1' = q1 + Clip3(-tc0, tc0, adj)
+ // adj = (q2 + common - (q1<<1)) >> 1
+ // same as q2 + common - (q1 * 2)
+ add (16) acc0<1>:w Q2 TempRow2(0)
+ mac (16) acc0<1>:w Q1 -2:w
+ shr (16) TempRow1(0)<1> acc0<16;16,1>:w 1:w
+
+ // tc clip to get tc_adj
+ cmp.g.f0.0 (16) null:w TempRow1(0) Mbaff_TC0(0) // Clip if delta' > tc
+ cmp.l.f0.1 (16) null:w TempRow1(0) -Mbaff_TC0(0) // Clip if delta' < -tc
+
+ (f0.0) mov (16) TempRow1(0)<1> Mbaff_TC0(0)
+ (f0.1) mov (16) TempRow1(0)<1> -Mbaff_TC0(0)
+
+ // q1' = q1 + tc_adj
+ add.sat (16) TempRow1B(0)<2> Q1 TempRow1(0) // q1+tc_adj
+ mov (16) NewQ1 TempRow1B(0) // q1'
+
+ //------------------------------------------------------------------------
+MBAFF_Y_ENDIF7:
+ endif
+
+ endif
+MBAFF_Y_ENDIF2:
+MBAFF_Y_ENDIF1:
+ endif
+
+RETURN
+
+#endif // !defined(__AVC_ILDB_LUMA_CORE_MBAFF__)
diff --git a/shaders/h264/ildb/AVC_ILDB_OpenGateway.asm b/shaders/h264/ildb/AVC_ILDB_OpenGateway.asm new file mode 100644 index 0000000..06d1933 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_OpenGateway.asm @@ -0,0 +1,43 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//----- Open a Message Gateway -----
+// The parent thread is the recipient thread
+
+#if defined(_DEBUG)
+ mov (1) EntrySignature:w 0x1111:w
+#endif
+
+mov (8) GatewayPayload<1>:ud r0.0<8;8,1>:ud // Init payload to r0
+
+// r50- (16 GRFs) are the GRFs child thread can wtite to.
+
+// Reg base is at bit 28:21, Gateway size is at [bit 10:8]
+// r6: 6 = 00000110
+//000 00000110 0000000000 100 00000000 ==> 0000 0000 1100 0000 0000 0100 0000 0000
+mov (1) RegBase_GatewaySize:ud 0x00C00400:ud // Reg base + Gateway size (16 GRFs)
+
+
+//000 00110010 0000000000 100 00000000 ==> 0000 0110 0100 0000 0000 0100 0000 0000
+//mov (1) RegBase_GatewaySize:ud 0x06400400:ud // Reg base (r50 = 0x640 byte offset) + Gateway size (16 GRFs)
+
+//mov (1) DispatchID:ub r0.20:ub // Dispatch ID
+mov (1) GatewayPayloadKey:uw 0x1212:uw // Key=0x1212
+
+// Message descriptor
+// bit 31 EOD
+// 27:24 FFID = 0x0011 for msg gateway
+// 23:20 msg length = 1 MRF
+// 19:16 Response length = 0
+// 14 AckReg = 1
+// 1:0 SubFuncID = 00 for OpenGateway
+// Message descriptor: 0 000 0011 0001 0000 + 0 1 000000000000 00 ==> 0000 0011 0001 0000 0100 0000 0000 0000
+// Send message to gateway: the ack message is put into response GRF r49 ==> Good for debugging
+send (8) GatewayResponse:ud m7 GatewayPayload<8;8,1>:ud MSG_GW OGWMSGDSC
+
+//----- End of Open a Message Gateway -----
diff --git a/shaders/h264/ildb/AVC_ILDB_Root_Field_UV.asm b/shaders/h264/ildb/AVC_ILDB_Root_Field_UV.asm new file mode 100644 index 0000000..759e80a --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Root_Field_UV.asm @@ -0,0 +1,9 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#include "AVC_ILDB_Root_UV.asm"
diff --git a/shaders/h264/ildb/AVC_ILDB_Root_Field_Y.asm b/shaders/h264/ildb/AVC_ILDB_Root_Field_Y.asm new file mode 100644 index 0000000..f88eede --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Root_Field_Y.asm @@ -0,0 +1,9 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#include "AVC_ILDB_Root_Y.asm"
diff --git a/shaders/h264/ildb/AVC_ILDB_Root_Mbaff_UV.asm b/shaders/h264/ildb/AVC_ILDB_Root_Mbaff_UV.asm new file mode 100644 index 0000000..01109be --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Root_Mbaff_UV.asm @@ -0,0 +1,140 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+/////////////////////////////////////////////////////////////////////////////////////
+// Kernel name: AVC_ILDB_Root_Mbaff.asm
+//
+// Root kernel serves as a scheduler for child threads.
+//
+//
+// ***** Note *****
+// Initial design bundle MB pair for each thread, and share AVC_ILDB_MB_Dep_Check.asm
+// with non mbaff kernels.
+//
+// Optimization will be done later, putting top and bottom MBs on separate threads.
+//
+//
+/////////////////////////////////////////////////////////////////////////////////////
+//
+// $Revision: 1 $
+// $Date: 10/19/06 5:06p $
+//
+
+// ----------------------------------------------------
+// AVC_ILDB_ROOT_MBAFF_UV
+// ----------------------------------------------------
+#define AVC_ILDB
+
+.kernel AVC_ILDB_ROOT_MBAFF_UV
+#if defined(COMBINED_KERNEL)
+ILDB_LABEL(AVC_ILDB_ROOT_UV):
+#endif
+
+#include "setupVPKernel.asm"
+#include "AVC_ILDB.inc"
+
+
+#if defined(_DEBUG)
+
+/////////////////////////////////////////////////////////////////////////////////////
+// Init URB space for running on RTL. It satisfies reading an unwritten URB entries.
+// Will remove it for production release.
+
+
+//mov (8) m1:ud 0x11111111:ud
+//mov (8) m2:ud 0x22222222:ud
+//mov (8) m3:ud 0x33333333:ud
+//mov (8) m4:ud 0x44444444:ud
+
+//mov (1) Temp1_W:w 0:w
+
+//ILDB_INIT_URB:
+//mul (1) URBOffset:w Temp1_W:w 4:w
+//shl (1) URBWriteMsgDescLow:uw URBOffset:w 4:w // Msg descriptor: URB write dest offset (9:4)
+//mov (1) URBWriteMsgDescHigh:uw 0x0650:uw // Msg descriptor: URB write 5 MRFs (m0 - m4)
+//#include "writeURB.asm"
+
+//add (1) Temp1_W:w Temp1_W:w 1:w // Increase block count
+//cmp.l.f0.0 (1) null Temp1_W:w MBsCntY:w // Check the block count limit
+//(f0.0) jmpi ILDB_INIT_URB // Loop back
+
+
+/////////////////////////////////////////////////////////////////////////////////////
+
+
+mov (1) EntrySignature:w 0xEFF0:w
+
+#endif
+//----------------------------------------------------------------------------------------------------------------
+
+// Set global variable
+mov (32) ChildParam:uw 0:uw // Reset local variables
+//mul (1) TotalBlocks:w MBsCntX:w MBsCntY:w // Total # of MB pairs
+//add (1) GatewayApertureE:w MBsCntY:w GatewayApertureB:w // Aperture End = aperture Head + BlockCntY
+
+
+// 2 URB entries for Y:
+// Entry 0 - Child thread R0Hdr
+// Entry 1 - input parameter to child kernel (child r1)
+
+#define URB_ENTRIES_PER_MB 2
+
+// URB_ENTRIES_PER_MB in differnt form, the final desired format is (URB_ENTRIES_PER_MB-1) << 10
+mov (1) URB_EntriesPerMB_2:w URB_ENTRIES_PER_MB-1:w
+shl (1) URB_EntriesPerMB_2:w URB_EntriesPerMB_2:w 10:w
+
+#define CHROMA_ROOT // Compiling flag for chroma only
+//mul (1) URBOffsetUVBase:w MBsCntY:w URB_ENTRIES_PER_MB:w // Right after Y entries
+
+// URB base for UV kernels
+#if defined(DEV_CL)
+ mov (1) URBOffsetUVBase:w 240:w
+#else
+ mov (1) URBOffsetUVBase:w 320:w
+#endif
+
+
+mov (1) ChildThreadsID:uw 3:uw
+
+shr (1) ThreadLimit:w MaxThreads:w 1:w // Initial luma thread limit to 50%
+mul (1) TotalBlocks:w MBsCntX:w MBsCntY:w // MBs to be processed count down from TotalBlocks
+
+//***** Init CT_R0Hdr fields that are common to all threads *************************
+mov (8) CT_R0Hdr.0:ud r0.0<8;8,1>:ud // Init to root R0 header
+mov (1) CT_R0Hdr.7:ud r0.6:ud // Copy Parent Thread Cnt; JJ did the change on 06/20/2006
+mov (1) CT_R0Hdr.31:ub 0:w // Reset the highest byte
+mov (1) CT_R0Hdr.3:ud 0x00000000
+mov (1) CT_R0Hdr.6:uw sr0.0:uw // sr0.0: state reg contains general thread states, e.g. EUID/TID.
+
+//***** Init ChildParam fields that are common to all threads ***********************
+mov (8) ChildParam<1>:ud RootParam<8;8,1>:ud // Copy all root parameters
+mov (4) CurCol<1>:w 0:w // Reset CurCol, CurRow
+add (2) LastCol<1>:w MBsCntX<2;2,1>:w -1:w // Get LastCol and LastRow
+
+mov (1) URBWriteMsgDesc:ud MSG_LEN(2)+URBWMSGDSC:ud
+
+//===================================================================================
+
+#include "AVC_ILDB_OpenGateway.asm" // Open root thread gateway for receiving notification
+
+#include "AVC_ILDB_Dep_Check.asm" // Check dependency and spawn all MBs
+
+//#include "AVC_ILDB_UpdateThrdLimit.asm" // Update thread limit in luma root thread via gateway
+
+#include "AVC_ILDB_CloseGateway.asm" // Close root thread gateway
+
+// Chroma root EOT = child send EOT : Request type = 1
+ END_CHILD_THREAD
+
+#undef CHROMA_ROOT
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Root_Mbaff_Y.asm b/shaders/h264/ildb/AVC_ILDB_Root_Mbaff_Y.asm new file mode 100644 index 0000000..ae80a36 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Root_Mbaff_Y.asm @@ -0,0 +1,140 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+/////////////////////////////////////////////////////////////////////////////////////
+// Kernel name: AVC_ILDB_Root_Mbaff.asm
+//
+// Root kernel serves as a scheduler for child threads.
+//
+//
+// ***** Note *****
+// Initial design bundle MB pair for each thread, and share AVC_ILDB_MB_Dep_Check.asm
+// with non mbaff kernels.
+//
+// Optimization will be done later, putting top and bottom MBs on separate threads.
+//
+//
+/////////////////////////////////////////////////////////////////////////////////////
+//
+// $Revision: 1 $
+// $Date: 10/19/06 5:06p $
+//
+
+// ----------------------------------------------------
+// AVC_ILDB_ROOT_MBAFF_Y
+// ----------------------------------------------------
+#define AVC_ILDB
+
+.kernel AVC_ILDB_ROOT_MBAFF_Y
+#if defined(COMBINED_KERNEL)
+ILDB_LABEL(AVC_ILDB_ROOT_Y):
+#endif
+
+#include "setupVPKernel.asm"
+#include "AVC_ILDB.inc"
+
+
+#if defined(_DEBUG)
+
+/////////////////////////////////////////////////////////////////////////////////////
+// Init URB space for running on RTL. It satisfies reading an unwritten URB entries.
+// Will remove it for production release.
+
+
+//mov (8) m1:ud 0x11111111:ud
+//mov (8) m2:ud 0x22222222:ud
+//mov (8) m3:ud 0x33333333:ud
+//mov (8) m4:ud 0x44444444:ud
+
+//mov (1) Temp1_W:w 0:w
+
+//ILDB_INIT_URB:
+//mul (1) URBOffset:w Temp1_W:w 4:w
+//shl (1) URBWriteMsgDescLow:uw URBOffset:w 4:w // Msg descriptor: URB write dest offset (9:4)
+//mov (1) URBWriteMsgDescHigh:uw 0x0650:uw // Msg descriptor: URB write 5 MRFs (m0 - m4)
+//#include "writeURB.asm"
+
+//add (1) Temp1_W:w Temp1_W:w 1:w // Increase block count
+//cmp.l.f0.0 (1) null Temp1_W:w MBsCntY:w // Check the block count limit
+//(f0.0) jmpi ILDB_INIT_URB // Loop back
+
+/////////////////////////////////////////////////////////////////////////////////////
+
+
+mov (1) EntrySignature:w 0xEFF0:w
+
+#endif
+//----------------------------------------------------------------------------------------------------------------
+
+// Set global variable
+mov (32) ChildParam:uw 0:uw // Reset local variables
+//mul (1) TotalBlocks:w MBsCntX:w MBsCntY:w // Total # of MB pairs
+//add (1) GatewayApertureE:w MBsCntY:w GatewayApertureB:w // Aperture End = aperture Head + BlockCntY
+
+
+// 2 URB entries for Y:
+// Entry 0 - Child thread R0Hdr
+// Entry 1 - input parameter to child kernel (child r1)
+
+#undef URB_ENTRIES_PER_MB
+#define URB_ENTRIES_PER_MB 2
+
+// URB_ENTRIES_PER_MB in differnt form, the final desired format is (URB_ENTRIES_PER_MB-1) << 10
+mov (1) URB_EntriesPerMB_2:w URB_ENTRIES_PER_MB-1:w
+shl (1) URB_EntriesPerMB_2:w URB_EntriesPerMB_2:w 10:w
+
+mov (1) ChildThreadsID:uw 1:uw // ChildThreadsID for chroma root
+
+shr (1) ThreadLimit:w MaxThreads:w 1:w // Initial luma thread limit to 50%
+mul (1) TotalBlocks:w MBsCntX:w MBsCntY:w // MBs to be processed count down from TotalBlocks
+
+//***** Init CT_R0Hdr fields that are common to all threads *************************
+mov (8) CT_R0Hdr.0:ud r0.0<8;8,1>:ud // Init to root R0 header
+mov (1) CT_R0Hdr.7:ud r0.6:ud // Copy Parent Thread Cnt; JJ did the change on 06/20/2006
+mov (1) CT_R0Hdr.31:ub 0:w // Reset the highest byte
+mov (1) CT_R0Hdr.3:ud 0x00000000
+mov (1) CT_R0Hdr.6:uw sr0.0:uw // sr0.0: state reg contains general thread states, e.g. EUID/TID.
+
+//***** Init ChildParam fields that are common to all threads ***********************
+mov (8) ChildParam<1>:ud RootParam<8;8,1>:ud // Copy all root parameters
+mov (4) CurCol<1>:w 0:w // Reset CurCol, CurRow
+add (2) LastCol<1>:w MBsCntX<2;2,1>:w -1:w // Get LastCol and LastRow
+
+mov (1) URBWriteMsgDesc:ud MSG_LEN(2)+URBWMSGDSC:ud
+
+//===================================================================================
+
+#include "AVC_ILDB_OpenGateway.asm" // Open root thread gateway for receiving notification
+
+#if defined(DEV_CL)
+ mov (1) URBOffset:uw 240:uw // Use chroma URB offset to spawn chroma root
+#else
+ mov (1) URBOffset:uw 320:uw // Use chroma URB offset to spawn chroma root
+#endif
+
+#include "AVC_ILDB_SpawnChromaRoot.asm" // Spawn chroma root
+
+mov (1) URBOffset:uw 0:uw // Use luma URB offset to spawn luma child
+mov (1) ChildThreadsID:uw 2:uw // Starting ChildThreadsID for luma child threads
+
+#include "AVC_ILDB_Dep_Check.asm" // Check dependency and spawn all MBs
+
+// Wait for UV root thread to finish
+ILDB_LABEL(WAIT_FOR_UV):
+cmp.l.f0.0 (1) null:w ThreadLimit:w MaxThreads:w
+(f0.0) jmpi ILDB_LABEL(WAIT_FOR_UV)
+
+#include "AVC_ILDB_CloseGateway.asm" // Close root thread gateway
+
+END_THREAD // End of root thread
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Root_UV.asm b/shaders/h264/ildb/AVC_ILDB_Root_UV.asm new file mode 100644 index 0000000..cb8e146 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Root_UV.asm @@ -0,0 +1,127 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: AVC_ILDB_Root_UV.asm
+//
+// Root kernel serves as a scheduler for child threads
+//
+// $Revision: 1 $
+// $Date: 10/19/06 5:06p $
+//
+
+// ----------------------------------------------------
+// AVC_ILDB_ROOT_UV
+// ----------------------------------------------------
+#define AVC_ILDB
+
+.kernel AVC_ILDB_ROOT_UV
+#if defined(COMBINED_KERNEL)
+ILDB_LABEL(AVC_ILDB_ROOT_UV):
+#endif
+
+#include "setupVPKernel.asm"
+#include "AVC_ILDB.inc"
+
+#if defined(_DEBUG)
+mov (1) EntrySignature:w 0xFF11:w
+#endif
+
+/////////////////////////////////////////////////////////////////////////////////////
+#if defined(_DEBUG)
+
+// Init URB space for running on RTL. It satisfies reading an unwritten URB entries.
+// Will remove it for production release.
+
+mov (8) m1:ud 0x55555555:ud
+mov (8) m2:ud 0x66666666:ud
+mov (8) m3:ud 0x77777777:ud
+mov (8) m4:ud 0x88888888:ud
+
+mov (1) Temp1_W:w MBsCntY:w
+shl (1) Temp2_W:w MBsCntY:w 1:w
+
+ILDB_LABEL(ILDB_INIT_URB_UV):
+mul (1) URBOffset:uw Temp1_W:uw 4:w // Each thread uses 4 URB entries (1 r0 + 1 inline + 2 data)
+mov (1) URBWriteMsgDesc:ud MSG_LEN(4)+URBWMSGDSC:ud // Msg descriptor: URB write msg length = 5
+#include "writeURB.asm"
+
+add (1) Temp1_W:w Temp1_W:w 1:w // Increase block count
+cmp.l.f0.0 (1) null Temp1_W:w Temp2_W:w // Check the block count limit
+(f0.0) jmpi ILDB_LABEL(ILDB_INIT_URB_UV) // Loop back
+
+mov (1) EntrySignature:w 0xFFF0:w
+
+#endif
+/////////////////////////////////////////////////////////////////////////////////////
+
+// Set global variable
+mov (32) ChildParam:uw 0:uw // Reset local variables, 2 GRFs
+//mul (1) TotalBlocks:w MBsCntX:w MBsCntY:w // Total # of blocks
+//add (1) GatewayApertureE:w MBsCntY:w GatewayApertureB:w // Aperture End = aperture Head + BlockCntY
+
+// 4 URB entries for Y:
+// Entry 0 - Child thread R0Hdr
+// Entry 1 - input parameter to child kernel (child r1)
+// Entry 2 - Prev MB data UV 2x8
+// Entry 3 - Unused
+
+#define URB_ENTRIES_PER_MB 4
+
+// URB_ENTRIES_PER_MB in differnt form, the final desired format is (URB_ENTRIES_PER_MB-1) << 10
+mov (1) URB_EntriesPerMB_2:w URB_ENTRIES_PER_MB-1:w
+shl (1) URB_EntriesPerMB_2:w URB_EntriesPerMB_2:w 10:w
+
+#define CHROMA_ROOT // Compiling flag for chroma only
+
+// URB base for UV kernels
+#if defined(DEV_CL)
+ mov (1) URBOffsetUVBase:w 240:w
+#else
+ mov (1) URBOffsetUVBase:w 320:w
+#endif
+
+
+mov (1) ChildThreadsID:uw 3:uw
+
+shr (1) ThreadLimit:w MaxThreads:w 1:w // Initial luma thread limit to 50%
+mul (1) TotalBlocks:w MBsCntX:w MBsCntY:w // MBs to be processed count down from TotalBlocks
+
+//***** Init CT_R0Hdr fields that are common to all threads *************************
+mov (8) CT_R0Hdr.0:ud r0.0<8;8,1>:ud // Init to root R0 header
+mov (1) CT_R0Hdr.7:ud r0.6:ud // Copy Parent Thread Cnt; JJ did the change on 06/20/2006
+mov (1) CT_R0Hdr.31:ub 0:w // Reset the highest byte
+mov (1) CT_R0Hdr.3:ud 0x00000000
+mov (1) CT_R0Hdr.6:uw sr0.0:uw // sr0.0: state reg contains general thread states, e.g. EUID/TID.
+
+//***** Init ChildParam fields that are common to all threads ***********************
+mov (8) ChildParam<1>:ud RootParam<8;8,1>:ud // Copy all root parameters
+mov (4) CurCol<1>:w 0:w // Reset CurCol, CurRow,
+add (2) LastCol<1>:w MBsCntX<2;2,1>:w -1:w // Get LastCol and LastRow
+
+mov (1) URBWriteMsgDesc:ud MSG_LEN(2)+URBWMSGDSC:ud
+
+//===================================================================================
+
+#include "AVC_ILDB_OpenGateway.asm" // Open gateway for receiving notification
+
+#include "AVC_ILDB_Dep_Check.asm" // Check dependency and spawn all luma child threads in parallel with chroma root
+
+//#include "AVC_ILDB_LumaThrdLimit.asm" // Update thread limit in luma root thread via gateway
+
+#include "AVC_ILDB_CloseGateway.asm" // Close root thread gateway
+
+// Chroma root EOT = child send EOT : Request type = 1
+ END_CHILD_THREAD
+
+#undef CHROMA_ROOT
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Root_Y.asm b/shaders/h264/ildb/AVC_ILDB_Root_Y.asm new file mode 100644 index 0000000..2274b93 --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Root_Y.asm @@ -0,0 +1,130 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: AVC_ILDB_Root_Y.asm
+//
+// Root kernel serves as a scheduler for child threads
+//
+// $Revision: 1 $
+// $Date: 10/19/06 5:06p $
+//
+
+// ----------------------------------------------------
+// AVC_ILDB_ROOT_Y
+// ----------------------------------------------------
+#define AVC_ILDB
+
+.kernel AVC_ILDB_ROOT_Y
+#if defined(COMBINED_KERNEL)
+ILDB_LABEL(AVC_ILDB_ROOT_Y):
+#endif
+
+#include "setupVPKernel.asm"
+#include "AVC_ILDB.inc"
+
+/////////////////////////////////////////////////////////////////////////////////////
+#if defined(_DEBUG)
+
+// Init URB space for running on RTL. It satisfies reading an unwritten URB entries.
+// Will remove it for production release.
+
+mov (8) m1:ud 0x11111111:ud
+mov (8) m2:ud 0x22222222:ud
+mov (8) m3:ud 0x33333333:ud
+mov (8) m4:ud 0x44444444:ud
+
+mov (1) Temp1_W:w 0:w
+
+ILDB_LABEL(ILDB_INIT_URB_Y):
+//mul (1) Temp2_W:w Temp1_W:w 4:w // URBOffset
+//shl (1) URBWriteMsgDescLow:uw Temp2_W:w 4:w // Msg descriptor: URB write dest offset (9:4)
+//mov (1) URBWriteMsgDescHigh:uw 0x0650:uw // Msg descriptor: URB write 5 MRFs (m0 - m4)
+
+//mul (1) URBOffset:uw Temp1_W:uw 4:w // Each thread uses 4 URB entries (1 r0 + 1 inline + 2 data)
+mul (1) URBOffset:uw Temp1_W:uw 2:w // Each thread uses 2 URB entries (1 r0 + 1 inline)
+mov (1) URBWriteMsgDesc:ud MSG_LEN(2)+URBWMSGDSC:ud // Msg descriptor: URB write msg length = 3
+#include "writeURB.asm"
+
+add (1) Temp1_W:w Temp1_W:w 1:w // Increase block count
+cmp.l.f0.0 (1) null Temp1_W:w MBsCntY:w // Check the block count limit
+(f0.0) jmpi ILDB_LABEL(ILDB_INIT_URB_Y) // Loop back
+
+mov (1) EntrySignature:w 0xFFF0:w
+
+#endif
+/////////////////////////////////////////////////////////////////////////////////////
+
+
+// Set global variable
+mov (32) ChildParam:uw 0:uw // Reset local variables, 2 GRFs
+//mul (1) TotalBlocks:w MBsCntX:w MBsCntY:w // Total # of blocks
+//add (1) GatewayApertureE:w MBsCntY:w GatewayApertureB:w // Aperture End = aperture Head + BlockCntY
+
+// 4 URB entries for Y:
+// Entry 0 - Child thread R0Hdr
+// Entry 1 - input parameter to child kernel (child r1)
+// Entry 2 - Prev MB data Y 4x16, col 1 and col 0
+// Entry 3 - Prev MB data Y 4x16, col 3 and col 2
+
+#undef URB_ENTRIES_PER_MB
+#define URB_ENTRIES_PER_MB 4
+
+// URB_ENTRIES_PER_MB in differnt form, the final desired format is (URB_ENTRIES_PER_MB-1) << 10
+mov (1) URB_EntriesPerMB_2:w URB_ENTRIES_PER_MB-1:w
+shl (1) URB_EntriesPerMB_2:w URB_EntriesPerMB_2:w 10:w
+
+shr (1) ThreadLimit:w MaxThreads:w 1:w // Initial luma thread limit to 50%
+mul (1) TotalBlocks:w MBsCntX:w MBsCntY:w // MBs to be processed count down from TotalBlocks
+
+//***** Init CT_R0Hdr fields that are common to all threads *************************
+mov (8) CT_R0Hdr.0:ud r0.0<8;8,1>:ud // Init to root R0 header
+mov (1) CT_R0Hdr.7:ud r0.6:ud // Copy Parent Thread Cnt; JJ did the change on 06/20/2006
+mov (1) CT_R0Hdr.31:ub 0:w // Reset the highest byte
+mov (1) CT_R0Hdr.3:ud 0x00000000
+mov (1) CT_R0Hdr.6:uw sr0.0:uw // sr0.0: state reg contains general thread states, e.g. EUID/TID.
+
+//***** Init ChildParam fields that are common to all threads ***********************
+mov (8) ChildParam<1>:ud RootParam<8;8,1>:ud // Copy all root parameters
+mov (4) CurCol<1>:w 0:w // Reset CurCol, CurRow,
+add (2) LastCol<1>:w MBsCntX<2;2,1>:w -1:w // Get LastCol and LastRow
+
+mov (1) URBWriteMsgDesc:ud MSG_LEN(2)+URBWMSGDSC:ud
+
+//===================================================================================
+
+#include "AVC_ILDB_OpenGateway.asm" // Open gateway for receiving notification
+
+#if defined(DEV_CL)
+ mov (1) URBOffset:uw 240:uw // Use chroma URB offset to spawn chroma root
+#else
+ mov (1) URBOffset:uw 320:uw // Use chroma URB offset to spawn chroma root
+#endif
+
+#include "AVC_ILDB_SpawnChromaRoot.asm" // Spawn chroma root
+
+mov (1) URBOffset:uw 0:uw // Use luma URB offset to spawn luma child
+mov (1) ChildThreadsID:uw 2:uw // Starting ChildThreadsID for luma child threads
+
+#include "AVC_ILDB_Dep_Check.asm" // Check dependency and spawn all luma child threads in parallel with chroma root
+
+
+// Wait for UV root thread to finish
+ILDB_LABEL(WAIT_FOR_UV):
+cmp.l.f0.0 (1) null:w ThreadLimit:w MaxThreads:w
+(f0.0) jmpi ILDB_LABEL(WAIT_FOR_UV)
+
+
+#include "AVC_ILDB_CloseGateway.asm" // Close root thread gateway
+
+END_THREAD // End of root thread
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
diff --git a/shaders/h264/ildb/AVC_ILDB_Spawn.asm b/shaders/h264/ildb/AVC_ILDB_Spawn.asm new file mode 100644 index 0000000..5b5c91e --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_Spawn.asm @@ -0,0 +1,22 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//=============== Spawn a child thread for a vertical child ===============
+
+#if defined(_DEBUG)
+ mov (1) EntrySignature:w 0x6666:w
+#endif
+
+ mul (1) URBOffset:uw CurRow:uw 2:w // 5:w // Each row uses 5 URB entries (R0, child R0, 3 GRFs of data from left MB)
+
+ mov (8) CT_R0Hdr.0:ud r0.0<8;8,1>:ud // Init to root R0 header
+
+ // R0.2: Interface Discriptor Ptr. Add offset 16 for next Interface Discriptor for child kernel
+ add (1) CT_R0Hdr.2:ud r0.2:ud IDesc_Child_Offset:w
+
+ #include "AVC_ILDB_SpawnChild.asm"
diff --git a/shaders/h264/ildb/AVC_ILDB_SpawnChild.asm b/shaders/h264/ildb/AVC_ILDB_SpawnChild.asm new file mode 100644 index 0000000..0f6950c --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_SpawnChild.asm @@ -0,0 +1,55 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//=============== Spawn a child thread for Luma or Chroma ===============
+
+ //----- Create child thread R0 header -----
+#if defined(_DEBUG)
+ mov (1) EntrySignature:w 0xAAAA:w
+#endif
+
+ //***** Set CT_R0Hdr fields that change for every thread
+
+ // Restore CT_R0Hdr.4:ud to r0.4:ud
+ mov (1) CT_R0Hdr.4:ud r0.4:ud
+
+ // R0.2: Interface Discriptor Ptr. Add a child offset for child kernel
+// add (1) CT_R0Hdr.2:ud r0.2:ud CHILD_OFFSET:w
+
+ // Assign a new Thread Count for this child
+ mov (1) CT_R0Hdr.6:ud ChildThreadsID:uw
+
+ //----- Prepare URB for launching a child thread -----
+ mov (16) m2.0:w ChildParam<16;16,1>:w
+
+ shr (1) MRF0.0:uw URBOffset:uw 1:w
+
+ add (1) ChildThreadsID:uw ChildThreadsID:uw 2:uw // Luma child=even, chroma child=odd
+
+ //--------------------------------------------------
+// #include "writeURB.asm"
+ send null:uw MRF0 null:ud URBWRITE URBWriteMsgDesc:ud // URB write
+
+ //--------------------------------------------------
+ // Set URB handle for child thread launching:
+ // URB handle Length (bit 15:10) - 0000 0000 0000 0000 yyyy yy00 0000 0000
+ // URB handle offset (bit 9:0) - 0000 0000 0000 0000 0000 00xx xxxx xxxx
+
+ or (1) CT_R0Hdr.4:ud URB_EntriesPerMB_2:w URBOffset:uw
+
+ // 2 URB entries:
+ // Entry 0 - CT_R0Hdr
+ // Entry 1 - input parameter to child kernel
+
+ //----- Spawn a child now -----
+ send (8) null:ud CT_R0Hdr null:ud TS TSMSGDSC
+// send (8) null:ud CT_Spawn_Reg null:ud 0x07100001
+
+
+ // Restore CT_R0Hdr.4:ud to r0.4:ud for next use
+// mov (1) CT_R0Hdr.4:ud r0.4:ud
diff --git a/shaders/h264/ildb/AVC_ILDB_SpawnChromaRoot.asm b/shaders/h264/ildb/AVC_ILDB_SpawnChromaRoot.asm new file mode 100644 index 0000000..cd5e57e --- /dev/null +++ b/shaders/h264/ildb/AVC_ILDB_SpawnChromaRoot.asm @@ -0,0 +1,47 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//=============== Spawn a chroma root thread ===============
+
+ //----- Create chroma root thread R0 header -----
+#if defined(_DEBUG)
+ mov (1) EntrySignature:w 0xAABA:w
+#endif
+
+
+
+ // Restore CT_R0Hdr.4:ud to r0.4:ud
+// mov (1) CT_R0Hdr.4:ud r0.4:ud
+
+ // R0.2: Interface Discriptor Ptr. Add child offset for child kernel
+ add (1) CT_R0Hdr.2:ud r0.2:ud CHROMA_ROOT_OFFSET:w
+
+ // Assign a new Thread Count for this child
+ mov (1) CT_R0Hdr.6:ud 1:w // ThreadID=1 for chroma root
+
+ //----- Copy luma root r1 for launching chroma root thread -----
+ mov (16) m2.0:w RootParam<16;16,1>:w
+
+ #include "writeURB.asm"
+
+ //--------------------------------------------------
+ // Set URB handle for child thread launching:
+ // URB handle Length (bit 15:10) - 0000 0000 0000 0000 yyyy yy00 0000 0000
+ // URB handle offset (bit 9:0) - 0000 0000 0000 0000 0000 00xx xxxx xxxx
+
+ or (1) CT_R0Hdr.4:ud URB_EntriesPerMB_2:w URBOffset:uw
+
+ // 2 URB entries:
+ // Entry 0 - CT_R0Hdr
+ // Entry 1 - input parameter to child kernel
+
+ //----- Spawn a child now -----
+ send (8) null:ud CT_R0Hdr null:ud TS TSMSGDSC
+
+ // Restore CT_R0Hdr.4:ud to r0.4:ud for next use
+ mov (1) CT_R0Hdr.4:ud r0.4:ud
diff --git a/shaders/h264/ildb/Child_Undefs.inc b/shaders/h264/ildb/Child_Undefs.inc new file mode 100644 index 0000000..fa3ade1 --- /dev/null +++ b/shaders/h264/ildb/Child_Undefs.inc @@ -0,0 +1,24 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Modual name: Child_Undefs.inc
+//
+// Undefine global symbols for new process in child thread
+//
+
+#undef P1
+#undef P2
+#undef P3
+#undef P4
+#undef P5
+#undef P6
+#undef P7
+#undef P8
+#undef EDGECNTLMAP
+#undef CLIP_NEGATIVE
+#undef CLIP_DONE
diff --git a/shaders/h264/ildb/ILDB_header.inc b/shaders/h264/ildb/ILDB_header.inc new file mode 100644 index 0000000..015bf28 --- /dev/null +++ b/shaders/h264/ildb/ILDB_header.inc @@ -0,0 +1,306 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__ILDB_HEADER__) // Make sure this file is only included once
+#define __ILDB_HEADER__
+
+// Module name: ILDB_header.inc
+//
+
+.default_execution_size (16)
+.default_register_type :ub
+
+#undef NULLREG
+#undef RETURN_REG
+#undef EOTMSGDSC
+#undef MSGSRC
+#undef END_THREAD
+#undef TSMSGDSC
+
+// ----------- Common constant definitions ------------
+//
+// Bit position constants
+//
+#define BIT0 0x01
+#define BIT1 0x02
+#define BIT2 0x04
+#define BIT3 0x08
+#define BIT4 0x10
+#define BIT5 0x20
+#define BIT6 0x40
+#define BIT7 0x80
+#define BIT8 0x0100
+#define BIT9 0x0200
+#define BIT10 0x0400
+#define BIT11 0x0800
+#define BIT12 0x1000
+#define BIT13 0x2000
+#define BIT14 0x4000
+#define BIT15 0x8000
+#define BIT16 0x00010000
+#define BIT17 0x00020000
+#define BIT18 0x00040000
+#define BIT19 0x00080000
+#define BIT20 0x00100000
+#define BIT21 0x00200000
+#define BIT22 0x00400000
+#define BIT23 0x00800000
+#define BIT24 0x01000000
+#define BIT25 0x02000000
+#define BIT26 0x04000000
+#define BIT27 0x08000000
+#define BIT28 0x10000000
+#define BIT29 0x20000000
+#define BIT30 0x40000000
+#define BIT31 0x80000000
+
+// Common constants
+//
+#define INST_SIZE 16 // Instruction size in byte
+
+#define GRFWIB 32 // GRF register width in byte
+#define GRFWIW 16 // GRF register width in word
+#define GRFWID 8 // GRF register width in dword
+
+#define TOP_FIELD 0
+#define BOTTOM_FIELD 1
+
+#define PREVIOUS_FRAME 0 // Previous frame
+#define CURRENT_FRAME 1 // Current frame
+#define NEXT_FRAME 2 // Next frame
+
+#define Y_ROW_WIDTH 16 // in bytes
+#define UV_ROW_WIDTH 8
+
+// Useful macros
+//
+#define REGION(Width,HStride) <Width*HStride;Width,HStride> // Region definition when ExecSize = Width
+
+#define NULLREG null<1>:d
+#define NULLREGW null<1>:w
+
+#define RETURN_REG r62 // Return pointer for all sub-routine calls (type DWORD)
+
+#define CALL(subFunc, skipInst) add (1) RETURN_REG<1>:ud ip:ud 1+skipInst*INST_SIZE \n\
+ jmpi (1) subFunc
+
+#define RETURN mov (1) ip:ud RETURN_REG<0;1,0>:ud // Return to calling module
+
+#define PRED_CALL(flag, subFunc, skipInst) add (1) RETURN_REG<1>:ud ip:ud 1+skipInst*INST_SIZE \n\
+ (flag) jmpi (1) subFunc
+
+
+// Definitions for surface states, GRF regions, and common data fields
+//
+// Note: Each kernel needs to define a specific symbol before including this
+// header file to ensure correct definitions.
+//
+
+
+
+#if defined(AVC_ILDB)
+
+.reg_count_total 64
+.reg_count_payload 4
+
+ // Binding Table Index
+ #define BI_CNTRL_DATA 0 // Control data map
+ #define BI_SRC_Y 1
+ #define BI_SRC_UV 2
+ #define BI_DEST_Y 3
+ #define BI_DEST_UV 4
+
+
+ //========== Left MB, 4x16 in r2 and r3 ==========
+ #define PREV_MB_Y_BASE 64 //2*GRFWIB // Byte offset to r2
+ .declare PREV_MB_YD Base=r2 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+ .declare PREV_MB_YW Base=r2 ElementSize=2 SrcRegion=REGION(8,1) Type=uw
+ .declare PREV_MB_YB Base=r2 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+
+ #define PREV_MB_U_BASE 64 //2*GRFWIB // seperate thread from Y // Byte offset to r2
+ .declare PREV_MB_UD Base=r2 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+ .declare PREV_MB_UW Base=r2 ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+ .declare PREV_MB_UB Base=r2 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+
+ #define PREV_MB_V_BASE 65 //2*GRFWIB+1 // NV12 // Byte offset to r2.1
+ .declare PREV_MB_VB Base=r2.1 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+
+ //========== Top MB, 16x4 in r4 and r5 ==========
+ #define TOP_MB_Y_BASE 128 //4*GRFWIB // Byte offset to r4
+ .declare TOP_MB_YD Base=r4 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+ .declare TOP_MB_YW Base=r4 ElementSize=2 SrcRegion=REGION(8,1) Type=uw
+ .declare TOP_MB_YB Base=r4 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+
+ #define TOP_MB_U_BASE 128 //4*GRFWIB // seperate thread from Y // Byte offset to r4
+ .declare TOP_MB_UD Base=r4 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+ .declare TOP_MB_UW Base=r4 ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+ .declare TOP_MB_UB Base=r4 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+
+ #define TOP_MB_V_BASE 129 //4*GRFWIB+1 // NV12 // Byte offset to r4.1
+ .declare TOP_MB_VB Base=r4.1 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+
+
+ //========== Current MB, 16x16 in r6-r13 ==========
+ #define SRC_MB_Y_BASE 192 //6*GRFWIB // Byte offset to r6
+ .declare SRC_YD Base=r6 ElementSize=4 SrcRegion=REGION(8,1) Type=ud // For read and write, 8 GRFs
+ .declare SRC_YW Base=r6 ElementSize=2 SrcRegion=REGION(8,1) Type=uw
+ .declare SRC_YB Base=r6 ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+
+ #define SRC_MB_U_BASE 192 //6*GRFWIB // seperate thread from Y // Byte offset to r6
+ .declare SRC_UD Base=r6 ElementSize=4 SrcRegion=REGION(8,1) Type=ud // For read and write, 2 GRFs
+ .declare SRC_UW Base=r6 ElementSize=2 SrcRegion=REGION(16,1) Type=uw // For read and write, 4 GRFs
+ .declare SRC_UB Base=r6 ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 2 GRFs
+
+ #define SRC_MB_V_BASE 193 // 6*GRFWIB+1 // NV12 // Byte offset to r6.1
+ .declare SRC_VD Base=r6.1 ElementSize=4 SrcRegion=REGION(8,1) Type=ud // For read and write, 2 GRFs
+ .declare SRC_VW Base=r6.1 ElementSize=2 SrcRegion=REGION(16,1) Type=uw // For read and write, 4 GRFs
+ .declare SRC_VB Base=r6.1 ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 2 GRFs
+
+#else // No kernel specified, define nothing.
+
+.reg_count_total 64
+.reg_count_payload 2
+
+ #define SRCAOFF r1.0:ud // Offset into alpha data
+ #define SRCOFF r1.1:ud // Offset into source YUV data
+ #define ORIX r1.4 // :w, H. origin of the destination block in pel
+ #define ORIY r1.5 // :w, V. origin of the destination block in pel
+
+#endif
+
+// ----------- Message Payload Header fields------------
+//
+#define IDP r0.2:ud // Interface Descriptor Pointer
+#define BTP r0.4:ud // Binding Table Pointer
+
+// ----------- Common Message Descriptor ------------
+//
+#ifdef DEV_ILK
+#define GW_DCN // Should be enabled only for ILK-B0 and beyond
+#define MSG_GW 0x03 // Message Gateway
+#define MSG_GW_EOT 0x23 // Message Gateway plus EOT bit set (For ILK only)
+#define DAPREAD 0x04 // Data Port Read Extended Message Descriptor,
+#define DAPWRITE 0x05 // Data Port Write Extended Message Descriptor,
+#define URBWRITE 0x06 // URB
+#define TS 0x07 // Thread Spawner Extended Message Descriptor
+#define TS_EOT 0x27 // End of Thread Extended Message Descriptor
+
+#define EOTMSGDSC 0x02000000 // End of Thread Message Descriptor /w URB handle dereferenced (used by root kernel)
+#define CHILD_EOTMSGDSC 0x02000012 // End of Child Thread Message Descriptor w/o URB handle dereferenced
+
+// Data Port Message Descriptor
+#define DWBRMSGDSC_RC 0x02086000 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_RC_TF 0x02086600 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_RC_BF 0x02086700 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_SC 0x0208A000 // DWORD Block Read Message Descriptor, reading from sampler cache = A.
+#define DWBRMSGDSC_SC_TF 0x0208E600 // DWORD Block Read Message Descriptor, reading top field from field mode sampler cache.
+#define DWBRMSGDSC_SC_BF 0x0208E700 // DWORD Block Read Message Descriptor, reading bottom field from field mode sampler cache.
+
+#define ILDBRMSGDSC 0x02085800 // AVC ILDB Control Data Read Msg Desc on Bearlake-C
+
+#define DWBWMSGDSC 0x02082000 // DWORD Block Write Message Descriptor
+#define DWBWMSGDSC_WC 0x0218A000 // DWORD Block Write Message Descriptor + write commit
+
+// URB Message Descriptor
+#define URBWMSGDSC 0x02080000 // URB Write Message Descriptor
+
+// Thread Spawner Message Descriptor
+#define TSMSGDSC 0x02000001
+
+// Message Gateway Message Descriptors
+#define OGWMSGDSC 0x02000000 // OpenGateway Message Descriptor
+#define CGWMSGDSC 0x02000001 // CloseGateway Message Descriptor
+#define FWDMSGDSC 0x02000002 // ForwardMsg Message Descriptor
+#define NOTIFYMSG 0x00008000 // Send notification with ForwardMsg message
+
+#define RESP_LEN(len) 0x100000*len
+#define MSG_LEN(len) 0x2000000*len
+
+#else // Pre DEV_ILK
+
+#define MSG_GW
+#define MSG_GW_EOT
+#define DAPREAD
+#define DAPWRITE
+#define URBWRITE
+#define TS
+#define TS_EOT
+
+#define EOTMSGDSC 0x87100000 // End of Thread Message Descriptor /w URB handle dereferenced (used by root kernel)
+#define CHILD_EOTMSGDSC 0x87100012 // End of Child Thread Message Descriptor w/o URB handle dereferenced
+
+// Data Port Message Descriptor
+#define DWBRMSGDSC_RC 0x04106000 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_RC_TF 0x04106600 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_RC_BF 0x04106700 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_SC 0x0410A000 // DWORD Block Read Message Descriptor, reading from sampler cache = A.
+#define DWBRMSGDSC_SC_TF 0x0410A600 // DWORD Block Read Message Descriptor, reading top field from field mode sampler cache.
+#define DWBRMSGDSC_SC_BF 0x0410A700 // DWORD Block Read Message Descriptor, reading bottom field from field mode sampler cache.
+
+#define ILDBRMSGDSC 0x04105800 // AVC ILDB Control Data Read Msg Desc on Bearlake-C
+
+#define DWBWMSGDSC 0x05102000 // DWORD Block Write Message Descriptor
+#define DWBWMSGDSC_WC 0x0511A000 // DWORD Block Write Message Descriptor + write commit
+
+// URB Message Descriptor
+#define URBWMSGDSC 0x06100000 // URB Write Message Descriptor
+
+// Thread Spawner Message Descriptor
+#define TSMSGDSC 0x07100001
+
+// Message Gateway Message Descriptors
+#define OGWMSGDSC 0x03100000 // OpenGateway Message Descriptor
+#define CGWMSGDSC 0x03100001 // CloseGateway Message Descriptor
+#define FWDMSGDSC 0x03100002 // ForwardMsg Message Descriptor
+#define NOTIFYMSG 0x00008000 // Send notification with ForwardMsg message
+
+#define RESP_LEN(len) 0x10000*len
+#define MSG_LEN(len) 0x100000*len
+
+// bits 15 - 0 = 01 011 000 00000000 = 0101 1000 0000 0000 = 5800
+// Render cache, AVC loop rd,
+#endif // DEV_ILK
+
+// Enable frame/field selection in message descriptor
+#define ENMSGDSCFM 0x400 // Enable MSGDSC to select frame surface
+#define ENMSGDSCTF 0x600 // Enable MSGDSC to select top field surface
+#define ENMSGDSCBF 0x700 // Enable MSGDSC to select bottom field surface
+
+#define END_THREAD send (8) NULLREG MSGHDR r0:ud TS_EOT EOTMSGDSC
+#define END_CHILD_THREAD send (8) NULLREG MSGHDR r0:ud TS_EOT CHILD_EOTMSGDSC
+
+// ----------- Message related register ------------
+//
+#define MSGHDR m1 // Message Payload Header
+#define MSGHDRY m1 // Message Payload Header register for Y data
+#define MSGHDRU m2 // Message Payload Header register for U data
+#define MSGHDRV m3 // Message Payload Header register for V data
+
+#define MSGHDRC m1 // Message Payload Header register for CUR MB
+#define MSGHDRL m2 // Message Payload Header register for LEFT MB
+#define MSGHDRT m3 // Message Payload Header register for TOP MB
+
+#define MSGHDRYA m4 // Second Message Payload Header register for Y data
+#define MSGSRC r63 // Message source register
+#define MSGDSC a0.0:ud // Message Descriptor register (type DWORD)
+
+#define MH_ORI MSGSRC.0 // DWORD block R/W message header block offset
+#define MH_ORIX MSGSRC.0 // DWORD block R/W message header X offset
+#define MH_ORIY MSGSRC.1 // DWORD block R/W message header Y offset
+#define MH_SIZE MSGSRC.2 // DWORD block R/W message header block width & height
+
+
+// M2 - M9 for message data payload
+.declare MSGPAYLOADB Base=m2 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare MSGPAYLOADW Base=m2 ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+.declare MSGPAYLOADD Base=m2 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare MSGPAYLOADF Base=m2 ElementSize=4 SrcRegion=REGION(8,1) Type=f
+
+// End of ILDB_header.inc
+
+#endif // !defined(__ILDB_HEADER__)
diff --git a/shaders/h264/ildb/Load_ILDB_Cntrl_Data.asm b/shaders/h264/ildb/Load_ILDB_Cntrl_Data.asm new file mode 100644 index 0000000..695ae4e --- /dev/null +++ b/shaders/h264/ildb/Load_ILDB_Cntrl_Data.asm @@ -0,0 +1,80 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_ILDB_Cntrl_Data.asm
+//
+// This module loads AVC ILDB control data for one MB.
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// CNTRL_DATA_D: CNTRL_DATA_D Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+//
+// Binding table index:
+// BI_CNTRL_DATA: Binding table index of control data surface
+//
+//----------------------------------------------------------------
+
+ // We need to get control data offset for the bottom MB in mbaff mode.
+ // That is, get f0.1=1 if MbaffFlag==1 && BotFieldFlag==1
+ and (1) CTemp1_W:uw BitFields:uw MbaffFlag+BotFieldFlag:uw // Mute all other bits
+
+ and.nz.f0.0 (1) null:w BitFields:w CntlDataExpFlag:w // Get CntlDataExpFlag
+
+ cmp.e.f0.1 (1) NULLREGW CTemp1_W:uw MbaffFlag+BotFieldFlag:uw // Check mbaff and bot flags
+
+ (f0.0) jmpi ILDB_LABEL(READ_BLC_CNTL_DATA)
+
+ // On Crestline, MB control data in memory occupy 64 DWs (expanded).
+// mov (1) MSGSRC.0<1>:ud 0:w { NoDDClr } // Block origin X
+// mov (1) MSGSRC.1<1>:ud CntrlDataOffsetY:ud { NoDDClr, NoDDChk } // Block origin Y
+// mov (1) MSGSRC.2<1>:ud 0x000F000F:ud { NoDDChk } // Block width and height (16x16=256 bytes)
+
+ mov (2) MSGSRC.0<1>:ud ORIX_CUR<2;2,1>:uw { NoDDClr } // Block origin X,Y
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud { NoDDChk } // Block width and height (16x16=256 bytes)
+
+ (f0.1) add (1) MSGSRC.1:ud MSGSRC.1:ud 16:w // +16 to for bottom MB in a pair
+
+ send (8) CNTRL_DATA_D(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD DWBRMSGDSC_SC+0x00080000+BI_CNTRL_DATA // Receive 8 GRFs
+ jmpi ILDB_LABEL(READ_CNTL_DATA_DONE)
+
+
+ILDB_LABEL(READ_BLC_CNTL_DATA):
+ // On Bearlake-C, MB control data in memory occupy 16 DWs. Data port returns 8 GRFs with expanded control data.
+
+ // Global offset
+ mov (1) MSGSRC.2:ud CntrlDataOffsetY:ud // CntrlDataOffsetY is the global offset
+
+ (f0.1) add (1) MSGSRC.2:ud MSGSRC.2:ud 64:w // +64 to the next MB control data (bot MB)
+
+ send (8) CNTRL_DATA_D(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(8)+ILDBRMSGDSC+BI_CNTRL_DATA // Receive 8 GRFs
+
+ILDB_LABEL(READ_CNTL_DATA_DONE):
+
+// End of load_ILDB_Cntrl_Data.asm
+
+
+
+
+// AVC ILDB control data message header format
+
+//DWord Bit Description
+//M0.7 31:0 Debug
+//M0.6 31:0 Debug
+//M0.5 31:8 Ignored
+// 7:0 Dispatch ID. // This ID is assigned by the fixed function unit and is a unique identifier for the thread. It is used to free up resources used by the thread upon thread completion.
+//M0.4 31:0 Ignored
+//M0.3 31:0 Ignored
+//M0.2 31:0 Global Offset. Specifies the global byte offset into the buffer.
+ // This offset must be OWord aligned (bits 3:0 MBZ) Format = U32 Range = [0,FFFFFFF0h]
+//M0.1 31:0 Ignored
+//M0.0 31:0 Ignored
+
+
+
diff --git a/shaders/h264/ildb/Load_ILDB_Cntrl_Data_16DW.asm b/shaders/h264/ildb/Load_ILDB_Cntrl_Data_16DW.asm new file mode 100644 index 0000000..4c91a29 --- /dev/null +++ b/shaders/h264/ildb/Load_ILDB_Cntrl_Data_16DW.asm @@ -0,0 +1,62 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_ILDB_Cntrl_Data_16DW.asm
+//
+// This module loads AVC ILDB 64DW control data for one MB CTG.
+// Dataport expands from 16DW to 64DW.
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// CNTRL_DATA_D: CNTRL_DATA_D Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+//
+// Binding table index:
+// BI_CNTRL_DATA: Binding table index of control data surface
+//
+//----------------------------------------------------------------
+
+// On CTG, MB control data in memory occupy 16 DWs. Data port returns 8 GRFs with expanded control data.
+
+#if defined(_MBAFF)
+ // We need to get control data offset for the bottom MB in mbaff mode.
+ // That is, get f0.1=1 if MbaffFlag==1 && BotFieldFlag==1
+// and (1) CTemp1_W:uw BitFields:uw MbaffFlag+BotFieldFlag:uw // Mute all other bits
+// cmp.e.f0.1 (1) NULLREGW CTemp1_W:uw MbaffFlag+BotFieldFlag:uw // Check mbaff and bot flags
+
+ and.ne.f0.1 (1) NULLREGW BitFields:uw BotFieldFlag:uw
+
+ // Global offset
+ mov (1) MSGSRC.2:ud CntrlDataOffsetY:ud
+
+ (f0.1) add (1) MSGSRC.2:ud MSGSRC.2:ud 64:w // +64 to the next MB control data (bot MB)
+#endif
+
+ send (8) CNTRL_DATA_D(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(8)+ILDBRMSGDSC+BI_CNTRL_DATA // Receive 8 GRFs
+
+// End of load_ILDB_Cntrl_Data_16DW.asm
+
+
+
+// AVC ILDB control data message header format
+
+//DWord Bit Description
+//M0.7 31:0 Debug
+//M0.6 31:0 Debug
+//M0.5 31:8 Ignored
+// 7:0 Dispatch ID. // This ID is assigned by the fixed function unit and is a unique identifier for the thread. It is used to free up resources used by the thread upon thread completion.
+//M0.4 31:0 Ignored
+//M0.3 31:0 Ignored
+//M0.2 31:0 Global Offset. Specifies the global byte offset into the buffer.
+ // This offset must be OWord aligned (bits 3:0 MBZ) Format = U32 Range = [0,FFFFFFF0h]
+//M0.1 31:0 Ignored
+//M0.0 31:0 Ignored
+
+
+
diff --git a/shaders/h264/ildb/Load_ILDB_Cntrl_Data_22DW.asm b/shaders/h264/ildb/Load_ILDB_Cntrl_Data_22DW.asm new file mode 100644 index 0000000..19e9a20 --- /dev/null +++ b/shaders/h264/ildb/Load_ILDB_Cntrl_Data_22DW.asm @@ -0,0 +1,36 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_ILDB_Cntrl_Data_22DW.asm
+//
+// ********** Apple only module **********
+//
+// This module loads AVC ILDB 22DW control data for one MB for CLN.
+// The reduced control data set is for progressive picture ONLY.
+//
+// Control data memory layout for each MB is 8x11 = 88 bytes.
+// It ocuppies 3 GRFs after reading in.
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// CNTRL_DATA_D: CNTRL_DATA_D Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 3 GRFs
+//
+// Binding table index:
+// BI_CNTRL_DATA: Binding table index of control data surface
+//
+//----------------------------------------------------------------
+
+ mul (1) MSGSRC.0<1>:ud ORIX:uw 8:uw { NoDDClr } // Block origin X
+ mul (1) MSGSRC.1<1>:ud ORIY:uw 11:uw { NoDDClr, NoDDChk } // Block origin Y
+ mov (1) MSGSRC.2<1>:ud 0x000A0007:ud { NoDDChk } // Block width and height (8x11=88 bytes)
+
+ send (8) CNTRL_DATA_D(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(3)+DWBRMSGDSC_SC+BI_CNTRL_DATA // Receive 3 GRFs
+
+// End of load_ILDB_Cntrl_Data_22DW.asm
diff --git a/shaders/h264/ildb/Load_ILDB_Cntrl_Data_64DW.asm b/shaders/h264/ildb/Load_ILDB_Cntrl_Data_64DW.asm new file mode 100644 index 0000000..b026afb --- /dev/null +++ b/shaders/h264/ildb/Load_ILDB_Cntrl_Data_64DW.asm @@ -0,0 +1,42 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_ILDB_Cntrl_Data_64DW.asm
+//
+// This module loads AVC ILDB 64DW control data for one MB for CLN.
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// CNTRL_DATA_D: CNTRL_DATA_D Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+//
+// Binding table index:
+// BI_CNTRL_DATA: Binding table index of control data surface
+//
+//----------------------------------------------------------------
+
+// On CLN, MB control data in memory occupy 64 DWs.
+
+#if defined(_MBAFF)
+ // We need to get control data offset for the bottom MB in mbaff mode.
+ // That is, set f0.1=1 if MbaffFlag==1 && BotFieldFlag==1
+ and (1) acc0.0:uw BitFields:uw MbaffFlag+BotFieldFlag:uw // Mute all other bits
+ cmp.e.f0.1 (1) NULLREGW acc0.0:uw MbaffFlag+BotFieldFlag:uw // Check mbaff and bot flags
+#endif // CTemp1_W
+
+ mov (2) MSGSRC.0<1>:ud ORIX_CUR<2;2,1>:uw { NoDDClr } // Block origin X,Y
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud { NoDDChk } // Block width and height (16x16=256 bytes)
+
+#if defined(_MBAFF)
+ (f0.1) add (1) MSGSRC.1:ud MSGSRC.1:ud 16:w // +16 to the bottom MB control data (bot MB)
+#endif
+
+ send (8) CNTRL_DATA_D(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(8)+DWBRMSGDSC_SC+BI_CNTRL_DATA // Receive 8 GRFs
+
+// End of load_ILDB_Cntrl_Data_64DW.asm
diff --git a/shaders/h264/ildb/Makefile.am b/shaders/h264/ildb/Makefile.am new file mode 100644 index 0000000..e69de29 --- /dev/null +++ b/shaders/h264/ildb/Makefile.am diff --git a/shaders/h264/ildb/Root_Undefs.inc b/shaders/h264/ildb/Root_Undefs.inc new file mode 100644 index 0000000..c0ee016 --- /dev/null +++ b/shaders/h264/ildb/Root_Undefs.inc @@ -0,0 +1,27 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Modual name: Root_Undefs.inc
+//
+// Undefine global symbols for new process in root thread
+//
+
+#undef READ_BI
+#undef WRITE_BI
+
+#undef ILDB_H_INDEPENDENT
+#undef ILDB_H_INDEPENDENT_CONT
+#undef ILDB_H_DEPENDENT
+#undef ILDB_H_DEPENDENT_SCAN
+#undef ILDB_H_NO_DEPENDENT
+
+#undef ILDB_V_INDEPENDENT
+#undef ILDB_V_INDEPENDENT_CONT
+#undef ILDB_V_DEPENDENT
+#undef ILDB_V_DEPENDENT_SCAN
+#undef ILDB_V_NO_DEPENDENT
diff --git a/shaders/h264/ildb/SetupVPKernel.asm b/shaders/h264/ildb/SetupVPKernel.asm new file mode 100644 index 0000000..c7d9634 --- /dev/null +++ b/shaders/h264/ildb/SetupVPKernel.asm @@ -0,0 +1,24 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Modual name: SetupVPKernel.asm
+//
+// Initial setup for running video-processing kernels
+//
+
+#include "ILDB_header.inc"
+
+//
+// Now, begin source code....
+//
+
+.code
+
+ mov (8) MSGSRC.0<1>:ud r0.0<8;8,1>:ud // Initialize message payload header with R0
+
+// End of SetupVPKernel
diff --git a/shaders/h264/ildb/TransposeNV12_16x16.asm b/shaders/h264/ildb/TransposeNV12_16x16.asm new file mode 100644 index 0000000..192a89f --- /dev/null +++ b/shaders/h264/ildb/TransposeNV12_16x16.asm @@ -0,0 +1,135 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////
+// Module name: TransposeNV12_16x16.asm
+//
+// Transpose a 16x16 NV12 MB. The output is also in NV12
+//
+//----------------------------------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region is :ub
+// SRC_YB: SRC_YB Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+// SRC_UW: SRC_UW Base=rxx ElementSize=2 SrcRegion=REGION(8,1) Type=uw // 4 GRFs
+//
+// Temp buffer:
+// BUF_B: BUF_B Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+// BUF_W: BUF_W Base=rxx ElementSize=2 SrcRegion=REGION(8,1) Type=uw // 4 GRFs
+//
+//////////////////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDDA:w
+#endif
+
+
+// Transpose Y (16x16 bytes)
+
+// The first step
+mov (16) BUF_B(0,0)<1> SRC_YB(0,0)<16;4,1>
+mov (16) BUF_B(0,16)<1> SRC_YB(2,0)<16;4,1>
+mov (16) BUF_B(1,0)<1> SRC_YB(4,0)<16;4,1>
+mov (16) BUF_B(1,16)<1> SRC_YB(6,0)<16;4,1>
+
+mov (16) BUF_B(2,0)<1> SRC_YB(0,4)<16;4,1>
+mov (16) BUF_B(2,16)<1> SRC_YB(2,4)<16;4,1>
+mov (16) BUF_B(3,0)<1> SRC_YB(4,4)<16;4,1>
+mov (16) BUF_B(3,16)<1> SRC_YB(6,4)<16;4,1>
+
+mov (16) BUF_B(4,0)<1> SRC_YB(0,8)<16;4,1>
+mov (16) BUF_B(4,16)<1> SRC_YB(2,8)<16;4,1>
+mov (16) BUF_B(5,0)<1> SRC_YB(4,8)<16;4,1>
+mov (16) BUF_B(5,16)<1> SRC_YB(6,8)<16;4,1>
+
+mov (16) BUF_B(6,0)<1> SRC_YB(0,12)<16;4,1>
+mov (16) BUF_B(6,16)<1> SRC_YB(2,12)<16;4,1>
+mov (16) BUF_B(7,0)<1> SRC_YB(4,12)<16;4,1>
+mov (16) BUF_B(7,16)<1> SRC_YB(6,12)<16;4,1>
+
+// The second step
+mov (16) SRC_YB(0,0)<1> BUF_B(0,0)<32;8,4>
+mov (16) SRC_YB(0,16)<1> BUF_B(0,1)<32;8,4>
+mov (16) SRC_YB(1,0)<1> BUF_B(0,2)<32;8,4>
+mov (16) SRC_YB(1,16)<1> BUF_B(0,3)<32;8,4>
+
+mov (16) SRC_YB(2,0)<1> BUF_B(2,0)<32;8,4>
+mov (16) SRC_YB(2,16)<1> BUF_B(2,1)<32;8,4>
+mov (16) SRC_YB(3,0)<1> BUF_B(2,2)<32;8,4>
+mov (16) SRC_YB(3,16)<1> BUF_B(2,3)<32;8,4>
+
+mov (16) SRC_YB(4,0)<1> BUF_B(4,0)<32;8,4>
+mov (16) SRC_YB(4,16)<1> BUF_B(4,1)<32;8,4>
+mov (16) SRC_YB(5,0)<1> BUF_B(4,2)<32;8,4>
+mov (16) SRC_YB(5,16)<1> BUF_B(4,3)<32;8,4>
+
+mov (16) SRC_YB(6,0)<1> BUF_B(6,0)<32;8,4>
+mov (16) SRC_YB(6,16)<1> BUF_B(6,1)<32;8,4>
+mov (16) SRC_YB(7,0)<1> BUF_B(6,2)<32;8,4>
+mov (16) SRC_YB(7,16)<1> BUF_B(6,3)<32;8,4>
+
+// Y is transposed.
+
+
+//////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+// Src U and V are mixed in NV12 format. U on even bytes, V on odd bytes.
+// Transpose by treating UV pair as a word.
+
+
+// Src U 8x8 and V 8x8 are mixed. (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |17 17 16 16 15 15 14 14 13 13 12 12 11 11 10 10 07 07 06 06 05 05 04 04 03 03 02 02 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |37 37 36 36 35 35 34 34 33 33 32 32 31 31 30 30 27 27 26 26 25 25 24 24 23 23 22 22 21 21 20 20|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |57 57 56 56 55 55 54 54 53 53 52 52 51 51 50 50 47 47 46 46 45 45 44 44 43 43 42 42 41 41 40 40|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 76 76 75 75 74 74 73 73 72 72 71 71 70 70 67 67 66 66 65 65 64 64 63 63 62 62 61 61 60 60|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// First step (16) <1>:w <==== <8;4,1>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |33 33 32 32 31 31 30 30 23 23 22 22 21 21 20 20 13 13 12 12 11 11 10 10 03 03 02 02 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 73 72 72 71 71 70 70 63 63 62 62 61 61 60 60 53 53 52 52 51 51 50 50 43 43 42 42 41 41 40 40|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |37 37 36 36 35 35 34 34 27 27 26 26 25 25 24 24 17 17 16 16 15 15 14 14 07 07 06 06 05 05 04 04|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 76 76 75 75 74 74 67 67 66 66 65 65 64 64 57 57 56 56 55 55 54 54 47 47 46 46 45 45 44 44|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// Transpose UV (8x8 words), The first step
+mov (16) BUF_W(0,0)<1> SRC_UW(0,0)<8;4,1>
+mov (16) BUF_W(1,0)<1> SRC_UW(2,0)<8;4,1>
+mov (16) BUF_W(2,0)<1> SRC_UW(0,4)<8;4,1>
+mov (16) BUF_W(3,0)<1> SRC_UW(2,4)<8;4,1>
+
+
+// Second step (16) <1>:w <=== <16;4,4>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 73 63 63 53 53 43 43 33 33 23 23 13 13 03 03 72 72 62 62 52 52 42 42 32 32 22 22 12 12 02 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |75 75 65 65 55 55 45 45 35 35 25 25 15 15 05 05 74 74 64 64 54 54 44 44 34 34 24 24 14 14 04 04|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 67 67 57 57 47 47 37 37 27 27 17 17 07 07 76 76 66 66 56 56 46 46 36 36 26 26 16 16 06 06|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// Transpose UV (8x8 words), The second step
+mov (8) SRC_UW(0,0)<1> BUF_W(0,0)<16;4,4>
+mov (8) SRC_UW(0,8)<1> BUF_W(0,1)<16;4,4>
+mov (8) SRC_UW(1,0)<1> BUF_W(0,2)<16;4,4>
+mov (8) SRC_UW(1,8)<1> BUF_W(0,3)<16;4,4>
+mov (8) SRC_UW(2,0)<1> BUF_W(2,0)<16;4,4>
+mov (8) SRC_UW(2,8)<1> BUF_W(2,1)<16;4,4>
+mov (8) SRC_UW(3,0)<1> BUF_W(2,2)<16;4,4>
+mov (8) SRC_UW(3,8)<1> BUF_W(2,3)<16;4,4>
+
+// U and V are now transposed and separated.
diff --git a/shaders/h264/ildb/TransposeNV12_4x16.asm b/shaders/h264/ildb/TransposeNV12_4x16.asm new file mode 100644 index 0000000..cb1dcbc --- /dev/null +++ b/shaders/h264/ildb/TransposeNV12_4x16.asm @@ -0,0 +1,94 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////
+// Module name: TransposeNV12_4x16.asm
+//
+// Transpose a 4x16 internal planar to 16x4 internal planar block
+//
+//----------------------------------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region is :ub
+// SRC_YB: SRC_YB Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+// SRC_UW: SRC_UB Base=rxx ElementSize=2 SrcRegion=REGION(8,1) Type=uw // 4 GRFs
+//
+// Temp buffer:
+// BUF_B: BUF_B Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+// BUF_W: BUF_W Base=rxx ElementSize=2 SrcRegion=REGION(8,1) Type=uw // 4 GRFs
+//
+//////////////////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDDB:w
+#endif
+
+// Transpose Y (4x16) right most 4 columns
+
+// The first step
+mov (16) BUF_B(0,0)<1> SRC_YB(0,0)<16;4,1> // Read 2 rows, write 1 row
+mov (16) BUF_B(0,16)<1> SRC_YB(2,0)<16;4,1>
+mov (16) BUF_B(1,0)<1> SRC_YB(4,0)<16;4,1>
+mov (16) BUF_B(1,16)<1> SRC_YB(6,0)<16;4,1>
+
+// The second step
+mov (16) BUF_B(2,0)<1> BUF_B(0,0)<32;8,4> // Read 2 rows, write 1 row
+mov (16) BUF_B(2,16)<1> BUF_B(0,1)<32;8,4>
+mov (16) BUF_B(3,0)<1> BUF_B(0,2)<32;8,4>
+mov (16) BUF_B(3,16)<1> BUF_B(0,3)<32;8,4>
+
+// Y is now transposed. the result is in BUF_B(2) and BUF_B(3).
+
+
+
+// Transpose UV (4x8), right most 2 columns in word
+// Use BUF_W(0) as temp buf
+
+// Src U 8x8 and V 8x8 are mixed. (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |17 17 16 16 15 15 14 14 13 13 12 12 11 11 10 10 07 07 06 06 05 05 04 04 03 03 02 02 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |37 37 36 36 35 35 34 34 33 33 32 32 31 31 30 30 27 27 26 26 25 25 24 24 23 23 22 22 21 21 20 20|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |57 57 56 56 55 55 54 54 53 53 52 52 51 51 50 50 47 47 46 46 45 45 44 44 43 43 42 42 41 41 40 40|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 76 76 75 75 74 74 73 73 72 72 71 71 70 70 67 67 66 66 65 65 64 64 63 63 62 62 61 61 60 60|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// First step (8) <1>:w <==== <8;2,1>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 70 70 61 61 60 60 51 51 50 50 41 41 40 40 31 31 30 30 21 21 20 20 11 11 10 10 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+mov (8) BUF_W(0,0)<1> SRC_UW(0,0)<8;2,1>
+mov (8) BUF_W(0,8)<1> SRC_UW(2,0)<8;2,1>
+
+// Second step (16) <1>:w <==== <1;8,2>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+mov (16) BUF_W(1,0)<1> BUF_W(0,0)<1;8,2>
+
+// UV are now transposed. the result is in BUF_W(1).
+
+
+
+//The first step
+//mov (16) BUF_B(0,0)<1> SRC_UW(0,0)<8;2,1> // Read 2 rows, write 1 row
+// The second step
+//mov (8) SRC_UB(4,0)<1> BUF_B(0,0)<16;8,2> // Read 1 row, write 1 row
+//mov (8) SRC_UB(4,8)<1> BUF_B(0,1)<16;8,2> // Read 1 row, write 1 row
+
+// Transpose V (8x8), right most 2 columns
+// The first step
+//mov (16) BUF_B(0,0)<1> SRC_VB(0,1)<8;2,1> // Read 2 rows, write 1 row
+// The second step
+//mov (8) SRC_UB(4,16)<1> BUF_B(0,0)<16;8,2> // Read 1 row, write 1 row
+//mov (8) SRC_UB(4,24)<1> BUF_B(0,1)<16;8,2> // Read 1 row, write 1 row
+
+// U and V are now transposed. the result is in BUF_B(4).
+
diff --git a/shaders/h264/ildb/Transpose_Cur_UV_2x8.asm b/shaders/h264/ildb/Transpose_Cur_UV_2x8.asm new file mode 100644 index 0000000..967e587 --- /dev/null +++ b/shaders/h264/ildb/Transpose_Cur_UV_2x8.asm @@ -0,0 +1,56 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////
+// Module name: Transpose_UV_2x8.asm
+//
+// Transpose UV 2x8 to 8x2 block (2x8U + 2x8V in NV12)
+//
+//----------------------------------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region is :ub
+// SRC_UW: SRC_UB Base=rxx ElementSize=2 SrcRegion=REGION(8,1) Type=uw // 4 GRFs
+//
+// Temp buffer:
+// BUF_W: BUF_W Base=rxx ElementSize=2 SrcRegion=REGION(8,1) Type=uw // 4 GRFs
+//
+//////////////////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDDB:w
+#endif
+
+// Transpose UV (4x8), right most 2 columns in word
+// Use BUF_W(0) as temp buf
+
+// Src U 8x8 and V 8x8 are mixed. (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |17 17 16 16 15 15 14 14 13 13 12 12 11 11 10 10 07 07 06 06 05 05 04 04 03 03 02 02 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |37 37 36 36 35 35 34 34 33 33 32 32 31 31 30 30 27 27 26 26 25 25 24 24 23 23 22 22 21 21 20 20|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |57 57 56 56 55 55 54 54 53 53 52 52 51 51 50 50 47 47 46 46 45 45 44 44 43 43 42 42 41 41 40 40|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 76 76 75 75 74 74 73 73 72 72 71 71 70 70 67 67 66 66 65 65 64 64 63 63 62 62 61 61 60 60|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// First step (8) <1>:w <==== <8;2,1>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 76 76 67 67 66 66 57 57 56 56 47 47 46 46 37 37 36 36 27 27 26 26 17 17 16 16 07 07 06 06|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+mov (8) LEFT_TEMP_W(0,0)<1> SRC_UW(0,6)<8;2,1> { NoDDClr }
+mov (8) LEFT_TEMP_W(0,8)<1> SRC_UW(2,6)<8;2,1> { NoDDChk }
+
+// Second step (16) <1>:w <==== <1;8,2>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 67 67 57 57 47 47 37 37 27 27 17 17 07 07 76 76 66 66 56 56 46 46 36 36 26 26 16 16 06 06|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+mov (16) LEFT_TEMP_W(1,0)<1> LEFT_TEMP_W(0,0)<1;8,2>
+
+// UV are now transposed. the result is in BUF_W(1)
diff --git a/shaders/h264/ildb/Transpose_Cur_UV_8x8.asm b/shaders/h264/ildb/Transpose_Cur_UV_8x8.asm new file mode 100644 index 0000000..dbb7e65 --- /dev/null +++ b/shaders/h264/ildb/Transpose_Cur_UV_8x8.asm @@ -0,0 +1,85 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////
+// Module name: Transpose_UV_8x8.asm
+//
+// Transpose a 8x8 UV block. (8x8U + 8x8V) The output is also in NV12
+//
+//----------------------------------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region is :ub
+// SRC_UW: SRC_UW Base=rxx ElementSize=2 SrcRegion=REGION(8,1) Type=uw // 4 GRFs
+//
+// Temp buffer:
+// BUF_W: BUF_W Base=rxx ElementSize=2 SrcRegion=REGION(8,1) Type=uw // 4 GRFs
+//
+//////////////////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDDA:w
+#endif
+
+//////////////////////////////////////////////////////////////////////////////////////////////////////////
+
+// Src U and V are mixed in NV12 format. U on even bytes, V on odd bytes.
+// Transpose by treating UV pair as a word.
+
+
+// Src U 8x8 and V 8x8 are mixed. (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |17 17 16 16 15 15 14 14 13 13 12 12 11 11 10 10 07 07 06 06 05 05 04 04 03 03 02 02 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |37 37 36 36 35 35 34 34 33 33 32 32 31 31 30 30 27 27 26 26 25 25 24 24 23 23 22 22 21 21 20 20|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |57 57 56 56 55 55 54 54 53 53 52 52 51 51 50 50 47 47 46 46 45 45 44 44 43 43 42 42 41 41 40 40|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 76 76 75 75 74 74 73 73 72 72 71 71 70 70 67 67 66 66 65 65 64 64 63 63 62 62 61 61 60 60|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// First step (16) <1>:w <==== <8;4,1>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |33 33 32 32 31 31 30 30 23 23 22 22 21 21 20 20 13 13 12 12 11 11 10 10 03 03 02 02 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 73 72 72 71 71 70 70 63 63 62 62 61 61 60 60 53 53 52 52 51 51 50 50 43 43 42 42 41 41 40 40|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |37 37 36 36 35 35 34 34 27 27 26 26 25 25 24 24 17 17 16 16 15 15 14 14 07 07 06 06 05 05 04 04|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 76 76 75 75 74 74 67 67 66 66 65 65 64 64 57 57 56 56 55 55 54 54 47 47 46 46 45 45 44 44|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// Transpose UV (8x8 words), The first step
+mov (16) CUR_TEMP_W(0,0)<1> SRC_UW(0,0)<8;4,1>
+mov (16) CUR_TEMP_W(1,0)<1> SRC_UW(2,0)<8;4,1>
+mov (16) CUR_TEMP_W(2,0)<1> SRC_UW(0,4)<8;4,1>
+mov (16) CUR_TEMP_W(3,0)<1> SRC_UW(2,4)<8;4,1>
+
+
+// Second step (16) <1>:w <=== <16;4,4>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 73 63 63 53 53 43 43 33 33 23 23 13 13 03 03 72 72 62 62 52 52 42 42 32 32 22 22 12 12 02 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |75 75 65 65 55 55 45 45 35 35 25 25 15 15 05 05 74 74 64 64 54 54 44 44 34 34 24 24 14 14 04 04|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |77 77 67 67 57 57 47 47 37 37 27 27 17 17 07 07 76 76 66 66 56 56 46 46 36 36 26 26 16 16 06 06|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// Transpose UV (8x8 words), The second step
+mov (8) SRC_UW(0,0)<1> CUR_TEMP_W(0,0)<16;4,4> { NoDDClr }
+mov (8) SRC_UW(0,8)<1> CUR_TEMP_W(0,1)<16;4,4> { NoDDChk }
+mov (8) SRC_UW(1,0)<1> CUR_TEMP_W(0,2)<16;4,4> { NoDDClr }
+mov (8) SRC_UW(1,8)<1> CUR_TEMP_W(0,3)<16;4,4> { NoDDChk }
+mov (8) SRC_UW(2,0)<1> CUR_TEMP_W(2,0)<16;4,4> { NoDDClr }
+mov (8) SRC_UW(2,8)<1> CUR_TEMP_W(2,1)<16;4,4> { NoDDChk }
+mov (8) SRC_UW(3,0)<1> CUR_TEMP_W(2,2)<16;4,4> { NoDDClr }
+mov (8) SRC_UW(3,8)<1> CUR_TEMP_W(2,3)<16;4,4> { NoDDChk }
+
+// U and V are now transposed and separated.
diff --git a/shaders/h264/ildb/Transpose_Cur_UV_Right_Most_2x8.asm b/shaders/h264/ildb/Transpose_Cur_UV_Right_Most_2x8.asm new file mode 100644 index 0000000..be7feba --- /dev/null +++ b/shaders/h264/ildb/Transpose_Cur_UV_Right_Most_2x8.asm @@ -0,0 +1,25 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Transpose Cur MB Right Most 2x8 to 8x2
+// Assume source is LEFT_TEMP_W(0), and detination is LEFT_TEMP_W(1)
+
+// Input from dport for transpose:
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 70 70 61 61 60 60 51 51 50 50 41 41 40 40 31 31 30 30 21 21 20 20 11 11 10 10 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// Output of transpose: <1> <=== <16;8,2>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// mov (8) LEFT_TEMP_W(1,0)<1> LEFT_TEMP_W(0,0)<16;8,2> { NoDDClr }
+// mov (8) LEFT_TEMP_W(1,8)<1> LEFT_TEMP_W(0,1)<16;8,2> { NoDDChk }
+
+ mov (16) LEFT_TEMP_W(1,0)<1> LEFT_TEMP_W(0,0)<1;8,2>
diff --git a/shaders/h264/ildb/Transpose_Cur_Y_16x16.asm b/shaders/h264/ildb/Transpose_Cur_Y_16x16.asm new file mode 100644 index 0000000..8c20f74 --- /dev/null +++ b/shaders/h264/ildb/Transpose_Cur_Y_16x16.asm @@ -0,0 +1,74 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////
+// Module name: Transpose_Y_16x16.asm
+//
+// Transpose Y 16x16 block.
+//
+//----------------------------------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region is :ub
+// SRC_YB: SRC_YB Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+//
+// Temp buffer:
+// CUR_TEMP_B: BUF_B Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+//
+//////////////////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDDA:w
+#endif
+
+
+// Transpose Y (16x16 bytes)
+
+// The first step
+mov (16) CUR_TEMP_B(0,0)<1> SRC_YB(0,0)<16;4,1> { NoDDClr }
+mov (16) CUR_TEMP_B(0,16)<1> SRC_YB(2,0)<16;4,1> { NoDDChk }
+mov (16) CUR_TEMP_B(1,0)<1> SRC_YB(4,0)<16;4,1> { NoDDClr }
+mov (16) CUR_TEMP_B(1,16)<1> SRC_YB(6,0)<16;4,1> { NoDDChk }
+
+mov (16) CUR_TEMP_B(2,0)<1> SRC_YB(0,4)<16;4,1> { NoDDClr }
+mov (16) CUR_TEMP_B(2,16)<1> SRC_YB(2,4)<16;4,1> { NoDDChk }
+mov (16) CUR_TEMP_B(3,0)<1> SRC_YB(4,4)<16;4,1> { NoDDClr }
+mov (16) CUR_TEMP_B(3,16)<1> SRC_YB(6,4)<16;4,1> { NoDDChk }
+
+mov (16) CUR_TEMP_B(4,0)<1> SRC_YB(0,8)<16;4,1> { NoDDClr }
+mov (16) CUR_TEMP_B(4,16)<1> SRC_YB(2,8)<16;4,1> { NoDDChk }
+mov (16) CUR_TEMP_B(5,0)<1> SRC_YB(4,8)<16;4,1> { NoDDClr }
+mov (16) CUR_TEMP_B(5,16)<1> SRC_YB(6,8)<16;4,1> { NoDDChk }
+
+mov (16) CUR_TEMP_B(6,0)<1> SRC_YB(0,12)<16;4,1> { NoDDClr }
+mov (16) CUR_TEMP_B(6,16)<1> SRC_YB(2,12)<16;4,1> { NoDDChk }
+mov (16) CUR_TEMP_B(7,0)<1> SRC_YB(4,12)<16;4,1> { NoDDClr }
+mov (16) CUR_TEMP_B(7,16)<1> SRC_YB(6,12)<16;4,1> { NoDDChk }
+
+// The second step
+mov (16) SRC_YB(0,0)<1> CUR_TEMP_B(0,0)<32;8,4> { NoDDClr }
+mov (16) SRC_YB(0,16)<1> CUR_TEMP_B(0,1)<32;8,4> { NoDDChk }
+mov (16) SRC_YB(1,0)<1> CUR_TEMP_B(0,2)<32;8,4> { NoDDClr }
+mov (16) SRC_YB(1,16)<1> CUR_TEMP_B(0,3)<32;8,4> { NoDDChk }
+
+mov (16) SRC_YB(2,0)<1> CUR_TEMP_B(2,0)<32;8,4> { NoDDClr }
+mov (16) SRC_YB(2,16)<1> CUR_TEMP_B(2,1)<32;8,4> { NoDDChk }
+mov (16) SRC_YB(3,0)<1> CUR_TEMP_B(2,2)<32;8,4> { NoDDClr }
+mov (16) SRC_YB(3,16)<1> CUR_TEMP_B(2,3)<32;8,4> { NoDDChk }
+
+mov (16) SRC_YB(4,0)<1> CUR_TEMP_B(4,0)<32;8,4> { NoDDClr }
+mov (16) SRC_YB(4,16)<1> CUR_TEMP_B(4,1)<32;8,4> { NoDDChk }
+mov (16) SRC_YB(5,0)<1> CUR_TEMP_B(4,2)<32;8,4> { NoDDClr }
+mov (16) SRC_YB(5,16)<1> CUR_TEMP_B(4,3)<32;8,4> { NoDDChk }
+
+mov (16) SRC_YB(6,0)<1> CUR_TEMP_B(6,0)<32;8,4> { NoDDClr }
+mov (16) SRC_YB(6,16)<1> CUR_TEMP_B(6,1)<32;8,4> { NoDDChk }
+mov (16) SRC_YB(7,0)<1> CUR_TEMP_B(6,2)<32;8,4> { NoDDClr }
+mov (16) SRC_YB(7,16)<1> CUR_TEMP_B(6,3)<32;8,4> { NoDDChk }
+
+// Y is transposed.
diff --git a/shaders/h264/ildb/Transpose_Cur_Y_4x16.asm b/shaders/h264/ildb/Transpose_Cur_Y_4x16.asm new file mode 100644 index 0000000..70c0b1c --- /dev/null +++ b/shaders/h264/ildb/Transpose_Cur_Y_4x16.asm @@ -0,0 +1,75 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//////////////////////////////////////////////////////////////////////////////////////////
+// Module name: Transpose_Y_4x16.asm
+//
+// Transpose a 4x16 internal planar to 16x4 internal planar block.
+// The src block is 16x16. Right moft 4 columns are transposed.
+//
+//----------------------------------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region is :ub
+// SRC_YB: SRC_YB Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+//
+// Temp buffer:
+// BUF_B: BUF_B Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+//
+//////////////////////////////////////////////////////////////////////////////////////////
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDDB:w
+#endif
+
+// Transpose Y (4x16) right most 4 columns
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |1f 1e 1d 1c 1b 1a 19 18 17 16 15 14 13 12 11 10 0f 0e 0d 0c 0b 0a 09 08 07 06 05 04 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |3f 3e 3d 3c 3b 3a 39 38 37 36 35 34 33 32 31 30 2f 2e 2d 2c 2b 2a 29 28 27 26 25 24 23 22 21 20|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |5f 5e 5d 5c 5b 5a 59 58 57 56 55 54 53 52 51 50 4f 4e 4d 4c 4b 4a 49 48 47 46 45 44 43 42 41 40|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |7f 7e 7d 7c 7b 7a 79 78 77 76 75 74 73 72 71 70 6f 6e 6d 6c 6b 6a 69 68 67 66 65 64 63 62 61 60|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |9f 9e 9d 9c 9b 9a 99 98 97 96 95 94 93 92 91 90 8f 8e 8d 8c 8b 8a 89 88 87 86 85 84 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |bf be bd bc bb ba b9 b8 b7 b6 b5 b4 b3 b2 b1 b0 af ae ad ac ab aa a9 a8 a7 a6 a5 a4 a3 a2 a1 a0|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |df de dd dc db da d9 d8 d7 d6 d5 d4 d3 d2 d1 d0 cf ce cd cc cb ca c9 c8 c7 c6 c5 c4 c3 c2 c1 c0|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |ff fe fd fc fb fa f9 f8 f7 f6 f5 f4 f3 f2 f1 f0 ef ee ed ec eb ea e9 e8 e7 e6 e5 e4 e3 e2 e1 e0|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// The first step
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |7f 7e 7d 7c 6f 6e 6d 6c 5f 5e 5d 5c 4f 4e 4d 4c 3f 3e 3d 3c 2f 2e 2d 2c 1f 1e 1d 1c 0f 0e 0d 0c|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |ff fe fd fc ef ee ed ec df de dd dc cf ce cd cc bf be bd bc af ae ad ac 9f 9e 9d 9c 8f 8e 8d 8c|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// The second step
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |fd ed dd cd bd ad 9d 8d 7d 6d 5d 4d 3d 2d 1d 0d fc ec dc cc bc ac 9c 8c 7c 6c 5c 4c 3c 2c 1c 0c|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |ff ef df cf bf af 9f 8f 7f 6f 5f 4f 3f 2f 1f 0f fe ee de ce be ae 9e 8e 7e 6e 5e 4e 3e 2e 1e 0e|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+
+mov (16) LEFT_TEMP_B(0,0)<1> SRC_YB(0,12)<16;4,1> { NoDDClr }
+mov (16) LEFT_TEMP_B(0,16)<1> SRC_YB(2,12)<16;4,1> { NoDDChk }
+mov (16) LEFT_TEMP_B(1,0)<1> SRC_YB(4,12)<16;4,1> { NoDDClr }
+mov (16) LEFT_TEMP_B(1,16)<1> SRC_YB(6,12)<16;4,1> { NoDDChk }
+
+// The second step
+mov (16) LEFT_TEMP_B(2,0)<1> LEFT_TEMP_B(0,0)<32;8,4> { NoDDClr }
+mov (16) LEFT_TEMP_B(2,16)<1> LEFT_TEMP_B(0,1)<32;8,4> { NoDDChk }
+mov (16) LEFT_TEMP_B(3,0)<1> LEFT_TEMP_B(0,2)<32;8,4> { NoDDClr }
+mov (16) LEFT_TEMP_B(3,16)<1> LEFT_TEMP_B(0,3)<32;8,4> { NoDDChk }
+
+// Y is now transposed. the result is in LEFT_TEMP_B(2) and LEFT_TEMP_B(3).
diff --git a/shaders/h264/ildb/Transpose_Cur_Y_Right_Most_4x16.asm b/shaders/h264/ildb/Transpose_Cur_Y_Right_Most_4x16.asm new file mode 100644 index 0000000..c458f85 --- /dev/null +++ b/shaders/h264/ildb/Transpose_Cur_Y_Right_Most_4x16.asm @@ -0,0 +1,31 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Transpose cur Y right most 4x16 to 16x4
+// Assume source is LEFT_TEMP_B(0), and detination is LEFT_TEMP_B(2)
+
+
+// Input received from dport:
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 72 71 70 63 62 61 60 53 52 51 50 43 42 41 40 33 32 31 30 23 22 21 20 13 12 11 10 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 f2 f1 f0 e3 e2 e1 e0 d3 d2 d1 d0 c3 c2 c1 c0 b3 b2 b1 b0 a3 a2 a1 a0 93 92 91 90 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// Output of transpose: <1> <= <32;8,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f1 e1 d1 c1 b1 a1 91 81 71 61 51 41 31 21 11 01 f0 e0 d0 c0 b0 a0 90 80 70 60 50 40 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 b3 a3 93 83 73 63 53 43 33 23 13 03 f2 e2 d2 c2 b2 a2 92 82 72 62 52 42 32 22 12 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+ // Transpose the data, also occupy 2 GRFs
+ mov (16) LEFT_TEMP_B(2)<1> LEFT_TEMP_B(0, 0)<32;8,4> { NoDDClr }
+ mov (16) LEFT_TEMP_B(2, 16)<1> LEFT_TEMP_B(0, 1)<32;8,4> { NoDDChk }
+ mov (16) LEFT_TEMP_B(3)<1> LEFT_TEMP_B(0, 2)<32;8,4> { NoDDClr }
+ mov (16) LEFT_TEMP_B(3, 16)<1> LEFT_TEMP_B(0, 3)<32;8,4> { NoDDChk }
diff --git a/shaders/h264/ildb/Transpose_Left_UV_2x8.asm b/shaders/h264/ildb/Transpose_Left_UV_2x8.asm new file mode 100644 index 0000000..678456e --- /dev/null +++ b/shaders/h264/ildb/Transpose_Left_UV_2x8.asm @@ -0,0 +1,28 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Transpose left MB 2x8 to 8x2
+// Assume source is LEFT_TEMP_W, and detination is PREV_MB_UW
+
+// Input from dport for transpose:
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 70 70 61 61 60 60 51 51 50 50 41 41 40 40 31 31 30 30 21 21 20 20 11 11 10 10 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// Output of transpose: <1> <=== <16;8,2>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// mov (8) PREV_MB_UW(0,0)<1> BUF_W(0,0)<16;8,2> { NoDDClr }
+// mov (8) PREV_MB_UW(0,8)<1> BUF_W(0,1)<16;8,2> { NoDDChk }
+
+// mov (8) PREV_MB_UW(0,0)<1> LEFT_TEMP_W(0,0)<16;8,2> { NoDDClr }
+// mov (8) PREV_MB_UW(0,8)<1> LEFT_TEMP_W(0,1)<16;8,2> { NoDDChk }
+
+ mov (16) PREV_MB_UW(0,0)<1> LEFT_TEMP_W(0,0)<1;8,2>
diff --git a/shaders/h264/ildb/Transpose_Left_Y_4x16.asm b/shaders/h264/ildb/Transpose_Left_Y_4x16.asm new file mode 100644 index 0000000..435996c --- /dev/null +++ b/shaders/h264/ildb/Transpose_Left_Y_4x16.asm @@ -0,0 +1,31 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Transpose left MB 4x16 to 16x4
+// Assume source is LEFT_TEMP_B, and detination is PREV_MB_YB
+
+
+// Input received from dport:
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 72 71 70 63 62 61 60 53 52 51 50 43 42 41 40 33 32 31 30 23 22 21 20 13 12 11 10 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 f2 f1 f0 e3 e2 e1 e0 d3 d2 d1 d0 c3 c2 c1 c0 b3 b2 b1 b0 a3 a2 a1 a0 93 92 91 90 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// Output of transpose: <1> <= <32;8,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f1 e1 d1 c1 b1 a1 91 81 71 61 51 41 31 21 11 01 f0 e0 d0 c0 b0 a0 90 80 70 60 50 40 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 b3 a3 93 83 73 63 53 43 33 23 13 03 f2 e2 d2 c2 b2 a2 92 82 72 62 52 42 32 22 12 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+ // Transpose the data, also occupy 2 GRFs
+ mov (16) PREV_MB_YB(0)<1> LEFT_TEMP_B(0, 0)<32;8,4> { NoDDClr }
+ mov (16) PREV_MB_YB(0, 16)<1> LEFT_TEMP_B(0, 1)<32;8,4> { NoDDChk }
+ mov (16) PREV_MB_YB(1)<1> LEFT_TEMP_B(0, 2)<32;8,4> { NoDDClr }
+ mov (16) PREV_MB_YB(1, 16)<1> LEFT_TEMP_B(0, 3)<32;8,4> { NoDDChk }
diff --git a/shaders/h264/ildb/loadNV12_16x16T.asm b/shaders/h264/ildb/loadNV12_16x16T.asm new file mode 100644 index 0000000..d5aa552 --- /dev/null +++ b/shaders/h264/ildb/loadNV12_16x16T.asm @@ -0,0 +1,53 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: loadNV12_16x16T.asm
+//
+// Load and transpose NV12 16x16 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud (U+V for NV12) // 4 GRFs
+//
+// Source region is :ub. The same region as :ud region
+// SRC_YB: SRC_YB Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8 GRFs
+// SRC_UB: SRC_UB Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 2 GRFs
+// SRC_VB: SRC_VB Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 2 GRFs
+//
+// Binding table index:
+// BI_SRC_Y: Binding table index of Y surface
+// BI_SRC_UV: Binding table index of UV surface (NV12)
+//
+// Temp buffer:
+// BUF_B: BUF_B Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD1:w
+#endif
+
+ // Read Y
+ mov (2) MSGSRC.0<1>:ud ORIX_CUR<2;2,1>:w // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud // Block width and height (16x16)
+ send (8) SRC_YD(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(8)+DWBRMSGDSC_RC+BI_SRC_Y // Read 8 GRFs
+
+ // Read U+V
+ asr (1) MSGSRC.1:ud MSGSRC.1:ud 1:w // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2<1>:ud 0x0007000F:ud // NV12 U+V block width and height (16x8)
+ send (8) SRC_UD(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(4)+DWBRMSGDSC_RC+BI_SRC_UV // Read 4 GRFs
+
+ #include "TransposeNV12_16x16.asm"
+
+// #include "Transpose_Y_16x16.asm"
+// #include "Transpose_NV12_UV_16x8.asm"
+
+// End of loadNV12_16x16T
diff --git a/shaders/h264/ildb/loadNV12_16x4.asm b/shaders/h264/ildb/loadNV12_16x4.asm new file mode 100644 index 0000000..a2e7dfd --- /dev/null +++ b/shaders/h264/ildb/loadNV12_16x4.asm @@ -0,0 +1,54 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module Name: Loadnv12_16X4.Asm
+//
+// Load Nv12 16X4 Block
+//
+//----------------------------------------------------------------
+// Symbols Need To Be Defined Before Including This Module
+//
+// Source Region In :Ud
+// Src_Yd: Src_Yd Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=Ud // 3 Grfs (2 For Y, 1 For U+V)
+//
+// Source Region Is :Ub. The Same Region As :Ud Region
+// Src_Yb: Src_Yb Base=Rxx Elementsize=1 Srcregion=Region(16,1) Type=Ub // 2 Grfs
+// Src_Ub: Src_Ub Base=Rxx Elementsize=1 Srcregion=Region(16,1) Type=Ub // 0.5 Grf
+// Src_Vb: Src_Vb Base=Rxx Elementsize=1 Srcregion=Region(16,1) Type=Ub // 0.5 Grf
+//
+// Binding Table Index:
+// Bi_Src_Y: Binding Table Index Of Y Surface
+// Bi_Src_UV: Binding Table Index Of UV Surface (Nv12)
+//
+// Temp Buffer:
+// Buf_D: Buf_D Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=Ud
+// Buf_B: Buf_B Base=Rxx Elementsize=1 Srcregion=Region(16,1) Type=Ub
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD2:w
+#endif
+
+ // Read Y
+ mov (2) MSGSRC.0<1>:ud ORIX<2;2,1>:w // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x0003000F:ud // Block width and height (16x4)
+ send (8) PREV_MB_YD(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(2)+DWBRMSGDSC_RC+BI_SRC_Y // Read 2 GRFs
+
+ // Read U+V
+ asr (1) MSGSRC.1:ud MSGSRC.1:ud 1:w // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2<1>:ud 0x0001000F:ud // NV12 U+V block width and height (16x2)
+
+ // Load NV12 U+V tp a temp buf
+ send (8) BUF_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(1)+DWBRMSGDSC_RC+BI_SRC_UV // Read 1 GRF
+
+ // Convert NV12 U+V to internal planar U and V and place them right after Y.
+// mov (16) SRC_UB(0,0)<1> BUF_B(0,0)<32;16,2>
+// mov (16) SRC_VB(0,0)<1> BUF_B(0,1)<32;16,2>
+
+// End of loadNV12_16x4.asm
diff --git a/shaders/h264/ildb/load_Cur_UV_8x8T.asm b/shaders/h264/ildb/load_Cur_UV_8x8T.asm new file mode 100644 index 0000000..25cb96c --- /dev/null +++ b/shaders/h264/ildb/load_Cur_UV_8x8T.asm @@ -0,0 +1,65 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_Cur_UV_8x8T.asm
+//
+// Load and transpose UV 8x8 block (NV12: 8x8U and 8x8V mixed)
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud (U+V for NV12) // 4 GRFs
+//
+// Binding table index:
+// BI_SRC_UV: Binding table index of UV surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD1:w
+#endif
+
+ // Read U+V blk
+#if defined(_PROGRESSIVE)
+ mov (1) MSGSRC.0:ud ORIX_CUR:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_CUR:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0007000F:ud { NoDDChk } // NV12 U+V block width and height (16x8 bytes)
+
+ //send (8) SRC_UD(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DWBRMSGDSC_SC+0x00040000+BI_SRC_UV
+ mov (1) MSGDSC RESP_LEN(4)+DWBRMSGDSC_SC+BI_SRC_UV:ud
+#endif
+
+#if defined(_FIELD)
+// cmp.z.f0.0 (1) NULLREGW PicTypeC:w 0:w // Get pic type flag
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+ // they are used later in this file
+
+ mov (1) MSGSRC.0:ud ORIX_CUR:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_CUR:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0007000F:ud { NoDDChk } // NV12 U+V block width and height (16x8 bytes)
+
+ // Set message descriptor
+
+ // Frame picture
+// (f0.0) mov (1) MSGDSC RESP_LEN(4)+DWBRMSGDSC_SC+BI_SRC_UV:ud // Read 4 GRFs from SRC_UV
+// (f0.0) jmpi load_UV_8x8T
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(4)+DWBRMSGDSC_SC_BF+BI_SRC_UV:ud // Read 4 GRFs from SRC_UV bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(4)+DWBRMSGDSC_SC_TF+BI_SRC_UV:ud // Read 4 GRFs from SRC_UV top field
+
+//load_UV_8x8T:
+
+#endif
+
+ send (8) SRC_UD(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// #include "Transpose_Cur_UV_8x8.asm"
+
+// End of load_UV_8x8T
diff --git a/shaders/h264/ildb/load_Cur_UV_8x8T_Mbaff.asm b/shaders/h264/ildb/load_Cur_UV_8x8T_Mbaff.asm new file mode 100644 index 0000000..82b7d9e --- /dev/null +++ b/shaders/h264/ildb/load_Cur_UV_8x8T_Mbaff.asm @@ -0,0 +1,62 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_Cur_UV_8x8T.asm
+//
+// Load and transpose UV 8x8 block (NV12: 8x8U and 8x8V mixed)
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud (U+V for NV12) // 4 GRFs
+//
+// Binding table index:
+// BI_SRC_UV: Binding table index of UV surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD1:w
+#endif
+ // FieldModeCurrentMbFlag determines how to access left MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+
+ // Read U+V
+ mov (1) MSGSRC.0:ud ORIX_CUR:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_CUR:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0007000F:ud { NoDDChk } // NV12 U+V block width and height (16x8 bytes)
+
+ // Set message descriptor
+
+ (f0.0) if (1) ILDB_LABEL(ELSE_UV_8X8T)
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(4)+DWBRMSGDSC_SC+BI_SRC_UV:ud // Read 4 GRFs from SRC_UV
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 8:w // Add vertical offset 8 for bot MB in MBAFF mode
+
+ILDB_LABEL(ELSE_UV_8X8T):
+ else (1) ILDB_LABEL(ENDIF_UV_8X8T)
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(4)+DWBRMSGDSC_SC_BF+BI_SRC_UV:ud // Read 4 GRFs from SRC_UV bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(4)+DWBRMSGDSC_SC_TF+BI_SRC_UV:ud // Read 4 GRFs from SRC_UV top field
+
+ asr (1) MSGSRC.1:d MSGSRC.1:d 1:w // Reduce y by half in field access mode
+
+ endif
+ILDB_LABEL(ENDIF_UV_8X8T):
+
+ send (8) SRC_UD(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// #include "Transpose_Cur_UV_8x8.asm"
+
+// End of load_UV_8x8T
diff --git a/shaders/h264/ildb/load_Cur_UV_Right_Most_2x8.asm b/shaders/h264/ildb/load_Cur_UV_Right_Most_2x8.asm new file mode 100644 index 0000000..426a518 --- /dev/null +++ b/shaders/h264/ildb/load_Cur_UV_Right_Most_2x8.asm @@ -0,0 +1,61 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module Name: Load_Cur_UV_Right_Most_2X8.Asm
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD0:w
+#endif
+
+#if defined(_PROGRESSIVE)
+ // Read U+V, (UV MB size = 16x8)
+ add (1) MSGSRC.0:ud ORIX_CUR:w 12:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_CUR:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x00070003:ud { NoDDChk } // NV12 U+V block width and height (4x8)
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(1)+DWBRMSGDSC_RC+BI_DEST_UV
+#endif
+
+#if defined(_FIELD) || defined(_MBAFF)
+
+ // FieldModeCurrentMbFlag determines how to access left MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+
+ // Read U+V
+ add (1) MSGSRC.0:ud ORIX_CUR:w 12:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_CUR:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x00070003:ud { NoDDChk } // NV12 U+V block width and height (4x8)
+
+ // Load NV12 U+V
+
+ // Set message descriptor
+
+ (f0.0) if (1) ILDB_LABEL(ELSE_Y_2x8T)
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC+BI_DEST_UV:ud // Read 1 GRF from SRC_UV
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 8:w // Add vertical offset 8 for bot MB in MBAFF mode
+
+ILDB_LABEL(ELSE_Y_2x8T):
+ else (1) ILDB_LABEL(ENDIF_Y_2x8T)
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_BF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_TF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y top field
+
+ endif
+ILDB_LABEL(ENDIF_Y_2x8T):
+
+ // Read 1 GRF from DEST surface as the above MB has been deblocked.
+// send (8) BUF_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud MSGDSC
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+#endif
+
diff --git a/shaders/h264/ildb/load_Cur_Y_16x16T.asm b/shaders/h264/ildb/load_Cur_Y_16x16T.asm new file mode 100644 index 0000000..d70b101 --- /dev/null +++ b/shaders/h264/ildb/load_Cur_Y_16x16T.asm @@ -0,0 +1,63 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_Y_16x16T.asm
+//
+// Load and transpose Y 16x16 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+//
+// Binding table index:
+// BI_SRC_Y: Binding table index of Y surface
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD1:w
+#endif
+ // Read Y
+
+#if defined(_PROGRESSIVE)
+ mov (2) MSGSRC.0<1>:ud ORIX_CUR<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud { NoDDChk } // Block width and height (16x16)
+
+ //send (8) SRC_YD(0)<1> MSGHDRC MSGSRC<8;8,1>:ud DWBRMSGDSC_SMPLR+0x00080000+BI_SRC_Y
+ mov (1) MSGDSC RESP_LEN(8)+DWBRMSGDSC_SC+BI_SRC_Y:ud
+#endif
+
+
+#if defined(_FIELD)
+// cmp.z.f0.0 (1) NULLREGW PicTypeC:w 0:w // Get pic type flag
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+ // they are used later in this file
+
+ mov (2) MSGSRC.0<1>:ud ORIX_CUR<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud { NoDDChk } // Block width and height (16x16)
+
+ // Set message descriptor
+ // Frame picture
+// (f0.0) mov (1) MSGDSC RESP_LEN(8)+DWBRMSGDSC_SC+BI_SRC_Y:ud // Read 8 GRFs from SRC_Y
+// (f0.0) jmpi load_Y_16x16T
+
+ // Non frame picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(8)+DWBRMSGDSC_SC_BF+BI_SRC_Y:ud // Read 8 GRFs from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(8)+DWBRMSGDSC_SC_TF+BI_SRC_Y:ud // Read 8 GRFs from SRC_Y top field
+
+//load_Y_16x16T:
+
+#endif
+
+ send (8) SRC_YD(0)<1> MSGHDRC MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// #include "Transpose_Cur_Y_16x16.asm"
+
+// End of load_Y_16x16T
diff --git a/shaders/h264/ildb/load_Cur_Y_16x16T_Mbaff.asm b/shaders/h264/ildb/load_Cur_Y_16x16T_Mbaff.asm new file mode 100644 index 0000000..f9c4745 --- /dev/null +++ b/shaders/h264/ildb/load_Cur_Y_16x16T_Mbaff.asm @@ -0,0 +1,62 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_Y_16x16T.asm
+//
+// Load and transpose Y 16x16 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+//
+// Binding table index:
+// BI_SRC_Y: Binding table index of Y surface
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD1:w
+#endif
+
+ // FieldModeCurrentMbFlag determines how to access left MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+
+ // Read Y
+ mov (2) MSGSRC.0<1>:d ORIX_CUR<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud { NoDDChk } // Block width and height (16x16)
+
+ // Set message descriptor, etc.
+
+ (f0.0) if (1) ILDB_LABEL(ELSE_Y_16x16T)
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(8)+DWBRMSGDSC_SC+BI_SRC_Y:ud // Read 8 GRFs from SRC_Y
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 16:w // Add vertical offset 16 for bot MB in MBAFF mode
+
+ILDB_LABEL(ELSE_Y_16x16T):
+ else (1) ILDB_LABEL(ENDIF_Y_16x16T)
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(8)+DWBRMSGDSC_SC_BF+BI_SRC_Y:ud // Read 8 GRFs from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(8)+DWBRMSGDSC_SC_TF+BI_SRC_Y:ud // Read 8 GRFs from SRC_Y top field
+
+ asr (1) MSGSRC.1:d MSGSRC.1:d 1:w // Reduce y by half in field access mode
+
+ endif
+ILDB_LABEL(ENDIF_Y_16x16T):
+
+ send (8) SRC_YD(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// #include "Transpose_Cur_Y_16x16.asm"
+
+// End of load_Y_16x16T
diff --git a/shaders/h264/ildb/load_Cur_Y_Right_Most_4x16.asm b/shaders/h264/ildb/load_Cur_Y_Right_Most_4x16.asm new file mode 100644 index 0000000..cd25ace --- /dev/null +++ b/shaders/h264/ildb/load_Cur_Y_Right_Most_4x16.asm @@ -0,0 +1,85 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_Cur_Y_Right_Most_4x16.asm
+//
+// Load luma cur MB right most 4x16 into LEFT_TEMP_B
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD0:w
+#endif
+
+
+#if defined(_PROGRESSIVE)
+ // Read Y
+ add (1) MSGSRC.0<1>:ud ORIX_CUR:w 12:w { NoDDClr } // Block origin, move right 12 bytes
+ mov (1) MSGSRC.1<1>:ud ORIY_CUR:w { NoDDClr, NoDDChk } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F0003:ud { NoDDChk } // Block width and height (4x16)
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRL MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(2)+DWBRMSGDSC_RC+BI_DEST_Y
+#endif
+
+
+#if defined(_FIELD) || defined(_MBAFF)
+
+ // FieldModeCurrentMbFlag determines how to access left MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+
+ // Read Y
+ add (1) MSGSRC.0<1>:ud ORIX_CUR:w 12:w { NoDDClr } // Block origin, move right 12 bytes
+ mov (1) MSGSRC.1<1>:ud ORIY_CUR:w { NoDDClr, NoDDChk } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F0003:ud { NoDDChk } // Block width and height (4x16)
+
+ // Set message descriptor, etc.
+
+ (f0.0) if (1) ILDB_LABEL(ELSE_Y_4x16T)
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 16:w // Add vertical offset 16 for bot MB in MBAFF mode
+
+ILDB_LABEL(ELSE_Y_4x16T):
+ else (1) ILDB_LABEL(ENDIF_Y_4x16T)
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_BF+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_TF+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y top field
+
+ endif
+ILDB_LABEL(ENDIF_Y_4x16T):
+
+// send (8) BUF_D(0)<1> MSGHDRY MSGSRC<8;8,1>:ud MSGDSC
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRL MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+#endif
+
+// Transpose 4x16 to 16x4
+
+// Input received from dport:
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 72 71 70 63 62 61 60 53 52 51 50 43 42 41 40 33 32 31 30 23 22 21 20 13 12 11 10 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 f2 f1 f0 e3 e2 e1 e0 d3 d2 d1 d0 c3 c2 c1 c0 b3 b2 b1 b0 a3 a2 a1 a0 93 92 91 90 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// Output of transpose: <1> <= <32;8,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f1 e1 d1 c1 b1 a1 91 81 71 61 51 41 31 21 11 01 f0 e0 d0 c0 b0 a0 90 80 70 60 50 40 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 b3 a3 93 83 73 63 53 43 33 23 13 03 f2 e2 d2 c2 b2 a2 92 82 72 62 52 42 32 22 12 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+/*
+ // Transpose the data, also occupy 2 GRFs
+ mov (16) PREV_MB_YB(0)<1> BUF_B(0, 0)<32;8,4> { NoDDClr }
+ mov (16) PREV_MB_YB(0, 16)<1> BUF_B(0, 1)<32;8,4> { NoDDChk }
+ mov (16) PREV_MB_YB(1)<1> BUF_B(0, 2)<32;8,4> { NoDDClr }
+ mov (16) PREV_MB_YB(1, 16)<1> BUF_B(0, 3)<32;8,4> { NoDDChk }
+*/
+// End of load_Y_4x16T
+
diff --git a/shaders/h264/ildb/load_Left_UV_2x8T.asm b/shaders/h264/ildb/load_Left_UV_2x8T.asm new file mode 100644 index 0000000..a5f622c --- /dev/null +++ b/shaders/h264/ildb/load_Left_UV_2x8T.asm @@ -0,0 +1,76 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module Name: Load_Left_UV_2X8T.Asm
+//
+// Load UV 8X2 Block
+//
+//----------------------------------------------------------------
+// Symbols ceed To be defined before including this module
+//
+// Source Region Is :UB
+// BUF_D: BUF_D Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=UD
+
+// Binding Table Index:
+// BI_SRC_UV: Binding Table Index Of UV Surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD0:w
+#endif
+
+#if defined(_PROGRESSIVE)
+ // Read U+V
+ mov (1) MSGSRC.0:ud ORIX_LEFT:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_LEFT:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x00070003:ud { NoDDChk } // NV12 U+V block width and height (4x8)
+
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(1)+DWBRMSGDSC_RC+BI_DEST_UV
+#endif
+
+#if defined(_FIELD) || defined(_MBAFF)
+
+ // FieldModeCurrentMbFlag determines how to access left MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+
+ // Read U+V
+ mov (1) MSGSRC.0:ud ORIX_LEFT:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_LEFT:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x00070003:ud { NoDDChk } // NV12 U+V block width and height (4x8)
+
+ // Load NV12 U+V
+
+ // Set message descriptor
+
+ (f0.0) if (1) ILDB_LABEL(ELSE_Y_2x8T)
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC+BI_DEST_UV:ud // Read 1 GRF from SRC_UV
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 8:w // Add vertical offset 8 for bot MB in MBAFF mode
+
+ILDB_LABEL(ELSE_Y_2x8T):
+ else (1) ILDB_LABEL(ENDIF_Y_2x8T)
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_BF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_TF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y top field
+
+ endif
+ILDB_LABEL(ENDIF_Y_2x8T):
+
+ // Read 1 GRF from DEST surface as the above MB has been deblocked.
+// send (8) BUF_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud MSGDSC
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+#endif
+
+// End of load_Left_UV_2x8T.asm
diff --git a/shaders/h264/ildb/load_Left_UV_2x8T_Mbaff.asm b/shaders/h264/ildb/load_Left_UV_2x8T_Mbaff.asm new file mode 100644 index 0000000..fefda4f --- /dev/null +++ b/shaders/h264/ildb/load_Left_UV_2x8T_Mbaff.asm @@ -0,0 +1,79 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module Name: Load_Left_UV_2X8T.Asm
+//
+// Load UV 8X2 Block
+//
+//----------------------------------------------------------------
+// Symbols ceed To be defined before including this module
+//
+// Source Region Is :UB
+// BUF_D: BUF_D Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=UD
+
+// Binding Table Index:
+// BI_SRC_UV: Binding Table Index Of UV Surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD0:w
+#endif
+
+ // FieldModeCurrentMbFlag determines how to access left MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+
+ // Read U+V
+ mov (1) MSGSRC.0:ud ORIX_LEFT:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_LEFT:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x00070003:ud { NoDDChk } // NV12 U+V block width and height (4x8)
+
+ // Load NV12 U+V
+
+ // Set message descriptor
+
+ (f0.0) if (1) ILDB_LABEL(ELSE_Y_2x8T)
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC+BI_DEST_UV:ud // Read 1 GRF from SRC_UV
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 8:w // Add vertical offset 8 for bot MB in MBAFF mode
+
+ILDB_LABEL(ELSE_Y_2x8T):
+ else (1) ILDB_LABEL(ENDIF_Y_2x8T)
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_BF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_TF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y top field
+
+ asr (1) MSGSRC.1:d MSGSRC.1:d 1:w // Reduce y by half in field access mode
+
+ endif
+ILDB_LABEL(ENDIF_Y_2x8T):
+
+ // Read 1 GRF from DEST surface as the above MB has been deblocked.
+// send (8) BUF_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud MSGDSC
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+
+// Input from dport for transpose:
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 70 70 61 61 60 60 51 51 50 50 41 41 40 40 31 31 30 30 21 21 20 20 11 11 10 10 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// Output of transpose: <1> <=== <16;8,2>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+/*
+ mov (8) PREV_MB_UW(0,0)<1> BUF_W(0,0)<16;8,2> { NoDDClr }
+ mov (8) PREV_MB_UW(0,8)<1> BUF_W(0,1)<16;8,2> { NoDDChk }
+*/
+// End of load_Left_UV_2x8T.asm
diff --git a/shaders/h264/ildb/load_Left_Y_4x16T.asm b/shaders/h264/ildb/load_Left_Y_4x16T.asm new file mode 100644 index 0000000..ab454f1 --- /dev/null +++ b/shaders/h264/ildb/load_Left_Y_4x16T.asm @@ -0,0 +1,96 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_Y_4x16T.asm
+//
+// Load luma left MB 4x16 and transpose 4x16 to 16x4.
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// PREV_MB_YD: PREV_MB_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 2 GRFs
+//
+// Binding table index:
+// BI_SRC_Y: Binding table index of Y surface
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD0:w
+#endif
+
+
+#if defined(_PROGRESSIVE)
+ // Read Y
+ mov (2) MSGSRC.0<1>:ud ORIX_LEFT<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F0003:ud { NoDDChk } // Block width and height (4x16)
+
+// mov (1) MSGDSC DWBRMSGDSC_RC+0x00020000+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRL MSGSRC<8;8,1>:ud DAPREAD RESP_LEN(2)+DWBRMSGDSC_RC+BI_DEST_Y
+#endif
+
+
+#if defined(_FIELD) || defined(_MBAFF)
+
+ // FieldModeCurrentMbFlag determines how to access left MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+
+ // Read Y
+ mov (2) MSGSRC.0<1>:ud ORIX_LEFT<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F0003:ud { NoDDChk } // Block width and height (4x16)
+
+ // Set message descriptor, etc.
+
+ (f0.0) if (1) ILDB_LABEL(ELSE_Y_4x16T)
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 16:w // Add vertical offset 16 for bot MB in MBAFF mode
+
+ILDB_LABEL(ELSE_Y_4x16T):
+ else (1) ILDB_LABEL(ENDIF_Y_4x16T)
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_BF+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_TF+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y top field
+
+ endif
+ILDB_LABEL(ENDIF_Y_4x16T):
+
+// send (8) BUF_D(0)<1> MSGHDRY MSGSRC<8;8,1>:ud MSGDSC
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRL MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+#endif
+
+// Transpose 4x16 to 16x4
+
+// Input received from dport:
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 72 71 70 63 62 61 60 53 52 51 50 43 42 41 40 33 32 31 30 23 22 21 20 13 12 11 10 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 f2 f1 f0 e3 e2 e1 e0 d3 d2 d1 d0 c3 c2 c1 c0 b3 b2 b1 b0 a3 a2 a1 a0 93 92 91 90 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// Output of transpose: <1> <= <32;8,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f1 e1 d1 c1 b1 a1 91 81 71 61 51 41 31 21 11 01 f0 e0 d0 c0 b0 a0 90 80 70 60 50 40 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 b3 a3 93 83 73 63 53 43 33 23 13 03 f2 e2 d2 c2 b2 a2 92 82 72 62 52 42 32 22 12 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+/*
+ // Transpose the data, also occupy 2 GRFs
+ mov (16) PREV_MB_YB(0)<1> BUF_B(0, 0)<32;8,4> { NoDDClr }
+ mov (16) PREV_MB_YB(0, 16)<1> BUF_B(0, 1)<32;8,4> { NoDDChk }
+ mov (16) PREV_MB_YB(1)<1> BUF_B(0, 2)<32;8,4> { NoDDClr }
+ mov (16) PREV_MB_YB(1, 16)<1> BUF_B(0, 3)<32;8,4> { NoDDChk }
+*/
+// End of load_Y_4x16T
+
diff --git a/shaders/h264/ildb/load_Left_Y_4x16T_Mbaff.asm b/shaders/h264/ildb/load_Left_Y_4x16T_Mbaff.asm new file mode 100644 index 0000000..95f73a3 --- /dev/null +++ b/shaders/h264/ildb/load_Left_Y_4x16T_Mbaff.asm @@ -0,0 +1,84 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: load_Y_4x16T.asm
+//
+// Load luma left MB 4x16 and transpose 4x16 to 16x4.
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// PREV_MB_YD: PREV_MB_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 2 GRFs
+//
+// Binding table index:
+// BI_SRC_Y: Binding table index of Y surface
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD0:w
+#endif
+
+ // FieldModeCurrentMbFlag determines how to access left MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+
+ // Read Y
+ mov (2) MSGSRC.0<1>:ud ORIX_LEFT<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F0003:ud { NoDDChk } // Block width and height (4x16)
+
+ // Set message descriptor, etc.
+
+ (f0.0) if (1) ELSE_Y_4x16T
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 16:w // Add vertical offset 16 for bot MB in MBAFF mode
+
+ELSE_Y_4x16T:
+ else (1) ENDIF_Y_4x16T
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_BF+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_TF+BI_DEST_Y:ud // Read 2 GRFs from DEST_Y top field
+
+ asr (1) MSGSRC.1:d MSGSRC.1:d 1:w // Reduce y by half in field access mode
+
+ endif
+ENDIF_Y_4x16T:
+
+// send (8) BUF_D(0)<1> MSGHDRY MSGSRC<8;8,1>:ud MSGDSC
+ send (8) LEFT_TEMP_D(0)<1> MSGHDRL MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+
+// Transpose 4x16 to 16x4
+
+// Input received from dport:
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 72 71 70 63 62 61 60 53 52 51 50 43 42 41 40 33 32 31 30 23 22 21 20 13 12 11 10 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 f2 f1 f0 e3 e2 e1 e0 d3 d2 d1 d0 c3 c2 c1 c0 b3 b2 b1 b0 a3 a2 a1 a0 93 92 91 90 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+// Output of transpose: <1> <= <32;8,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f1 e1 d1 c1 b1 a1 91 81 71 61 51 41 31 21 11 01 f0 e0 d0 c0 b0 a0 90 80 70 60 50 40 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 b3 a3 93 83 73 63 53 43 33 23 13 03 f2 e2 d2 c2 b2 a2 92 82 72 62 52 42 32 22 12 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+/*
+ // Transpose the data, also occupy 2 GRFs
+ mov (16) PREV_MB_YB(0)<1> BUF_B(0, 0)<32;8,4> { NoDDClr }
+ mov (16) PREV_MB_YB(0, 16)<1> BUF_B(0, 1)<32;8,4> { NoDDChk }
+ mov (16) PREV_MB_YB(1)<1> BUF_B(0, 2)<32;8,4> { NoDDClr }
+ mov (16) PREV_MB_YB(1, 16)<1> BUF_B(0, 3)<32;8,4> { NoDDChk }
+*/
+// End of load_Y_4x16T
diff --git a/shaders/h264/ildb/load_Top_UV_8x2.asm b/shaders/h264/ildb/load_Top_UV_8x2.asm new file mode 100644 index 0000000..844291f --- /dev/null +++ b/shaders/h264/ildb/load_Top_UV_8x2.asm @@ -0,0 +1,70 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module Name: Load_Top_UV_8X2.Asm
+//
+// Load UV 8X2 Block
+//
+//----------------------------------------------------------------
+// Symbols ceed To be defined before including this module
+//
+// Source Region Is :UB
+// BUF_D: BUF_D Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=UD
+
+// Binding Table Index:
+// BI_SRC_UV: Binding Table Index Of UV Surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD2:w
+#endif
+
+#if defined(_PROGRESSIVE)
+ // Read U+V
+ mov (1) MSGSRC.0:ud ORIX_TOP:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_TOP:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0001000F:ud { NoDDChk } // NV12 U+V block width and height (16x2)
+
+ // Read 1 GRF from DEST surface as the above MB has been deblocked.
+ //send (8) TOP_MB_UD(0)<1> MSGHDRU MSGSRC<8;8,1>:ud RESP_LEN(1)+DWBRMSGDSC_RC+BI_DEST_UV
+ mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC+BI_DEST_UV:ud
+#endif
+
+#if defined(_FIELD)
+
+// cmp.z.f0.0 (1) NULLREGW PicTypeC:w 0:w // Get pic type flag
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+ // They are used later in this file
+
+ // Read U+V
+ mov (1) MSGSRC.0:ud ORIX_TOP:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_TOP:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0001000F:ud { NoDDChk } // NV12 U+V block width and height (16x2)
+
+ // Load NV12 U+V
+
+ // Set message descriptor
+ // Frame picture
+// (f0.0) mov (1) MSGDSC DWBRMSGDSC_RC+0x00010000+BI_DEST_UV:ud // Read 1 GRF from SRC_UV
+// (f0.0) jmpi Load_Top_UV_8x2
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_BF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_TF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y top field
+
+//Load_Top_UV_8x2:
+
+ // Read 1 GRF from DEST surface as the above MB has been deblocked.
+// send (8) PREV_MB_UD(0)<1> MSGHDRU MSGSRC<8;8,1>:ud MSGDSC
+
+#endif
+
+ send (8) TOP_MB_UD(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// End of load_Top_UV_8x2.asm
diff --git a/shaders/h264/ildb/load_Top_UV_8x2_Mbaff.asm b/shaders/h264/ildb/load_Top_UV_8x2_Mbaff.asm new file mode 100644 index 0000000..d60aa4e --- /dev/null +++ b/shaders/h264/ildb/load_Top_UV_8x2_Mbaff.asm @@ -0,0 +1,79 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module Name: Load_Top_UV_8X2.Asm
+//
+// Load UV 8X2 Block
+//
+//----------------------------------------------------------------
+// Symbols ceed To be defined before including this module
+//
+// Source Region Is :UB
+// BUF_D: BUF_D Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=UD
+
+// Binding Table Index:
+// BI_SRC_UV: Binding Table Index Of UV Surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD2:w
+#endif
+
+ // FieldModeCurrentMbFlag determines how to access above MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w
+
+ // Read U+V
+ mov (1) MSGSRC.0:ud ORIX_TOP:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:d ORIY_TOP:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0001000F:ud { NoDDChk } // NV12 U+V block width and height (16x2)
+
+ // Load NV12 U+V
+
+ // Set message descriptor
+
+ (f0.0) if (1) ELSE_UV_8X2
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC+BI_DEST_UV:ud // Read 1 GRF from SRC_UV
+
+ // Add vertical offset 8 for bot MB in MBAFF mode
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 8:w
+
+ // Dual field mode setup
+ and.z.f0.1 (1) NULLREGW DualFieldMode:w 1:w
+ (f0.1) jmpi NOT_DUAL_FIELD_UV
+
+ add (1) MSGSRC.1:d MSGSRC.1:d -2:w { NoDDClr } // Load 4 lines in stead of 2
+ mov (1) MSGSRC.2:ud 0x0003000F:ud { NoDDChk } // New block width and height (16x8)
+
+ add (1) MSGDSC MSGDSC RESP_LEN(1):ud // 1 more GRF to receive
+
+NOT_DUAL_FIELD_UV:
+
+ELSE_UV_8X2:
+ else (1) ENDIF_UV_8X2
+
+ // Field picture
+ asr (1) MSGSRC.1:d ORIY_CUR:w 2:w // asr 1: NV12 U+V block origin y = half of Y comp
+ // asr 1: Reduce y by half in field access mode
+
+ (f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_BF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC_TF+BI_DEST_UV:ud // Read 1 GRF from SRC_Y top field
+
+ add (1) MSGSRC.1:d MSGSRC.1:d -2:w // for last 2 rows of above MB
+
+ endif
+ENDIF_UV_8X2:
+
+ // Read 1 GRF from DEST surface as the above MB has been deblocked.
+ send (8) PREV_MB_UD(0)<1> MSGHDRU MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// End of load_Top_UV_8x2.asm
diff --git a/shaders/h264/ildb/load_Top_Y_16x4.asm b/shaders/h264/ildb/load_Top_Y_16x4.asm new file mode 100644 index 0000000..7590d62 --- /dev/null +++ b/shaders/h264/ildb/load_Top_Y_16x4.asm @@ -0,0 +1,70 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module Name: Load_Y_16X4.asm
+//
+// Load Y 16X4 Block to PREV_MB_YD
+//
+//----------------------------------------------------------------
+// Symbols Need To Be Defined Before Including This Module
+//
+// Source Region In :Ud
+// Src_YD: Src_Yd Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=Ud // 3 Grfs (2 For Y, 1 For U+V)
+//
+// Source Region Is :Ub. The Same Region As :Ud Region
+// Src_YB: Src_Yb Base=Rxx Elementsize=1 Srcregion=Region(16,1) Type=Ub // 2 Grfs
+//
+// Binding Table Index:
+// Bi_Src_Y: Binding Table Index Of Y Surface
+//
+// Temp Buffer:
+// Buf_D: Buf_D Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=Ud
+// Buf_B: Buf_B Base=Rxx Elementsize=1 Srcregion=Region(16,1) Type=Ub
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD2:w
+#endif
+
+#if defined(_PROGRESSIVE)
+ // Read Y
+ mov (2) MSGSRC.0<1>:ud ORIX_TOP<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x0003000F:ud { NoDDChk } // Block width and height (16x4)
+
+ mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC+BI_DEST_Y:ud // Read 2 GRFs from SRC_Y
+#endif
+
+#if defined(_FIELD)
+
+// cmp.z.f0.0 (1) NULLREGW PicTypeC:w 0:w // Get pic type flag
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+ // they are used later in this file
+
+ mov (2) MSGSRC.0<1>:ud ORIX_TOP<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x0003000F:ud { NoDDChk } // Block width and height (16x4)
+
+ // Set message descriptor
+
+ // Frame picture
+// (f0.0) mov (1) MSGDSC DWBRMSGDSC_RC+0x00020000+BI_DEST_Y:ud // Read 2 GRFs from SRC_Y
+// (f0.0) jmpi load_Y_16x4
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_BF+BI_DEST_Y:ud // Read 2 GRFs from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_TF+BI_DEST_Y:ud // Read 2 GRFs from SRC_Y top field
+
+//load_Y_16x4:
+ // Read 2 GRFs from DEST surface, as the above MB has been deblocked
+// send (8) PREV_MB_YD(0)<1> MSGHDRY MSGSRC<8;8,1>:ud MSGDSC
+
+#endif
+
+ send (8) TOP_MB_YD(0)<1> MSGHDRT MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// End of load_Y_16x4.asm
diff --git a/shaders/h264/ildb/load_Top_Y_16x4_Mbaff.asm b/shaders/h264/ildb/load_Top_Y_16x4_Mbaff.asm new file mode 100644 index 0000000..0237882 --- /dev/null +++ b/shaders/h264/ildb/load_Top_Y_16x4_Mbaff.asm @@ -0,0 +1,81 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module Name: Load_Y_16X4.asm
+//
+// Load Y 16X4 Block to PREV_MB_YD
+//
+//----------------------------------------------------------------
+// Symbols Need To Be Defined Before Including This Module
+//
+// Source Region In :Ud
+// Src_YD: Src_Yd Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=Ud // 3 Grfs (2 For Y, 1 For U+V)
+//
+// Source Region Is :Ub. The Same Region As :Ud Region
+// Src_YB: Src_Yb Base=Rxx Elementsize=1 Srcregion=Region(16,1) Type=Ub // 2 Grfs
+//
+// Binding Table Index:
+// Bi_Src_Y: Binding Table Index Of Y Surface
+//
+// Temp Buffer:
+// Buf_D: Buf_D Base=Rxx Elementsize=4 Srcregion=Region(8,1) Type=Ud
+// Buf_B: Buf_B Base=Rxx Elementsize=1 Srcregion=Region(16,1) Type=Ub
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD2:w
+#endif
+ // FieldModeCurrentMbFlag determines how to access above MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w
+
+ // Read Y
+ mov (2) MSGSRC.0<1>:ud ORIX_TOP<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x0003000F:ud { NoDDChk } // Block width and height (16x4)
+
+ // Set message descriptor
+
+ (f0.0) if (1) ELSE_Y_16x4
+
+ // Frame picture
+ mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC+BI_DEST_Y:ud // Read 2 GRFs from SRC_Y
+
+ // Add vertical offset 16 for bot MB in MBAFF mode
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 16:w
+
+ // Dual field mode setup
+ and.z.f0.1 (1) NULLREGW DualFieldMode:w 1:w
+ (f0.1) jmpi NOT_DUAL_FIELD
+
+ add (1) MSGSRC.1:d MSGSRC.1:d -4:w { NoDDClr } // Load 8 lines in above MB
+ mov (1) MSGSRC.2:ud 0x0007000F:ud { NoDDChk } // New block width and height (16x8)
+
+ add (1) MSGDSC MSGDSC RESP_LEN(2):ud // 2 more GRF to receive
+
+NOT_DUAL_FIELD:
+
+ELSE_Y_16x4:
+ else (1) ENDIF_Y_16x4
+
+ asr (1) MSGSRC.1:d ORIY_CUR:w 1:w // Reduce y by half in field access mode
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_BF+BI_DEST_Y:ud // Read 2 GRFs from SRC_Y bottom field
+ (-f0.1) mov (1) MSGDSC RESP_LEN(2)+DWBRMSGDSC_RC_TF+BI_DEST_Y:ud // Read 2 GRFs from SRC_Y top field
+
+ add (1) MSGSRC.1:d MSGSRC.1:d -4:w // for last 4 rows of above MB
+
+ endif
+ENDIF_Y_16x4:
+
+ // Read 2 GRFs from DEST surface, as the above MB has been deblocked
+ send (8) PREV_MB_YD(0)<1> MSGHDRY MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// End of load_Y_16x4.asm
diff --git a/shaders/h264/ildb/saveNV12_16x16.asm b/shaders/h264/ildb/saveNV12_16x16.asm new file mode 100644 index 0000000..1cbe27b --- /dev/null +++ b/shaders/h264/ildb/saveNV12_16x16.asm @@ -0,0 +1,53 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: saveNV12_16x16.asm
+//
+// Save a NV12 16x16 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 4 GRF
+//
+// Binding table index:
+// BI_DEST_Y: Binding table index of Y surface
+// BI_DEST_UV: Binding table index of UV surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD4:w
+#endif
+
+
+ mov (2) MSGSRC.0<1>:ud ORIX_CUR<2;2,1>:w // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud // Block width and height (16x16)
+
+ // Pack Y
+ mov (16) MSGPAYLOADD(0)<1> SRC_YD(0) // Compressed inst
+ mov (16) MSGPAYLOADD(2)<1> SRC_YD(2)
+ mov (16) MSGPAYLOADD(4)<1> SRC_YD(4)
+ mov (16) MSGPAYLOADD(6)<1> SRC_YD(6)
+
+ send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(8)+DWBWMSGDSC+BI_DEST_Y // Write 8 GRFs
+
+
+
+ asr (1) MSGSRC.1:ud MSGSRC.1:ud 1:w // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2<1>:ud 0x0007000F:ud // NV12 U+V block width and height (16x8)
+
+ mov (16) MSGPAYLOADD(0)<1> SRC_UD(0) // Compressed inst
+ mov (16) MSGPAYLOADD(2)<1> SRC_UD(2)
+
+ send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(4)+DWBWMSGDSC+BI_DEST_UV // Write 4 GRFs
+
+
+// End of saveNV12_16x16.asm
diff --git a/shaders/h264/ildb/saveNV12_16x4.asm b/shaders/h264/ildb/saveNV12_16x4.asm new file mode 100644 index 0000000..3a99995 --- /dev/null +++ b/shaders/h264/ildb/saveNV12_16x4.asm @@ -0,0 +1,50 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: saveNV12_16x4.asm
+//
+// Save a NV12 16x4 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 2 GRFs
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 1 GRF
+//
+// Binding table index:
+// BI_DEST_Y: Binding table index of Y surface
+// BI_DEST_UV: Binding table index of UV surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD5:w
+#endif
+
+ mov (2) MSGSRC.0<1>:ud ORIX_TOP<2;2,1>:w // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x0003000F:ud // Block width and height (16x4)
+
+ // Pack Y
+ mov (16) MSGPAYLOADD(0)<1> SRC_YD(0) // Compressed inst
+
+ send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(2)+DWBWMSGDSC+BI_DEST_Y // Write 2 GRFs
+
+
+ asr (1) MSGSRC.1:ud MSGSRC.1:ud 1:w // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2<1>:ud 0x0001000F:ud // NV12 U+V block width and height (16x2)
+
+ // Pack U and V
+// mov (16) MSGPAYLOADB(0,0)<2> SRC_UB(0,0)
+// mov (16) MSGPAYLOADB(0,1)<2> SRC_VB(0,0)
+
+ mov (8) MSGPAYLOADD(0,0)<1> SRC_UD(0)
+
+ send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(1)+DWBWMSGDSC+BI_DEST_UV // Write 1 GRF
+
+// End of saveNV12_16x4.asm
diff --git a/shaders/h264/ildb/saveNV12_16x4T.asm b/shaders/h264/ildb/saveNV12_16x4T.asm new file mode 100644 index 0000000..66085d1 --- /dev/null +++ b/shaders/h264/ildb/saveNV12_16x4T.asm @@ -0,0 +1,113 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: saveNV12_16x4T.asm
+//
+// Transpose 16x4 to 4x16 YNV12 data and write to memory
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Left MB region:
+// PREV_MB_YB: Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+// PREV_MB_UW: Base=ryy ElementSize=2 SrcRegion=REGION(8,1) Type=uw
+
+// Binding table index:
+// BI_SRC_Y: Binding table index of Y surface
+// BI_SRC_UV: Binding table index of UV surface (NV12)
+//
+// Temp buffer:
+// BUF_B: BUF_B Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+// BUF_W: BUF_W Base=rxx ElementSize=1 SrcRegion=REGION(8,1) Type=uw
+//
+//
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD6:w
+#endif
+
+ mov (2) MSGSRC.0<1>:ud ORIX_LEFT<2;2,1>:w // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F0003:ud // 4x16
+
+// Transpose Y, save them to MRFs
+
+// 16x4 Y src in GRF (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f1 e1 d1 c1 b1 a1 91 81 71 61 51 41 31 21 11 01 f0 e0 d0 c0 b0 a0 90 80 70 60 50 40 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 b3 a3 93 83 73 63 53 43 33 23 13 03 f2 e2 d2 c2 b2 a2 92 82 72 62 52 42 32 22 12 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// First step (16) <1> <=== <16;4,1>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 63 53 43 72 62 52 42 71 61 51 41 70 60 50 40 33 23 13 03 32 22 12 02 31 21 11 01 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 f2 e2 d2 c2 f1 e1 d1 c1 f0 e0 d0 c0 b3 a3 93 83 b2 a2 92 82 b1 a1 91 81 b0 a0 90 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+ // The first step
+ mov (16) BUF_B(0,0)<1> PREV_MB_YB(0,0)<16;4,1>
+ mov (16) BUF_B(0,16)<1> PREV_MB_YB(0,4)<16;4,1>
+ mov (16) BUF_B(1,0)<1> PREV_MB_YB(0,8)<16;4,1>
+ mov (16) BUF_B(1,16)<1> PREV_MB_YB(0,12)<16;4,1>
+
+//
+// Second step (16) <1> <=== <1;4,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 72 71 70 63 62 61 60 53 52 51 50 43 42 41 40 33 32 31 30 23 22 21 20 13 12 11 10 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 f2 f1 f0 e3 e2 e1 e0 d3 d2 d1 d0 c3 c2 c1 c0 b3 b2 b1 b0 a3 a2 a1 a0 93 92 91 90 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+ // The second step
+// mov (16) MSGPAYLOADB(0,0)<1> BUF_B(0,0)<32;8,4> // Read 2 rows, write 1 row
+// mov (16) MSGPAYLOADB(0,16)<1> BUF_B(0,1)<32;8,4>
+// mov (16) MSGPAYLOADB(1,0)<1> BUF_B(0,2)<32;8,4>
+// mov (16) MSGPAYLOADB(1,16)<1> BUF_B(0,3)<32;8,4>
+
+ mov (16) MSGPAYLOADB(0,0)<1> BUF_B(0,0)<1;4,4>
+ mov (16) MSGPAYLOADB(0,16)<1> BUF_B(0,16)<1;4,4>
+ mov (16) MSGPAYLOADB(1,0)<1> BUF_B(1,0)<1;4,4>
+ mov (16) MSGPAYLOADB(1,16)<1> BUF_B(1,16)<1;4,4>
+
+// Transposed Y in 4x16 is ready for writting to dataport.
+//
+ send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(2)+DWBWMSGDSC+BI_DEST_Y // Write 2 GRFs
+
+
+
+/////////////////////////////////////////////////////////////////////////////////////////////////////
+
+ // Transpose U/V, save them to MRFs in NV12 format
+ asr (1) MSGSRC.1:ud MSGSRC.1:ud 1:w // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2<1>:ud 0x00070003:ud // NV12 U+V block width and height (4x8)
+
+
+// 16x2 UV src in GRF (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// First step (8) <1> <=== <8;4,1>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 70 70 60 60 50 50 40 40 31 31 21 21 11 11 01 01 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+ mov (8) BUF_W(0,0)<1> PREV_MB_UW(0,0)<8;4,1>
+ mov (8) BUF_W(0,8)<1> PREV_MB_UW(0,4)<8;4,1>
+
+// Second step (8) <1> <=== <1;2,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 70 70 61 61 60 60 51 51 50 50 41 41 40 40 31 31 30 30 21 21 20 20 11 11 10 10 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+ mov (8) MSGPAYLOADW(0,0)<1> BUF_W(0,0)<1;2,4>
+ mov (8) MSGPAYLOADW(0,8)<1> BUF_W(0,8)<1;2,4>
+
+// Transposed U+V in NV12 in 4x8 is ready for writting to dataport.
+
+ send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(1)+DWBWMSGDSC+BI_DEST_UV // Write 1 GRF
+
+
diff --git a/shaders/h264/ildb/save_Cur_UV_8x8.asm b/shaders/h264/ildb/save_Cur_UV_8x8.asm new file mode 100644 index 0000000..39d4b09 --- /dev/null +++ b/shaders/h264/ildb/save_Cur_UV_8x8.asm @@ -0,0 +1,53 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Cur_UV_8x8.asm
+//
+// Save UV 8x8 block (8x8U + 8x8V in NV12)
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 4 GRF
+//
+// Binding table index:
+// BI_DEST_UV: Binding table index of UV surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD4:w
+#endif
+
+#if defined(_FIELD)
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+#endif
+
+ mov (1) MSGSRC.0:ud ORIX_CUR:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_CUR:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0007000F:ud { NoDDChk } // NV12 U+V block width and height (16x8)
+
+ mov (16) MSGPAYLOADD(0)<1> SRC_UD(0) // Compressed inst
+ mov (16) MSGPAYLOADD(2)<1> SRC_UD(2)
+
+#if defined(_PROGRESSIVE)
+ mov (1) MSGDSC MSG_LEN(4)+DWBWMSGDSC+BI_DEST_UV:ud
+// send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DWBWMSGDSC+0x00400000+BI_DEST_UV
+#endif
+
+#if defined(_FIELD)
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(4)+DWBWMSGDSC+ENMSGDSCBF+BI_DEST_UV:ud // Write 4 GRFs to DEST_UV bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(4)+DWBWMSGDSC+ENMSGDSCTF+BI_DEST_UV:ud // Write 4 GRFs to DEST_UV top field
+
+#endif
+
+ send (8) null:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+// End of save_Cur_UV_8x8.asm
diff --git a/shaders/h264/ildb/save_Cur_UV_8x8_Mbaff.asm b/shaders/h264/ildb/save_Cur_UV_8x8_Mbaff.asm new file mode 100644 index 0000000..6f469ea --- /dev/null +++ b/shaders/h264/ildb/save_Cur_UV_8x8_Mbaff.asm @@ -0,0 +1,62 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Cur_UV_8x8.asm
+//
+// Save UV 8x8 block (8x8U + 8x8V in NV12)
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 4 GRF
+//
+// Binding table index:
+// BI_DEST_UV: Binding table index of UV surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD4:w
+#endif
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w
+
+ mov (1) MSGSRC.0:ud ORIX_CUR:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_CUR:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0007000F:ud { NoDDChk } // NV12 U+V block width and height (16x8)
+
+ mov (16) MSGPAYLOADD(0)<1> SRC_UD(0) // Compressed inst
+ mov (16) MSGPAYLOADD(2)<1> SRC_UD(2)
+
+ // Set message descriptor
+
+ (f0.0) if (1) ELSE_UV_8X8
+
+ // Frame picture
+ mov (1) MSGDSC MSG_LEN(4)+DWBWMSGDSC+BI_DEST_UV:ud // Write 4 GRFs to DEST_UV
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 8:w // Add vertical offset 8 for bot MB in MBAFF mode
+
+ELSE_UV_8X8:
+ else (1) ENDIF_UV_8X8
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(4)+DWBWMSGDSC+ENMSGDSCBF+BI_DEST_UV:ud // Write 4 GRFs to DEST_UV bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(4)+DWBWMSGDSC+ENMSGDSCTF+BI_DEST_UV:ud // Write 4 GRFs to DEST_UV top field
+
+ asr (1) MSGSRC.1:d MSGSRC.1:d 1:w // Reduce y by half in field access mode
+
+ endif
+ENDIF_UV_8X8:
+
+ send (8) null:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+// End of save_Cur_UV_8x8.asm
diff --git a/shaders/h264/ildb/save_Cur_Y_16x16.asm b/shaders/h264/ildb/save_Cur_Y_16x16.asm new file mode 100644 index 0000000..fac22cd --- /dev/null +++ b/shaders/h264/ildb/save_Cur_Y_16x16.asm @@ -0,0 +1,56 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Cur_Y_16x16.asm
+//
+// Save a Y 16x16 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+//
+// Binding table index:
+// BI_DEST_Y: Binding table index of Y surface
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD4:w
+#endif
+
+#if defined(_FIELD)
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+#endif
+
+ mov (2) MSGSRC.0<1>:ud ORIX_CUR<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud { NoDDChk } // Block width and height (16x16)
+
+ // Pack Y
+ mov (16) MSGPAYLOADD(0)<1> SRC_YD(0) // Compressed inst
+ mov (16) MSGPAYLOADD(2)<1> SRC_YD(2)
+ mov (16) MSGPAYLOADD(4)<1> SRC_YD(4)
+ mov (16) MSGPAYLOADD(6)<1> SRC_YD(6)
+
+
+#if defined(_PROGRESSIVE)
+ mov (1) MSGDSC MSG_LEN(8)+DWBWMSGDSC+BI_DEST_Y:ud
+// send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DWBWMSGDSC+0x00800000+BI_DEST_Y
+#endif
+
+#if defined(_FIELD)
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(8)+DWBWMSGDSC+ENMSGDSCBF+BI_DEST_Y:ud // Write 8 GRFs to DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(8)+DWBWMSGDSC+ENMSGDSCTF+BI_DEST_Y:ud // Write 8 GRFs to DEST_Y top field
+
+#endif
+
+ send (8) null:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+// End of save_Cur_Y_16x16.asm
diff --git a/shaders/h264/ildb/save_Cur_Y_16x16_Mbaff.asm b/shaders/h264/ildb/save_Cur_Y_16x16_Mbaff.asm new file mode 100644 index 0000000..6ab78dc --- /dev/null +++ b/shaders/h264/ildb/save_Cur_Y_16x16_Mbaff.asm @@ -0,0 +1,64 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Cur_Y_16x16.asm
+//
+// Save a Y 16x16 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 8 GRFs
+//
+// Binding table index:
+// BI_DEST_Y: Binding table index of Y surface
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD4:w
+#endif
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w
+
+ mov (2) MSGSRC.0<1>:ud ORIX_CUR<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F000F:ud { NoDDChk } // Block width and height (16x16 or 12x16)
+
+ // Pack Y
+ mov (16) MSGPAYLOADD(0)<1> SRC_YD(0) // Compressed inst
+ mov (16) MSGPAYLOADD(2)<1> SRC_YD(2)
+ mov (16) MSGPAYLOADD(4)<1> SRC_YD(4)
+ mov (16) MSGPAYLOADD(6)<1> SRC_YD(6)
+
+ // Set message descriptor
+
+ (f0.0) if (1) ELSE_Y_16x16
+
+ // Frame picture
+ mov (1) MSGDSC MSG_LEN(8)+DWBWMSGDSC+BI_DEST_Y:ud // Write 8 GRFs to DEST_Y
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 16:w // Add vertical offset 16 for bot MB in MBAFF mode
+
+ELSE_Y_16x16:
+ else (1) ENDIF_Y_16x16
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(8)+DWBWMSGDSC+ENMSGDSCBF+BI_DEST_Y:ud // Write 8 GRFs to DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(8)+DWBWMSGDSC+ENMSGDSCTF+BI_DEST_Y:ud // Write 8 GRFs to DEST_Y top field
+
+ asr (1) MSGSRC.1:d MSGSRC.1:d 1:w // Reduce y by half in field access mode
+
+ endif
+ENDIF_Y_16x16:
+
+ send (8) null:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+// End of save_Cur_Y_16x16.asm
diff --git a/shaders/h264/ildb/save_Left_UV_8x2T.asm b/shaders/h264/ildb/save_Left_UV_8x2T.asm new file mode 100644 index 0000000..172002e --- /dev/null +++ b/shaders/h264/ildb/save_Left_UV_8x2T.asm @@ -0,0 +1,72 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Left_UV_8x2T.asm
+//
+// Transpose 8x2 to 2x8 UV data and write to memory
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Left MB region:
+// PREV_MB_UW: Base=ryy ElementSize=2 SrcRegion=REGION(8,1) Type=uw
+
+// Binding table index:
+// BI_SRC_UV: Binding table index of UV surface (NV12)
+//
+// Temp buffer:
+// BUF_W: BUF_W Base=rxx ElementSize=1 SrcRegion=REGION(8,1) Type=uw
+//
+//
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD6:w
+#endif
+
+#if defined(_FIELD)
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+#endif
+
+ // Transpose U/V, save them to MRFs in NV12 format
+ mov (1) MSGSRC.0:ud ORIX_LEFT:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_LEFT:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x00070003:ud { NoDDChk } // NV12 U+V block width and height (4x8)
+
+
+// 16x2 UV src in GRF (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// First step (8) <1> <=== <8;4,1>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 70 70 60 60 50 50 40 40 31 31 21 21 11 11 01 01 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+ mov (8) LEFT_TEMP_W(0,0)<1> PREV_MB_UW(0,0)<8;4,1> { NoDDClr }
+ mov (8) LEFT_TEMP_W(0,8)<1> PREV_MB_UW(0,4)<8;4,1> { NoDDChk }
+
+// Second step (8) <1> <=== <1;2,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 70 70 61 61 60 60 51 51 50 50 41 41 40 40 31 31 30 30 21 21 20 20 11 11 10 10 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+ mov (8) MSGPAYLOADW(0,0)<1> LEFT_TEMP_W(0,0)<1;2,4>
+ mov (8) MSGPAYLOADW(0,8)<1> LEFT_TEMP_W(0,8)<1;2,4>
+
+// Transposed U+V in NV12 in 4x8 is ready for writting to dataport.
+
+#if defined(_PROGRESSIVE)
+ mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC+BI_DEST_UV:ud
+// send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DWBWMSGDSC+0x00100000+BI_DEST_UV
+#endif
+
+#if defined(_FIELD)
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC+ENMSGDSCBF+BI_DEST_UV:ud // Write 1 GRF to DEST_UV bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC+ENMSGDSCTF+BI_DEST_UV:ud // Write 1 GRF to DEST_UV top field
+
+#endif
+ send (8) null:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
diff --git a/shaders/h264/ildb/save_Left_UV_8x2T_Mbaff.asm b/shaders/h264/ildb/save_Left_UV_8x2T_Mbaff.asm new file mode 100644 index 0000000..f98b311 --- /dev/null +++ b/shaders/h264/ildb/save_Left_UV_8x2T_Mbaff.asm @@ -0,0 +1,82 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Left_UV_8x2T.asm
+//
+// Transpose 8x2 to 2x8 UV data and write to memory
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Left MB region:
+// PREV_MB_UW: Base=ryy ElementSize=2 SrcRegion=REGION(8,1) Type=uw
+
+// Binding table index:
+// BI_SRC_UV: Binding table index of UV surface (NV12)
+//
+// Temp buffer:
+// BUF_W: BUF_W Base=rxx ElementSize=1 SrcRegion=REGION(8,1) Type=uw
+//
+//
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD6:w
+#endif
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w
+
+ // Transpose U/V, save them to MRFs in NV12 format
+ mov (1) MSGSRC.0:ud ORIX_LEFT:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_LEFT:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x00070003:ud { NoDDChk } // NV12 U+V block width and height (4x8)
+
+
+// 16x2 UV src in GRF (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 31 31 21 21 11 11 01 01 70 70 60 60 50 50 40 40 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// First step (8) <1> <=== <8;4,1>:w
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 61 61 51 51 41 41 70 70 60 60 50 50 40 40 31 31 21 21 11 11 01 01 30 30 20 20 10 10 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+ mov (8) BUF_W(0,0)<1> PREV_MB_UW(0,0)<8;4,1> { NoDDClr }
+ mov (8) BUF_W(0,8)<1> PREV_MB_UW(0,4)<8;4,1> { NoDDChk }
+
+// Second step (8) <1> <=== <1;2,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |71 71 70 70 61 61 60 60 51 51 50 50 41 41 40 40 31 31 30 30 21 21 20 20 11 11 10 10 01 01 00 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+ mov (8) MSGPAYLOADW(0,0)<1> BUF_W(0,0)<1;2,4>
+ mov (8) MSGPAYLOADW(0,8)<1> BUF_W(0,8)<1;2,4>
+
+// Transposed U+V in NV12 in 4x8 is ready for writting to dataport.
+
+ // Set message descriptor
+
+ (f0.0) if (1) ELSE_UV_8X2T
+
+ // Frame picture
+ mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC+BI_DEST_UV:ud // Write 1 GRF to DEST_UV
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 8:w // Add vertical offset 8 for bot MB in MBAFF mode
+
+ELSE_UV_8X2T:
+ else (1) ENDIF_UV_8X2T
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC+ENMSGDSCBF+BI_DEST_UV:ud // Write 1 GRF to DEST_UV bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC+ENMSGDSCTF+BI_DEST_UV:ud // Write 1 GRF to DEST_UV top field
+
+ asr (1) MSGSRC.1:d MSGSRC.1:d 1:w // Reduce y by half in field access mode
+
+ endif
+ENDIF_UV_8X2T:
+
+ send (8) null:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
diff --git a/shaders/h264/ildb/save_Left_Y_16x4T.asm b/shaders/h264/ildb/save_Left_Y_16x4T.asm new file mode 100644 index 0000000..84d81e9 --- /dev/null +++ b/shaders/h264/ildb/save_Left_Y_16x4T.asm @@ -0,0 +1,89 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Left_Y_16x4T.asm
+//
+// Transpose 16x4 to 4x16 Y data and write to memory
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Left MB region:
+// PREV_MB_YB: Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+
+// Binding table index:
+// BI_SRC_Y: Binding table index of Y surface
+//
+// Temp buffer:
+// BUF_B: BUF_B Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+//
+//
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD6:w
+#endif
+
+#if defined(_FIELD)
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+#endif
+
+ mov (2) MSGSRC.0<1>:ud ORIX_LEFT<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F0003:ud { NoDDChk } // 4x16
+
+// Transpose Y, save them to MRFs
+
+// 16x4 Y src in GRF (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f1 e1 d1 c1 b1 a1 91 81 71 61 51 41 31 21 11 01 f0 e0 d0 c0 b0 a0 90 80 70 60 50 40 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 b3 a3 93 83 73 63 53 43 33 23 13 03 f2 e2 d2 c2 b2 a2 92 82 72 62 52 42 32 22 12 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// First step (16) <1> <=== <16;4,1>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 63 53 43 72 62 52 42 71 61 51 41 70 60 50 40 33 23 13 03 32 22 12 02 31 21 11 01 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 f2 e2 d2 c2 f1 e1 d1 c1 f0 e0 d0 c0 b3 a3 93 83 b2 a2 92 82 b1 a1 91 81 b0 a0 90 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+ // The first step
+ mov (16) LEFT_TEMP_B(0,0)<1> PREV_MB_YB(0,0)<16;4,1> { NoDDClr }
+ mov (16) LEFT_TEMP_B(0,16)<1> PREV_MB_YB(0,4)<16;4,1> { NoDDChk }
+ mov (16) LEFT_TEMP_B(1,0)<1> PREV_MB_YB(0,8)<16;4,1> { NoDDClr }
+ mov (16) LEFT_TEMP_B(1,16)<1> PREV_MB_YB(0,12)<16;4,1> { NoDDChk }
+
+//
+// Second step (16) <1> <=== <1;4,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 72 71 70 63 62 61 60 53 52 51 50 43 42 41 40 33 32 31 30 23 22 21 20 13 12 11 10 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 f2 f1 f0 e3 e2 e1 e0 d3 d2 d1 d0 c3 c2 c1 c0 b3 b2 b1 b0 a3 a2 a1 a0 93 92 91 90 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+ // The second step
+ mov (16) MSGPAYLOADB(0,0)<1> LEFT_TEMP_B(0,0)<1;4,4>
+ mov (16) MSGPAYLOADB(0,16)<1> LEFT_TEMP_B(0,16)<1;4,4>
+ mov (16) MSGPAYLOADB(1,0)<1> LEFT_TEMP_B(1,0)<1;4,4>
+ mov (16) MSGPAYLOADB(1,16)<1> LEFT_TEMP_B(1,16)<1;4,4>
+
+// Transposed Y in 4x16 is ready for writting to dataport.
+
+
+#if defined(_PROGRESSIVE)
+ mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC+BI_DEST_Y:ud
+// send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DWBWMSGDSC+0x00200000+BI_DEST_Y
+#endif
+
+#if defined(_FIELD)
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC+ENMSGDSCBF+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC+ENMSGDSCTF+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y top field
+
+#endif
+
+ send (8) null:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
diff --git a/shaders/h264/ildb/save_Left_Y_16x4T_Mbaff.asm b/shaders/h264/ildb/save_Left_Y_16x4T_Mbaff.asm new file mode 100644 index 0000000..977ad4f --- /dev/null +++ b/shaders/h264/ildb/save_Left_Y_16x4T_Mbaff.asm @@ -0,0 +1,101 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Left_Y_16x4T.asm
+//
+// Transpose 16x4 to 4x16 Y data and write to memory
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Left MB region:
+// PREV_MB_YB: Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+
+// Binding table index:
+// BI_SRC_Y: Binding table index of Y surface
+//
+// Temp buffer:
+// BUF_B: BUF_B Base=rxx ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+//
+//
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD6:w
+#endif
+
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w
+
+ mov (2) MSGSRC.0<1>:ud ORIX_LEFT<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x000F0003:ud { NoDDChk } // 4x16
+
+// Transpose Y, save them to MRFs
+
+// 16x4 Y src in GRF (each pix is specified as yx)
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f1 e1 d1 c1 b1 a1 91 81 71 61 51 41 31 21 11 01 f0 e0 d0 c0 b0 a0 90 80 70 60 50 40 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 b3 a3 93 83 73 63 53 43 33 23 13 03 f2 e2 d2 c2 b2 a2 92 82 72 62 52 42 32 22 12 02|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+// First step (16) <1> <=== <16;4,1>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 63 53 43 72 62 52 42 71 61 51 41 70 60 50 40 33 23 13 03 32 22 12 02 31 21 11 01 30 20 10 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 e3 d3 c3 f2 e2 d2 c2 f1 e1 d1 c1 f0 e0 d0 c0 b3 a3 93 83 b2 a2 92 82 b1 a1 91 81 b0 a0 90 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+
+ // The first step
+ mov (16) BUF_B(0,0)<1> PREV_MB_YB(0,0)<16;4,1> { NoDDClr }
+ mov (16) BUF_B(0,16)<1> PREV_MB_YB(0,4)<16;4,1> { NoDDChk }
+ mov (16) BUF_B(1,0)<1> PREV_MB_YB(0,8)<16;4,1> { NoDDClr }
+ mov (16) BUF_B(1,16)<1> PREV_MB_YB(0,12)<16;4,1> { NoDDChk }
+
+//
+// Second step (16) <1> <=== <1;4,4>
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |73 72 71 70 63 62 61 60 53 52 51 50 43 42 41 40 33 32 31 30 23 22 21 20 13 12 11 10 03 02 01 00|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+// |f3 f2 f1 f0 e3 e2 e1 e0 d3 d2 d1 d0 c3 c2 c1 c0 b3 b2 b1 b0 a3 a2 a1 a0 93 92 91 90 83 82 81 80|
+// +-----------------------+-----------------------+-----------------------+-----------------------+
+//
+ // The second step
+ mov (16) MSGPAYLOADB(0,0)<1> BUF_B(0,0)<1;4,4>
+ mov (16) MSGPAYLOADB(0,16)<1> BUF_B(0,16)<1;4,4>
+ mov (16) MSGPAYLOADB(1,0)<1> BUF_B(1,0)<1;4,4>
+ mov (16) MSGPAYLOADB(1,16)<1> BUF_B(1,16)<1;4,4>
+
+// Transposed Y in 4x16 is ready for writting to dataport.
+
+ //***** Left MB is loaded the same as indicated by FieldModeCurrentMbFlag.
+
+ // Set message descriptor
+
+ (f0.0) if (1) ELSE_Y_16x4T
+
+ // Frame picture
+ mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y
+
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 16:w // Add vertical offset 16 for bot MB in MBAFF mode
+
+ELSE_Y_16x4T:
+ else (1) ENDIF_Y_16x4T
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC+ENMSGDSCBF+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC+ENMSGDSCTF+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y top field
+
+ asr (1) MSGSRC.1:d MSGSRC.1:d 1:w // Reduce y by half in field access mode
+
+ endif
+ENDIF_Y_16x4T:
+
+ send (8) null:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+
+
diff --git a/shaders/h264/ildb/save_Top_UV_8x2.asm b/shaders/h264/ildb/save_Top_UV_8x2.asm new file mode 100644 index 0000000..5263c35 --- /dev/null +++ b/shaders/h264/ildb/save_Top_UV_8x2.asm @@ -0,0 +1,52 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Top_UV_8x2.asm
+//
+// Save UV 8x2 block (8x2U + 8x2V in NV12)
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 1 GRF
+//
+// Binding table index:
+// BI_DEST_UV: Binding table index of UV surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD5:w
+#endif
+
+#if defined(_FIELD)
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+#endif
+
+ mov (1) MSGSRC.0:ud ORIX_TOP:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_TOP:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0001000F:ud { NoDDChk } // NV12 U+V block width and height (16x2)
+
+ mov (8) MSGPAYLOADD(0,0)<1> TOP_MB_UD(0)
+
+
+#if defined(_PROGRESSIVE)
+ mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC_WC+BI_DEST_UV:ud
+// send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DWBWMSGDSC+0x00100000+BI_DEST_UV
+#endif
+
+#if defined(_FIELD)
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC_WC+ENMSGDSCBF+BI_DEST_UV:ud // Write 1 GRF to DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC_WC+ENMSGDSCTF+BI_DEST_UV:ud // Write 1 GRF to DEST_Y top field
+
+#endif
+
+ send (8) WritebackResponse(0)<1> MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+// End of save_Top_UV_8x2.asm
diff --git a/shaders/h264/ildb/save_Top_UV_8x2_Mbaff.asm b/shaders/h264/ildb/save_Top_UV_8x2_Mbaff.asm new file mode 100644 index 0000000..ef2ba84 --- /dev/null +++ b/shaders/h264/ildb/save_Top_UV_8x2_Mbaff.asm @@ -0,0 +1,69 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Top_UV_8x2.asm
+//
+// Save UV 8x2 block (8x2U + 8x2V in NV12)
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_UD: SRC_UD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 1 GRF
+//
+// Binding table index:
+// BI_DEST_UV: Binding table index of UV surface (NV12)
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD5:w
+#endif
+ and.z.f0.1 (8) NULLREGW DualFieldMode<0;1,0>:w 1:w
+
+ // FieldModeCurrentMbFlag determines how to access above MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ // Pack U and V
+ mov (1) MSGSRC.0:ud ORIX_TOP:w { NoDDClr } // Block origin
+ asr (1) MSGSRC.1:ud ORIY_TOP:w 1:w { NoDDClr, NoDDChk } // NV12 U+V block origin y = half of Y comp
+ mov (1) MSGSRC.2:ud 0x0001000F:ud { NoDDChk } // NV12 U+V block width and height (16x2)
+
+ // Dual field mode
+ (f0.1) mov (8) MSGPAYLOADD(0)<1> PREV_MB_UD(0)
+ (-f0.1) mov (8) MSGPAYLOADD(0)<1> PREV_MB_UD(1) // for dual field mode, write last 2 rows
+
+ // Set message descriptor
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w
+
+ (f0.0) if (1) ELSE_UV_8X2_SAVE
+
+ // Frame picture
+ mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC_WC+BI_DEST_UV:ud // Write 1 GRFs to DEST_UV
+
+ // Add vertical offset 8 for bot MB in MBAFF mode
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 8:w
+
+ELSE_UV_8X2_SAVE:
+ else (1) ENDIF_UV_8X2_SAVE
+
+ asr (1) MSGSRC.1:d ORIY_CUR:w 2:w // asr 1: NV12 U+V block origin y = half of Y comp
+ // asr 1: Reduce y by half in field access mode
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC_WC+ENMSGDSCBF+BI_DEST_UV:ud // Write 1 GRF to DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(1)+DWBWMSGDSC_WC+ENMSGDSCTF+BI_DEST_UV:ud // Write 1 GRF to DEST_Y top field
+
+ add (1) MSGSRC.1:d MSGSRC.1:d -2:w // for last 4 rows of above MB
+
+ endif
+ENDIF_UV_8X2_SAVE:
+
+ send (8) WritebackResponse(0)<1> MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+// End of save_Top_UV_8x2.asm
diff --git a/shaders/h264/ildb/save_Top_Y_16x4.asm b/shaders/h264/ildb/save_Top_Y_16x4.asm new file mode 100644 index 0000000..8889087 --- /dev/null +++ b/shaders/h264/ildb/save_Top_Y_16x4.asm @@ -0,0 +1,52 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Top_Y_16x4.asm
+//
+// Save a Y 16x4 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 2 GRFs
+//
+// Binding table index:
+// BI_DEST_Y: Binding table index of Y surface
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD5:w
+#endif
+
+#if defined(_FIELD)
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w // Get bottom field flag
+#endif
+
+ mov (2) MSGSRC.0<1>:ud ORIX_TOP<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x0003000F:ud { NoDDChk } // Block width and height (16x4)
+
+ // Pack Y
+ mov (16) MSGPAYLOADD(0)<1> TOP_MB_YD(0) // Compressed inst
+
+
+#if defined(_PROGRESSIVE)
+ mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC_WC+BI_DEST_Y:ud
+// send (8) NULLREG MSGHDR MSGSRC<8;8,1>:ud DWBWMSGDSC+0x00200000+BI_DEST_Y
+#endif
+
+#if defined(_FIELD)
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC_WC+ENMSGDSCBF+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC_WC+ENMSGDSCTF+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y top field
+
+#endif
+
+ send (8) WritebackResponse(0)<1> MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+// End of save_Top_Y_16x4.asm
diff --git a/shaders/h264/ildb/save_Top_Y_16x4_Mbaff.asm b/shaders/h264/ildb/save_Top_Y_16x4_Mbaff.asm new file mode 100644 index 0000000..d8bb9a7 --- /dev/null +++ b/shaders/h264/ildb/save_Top_Y_16x4_Mbaff.asm @@ -0,0 +1,69 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_Top_Y_16x4.asm
+//
+// Save a Y 16x4 block
+//
+//----------------------------------------------------------------
+// Symbols need to be defined before including this module
+//
+// Source region in :ud
+// SRC_YD: SRC_YD Base=rxx ElementSize=4 SrcRegion=REGION(8,1) Type=ud // 2 GRFs
+//
+// Binding table index:
+// BI_DEST_Y: Binding table index of Y surface
+//
+//----------------------------------------------------------------
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0xDDD5:w
+#endif
+
+ and.z.f0.1 (16) NULLREGW DualFieldMode<0;1,0>:w 1:w
+
+ // FieldModeCurrentMbFlag determines how to access above MB
+ and.z.f0.0 (1) null:w r[ECM_AddrReg, BitFlags]:ub FieldModeCurrentMbFlag:w
+
+ mov (2) MSGSRC.0<1>:ud ORIX_TOP<2;2,1>:w { NoDDClr } // Block origin
+ mov (1) MSGSRC.2<1>:ud 0x0003000F:ud { NoDDChk } // Block width and height (16x4)
+
+ // Pack Y
+ // Dual field mode
+ (f0.1) mov (16) MSGPAYLOADD(0)<1> PREV_MB_YD(0) // Compressed inst
+ (-f0.1) mov (16) MSGPAYLOADD(0)<1> PREV_MB_YD(2) // for dual field mode, write last 4 rows
+
+ // Set message descriptor
+
+ and.nz.f0.1 (1) NULLREGW BitFields:w BotFieldFlag:w
+
+ (f0.0) if (1) ELSE_Y_16x4_SAVE
+
+ // Frame picture
+ mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC_WC+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y
+
+ // Add vertical offset 16 for bot MB in MBAFF mode
+ (f0.1) add (1) MSGSRC.1:d MSGSRC.1:d 16:w
+
+ELSE_Y_16x4_SAVE:
+ else (1) ENDIF_Y_16x4_SAVE
+
+ asr (1) MSGSRC.1:d ORIY_CUR:w 1:w // Reduce y by half in field access mode
+
+ // Field picture
+ (f0.1) mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC_WC+ENMSGDSCBF+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y bottom field
+ (-f0.1) mov (1) MSGDSC MSG_LEN(2)+DWBWMSGDSC_WC+ENMSGDSCTF+BI_DEST_Y:ud // Write 2 GRFs to DEST_Y top field
+
+ add (1) MSGSRC.1:d MSGSRC.1:d -4:w // for last 4 rows of above MB
+
+ endif
+ENDIF_Y_16x4_SAVE:
+
+ send (8) WritebackResponse(0)<1> MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+// End of save_Top_Y_16x4.asm
diff --git a/shaders/h264/ildb/writeURB.asm b/shaders/h264/ildb/writeURB.asm new file mode 100644 index 0000000..c75b220 --- /dev/null +++ b/shaders/h264/ildb/writeURB.asm @@ -0,0 +1,38 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: WriteURB.asm
+//
+// General purpose module to write data to URB using the URB handle/offset in r0
+//
+//----------------------------------------------------------------
+// Assume:
+// - a0.0 and a0.1 is meg desc, has been assign with URB offset and msg size
+// - MRFs are alrady assigned with data.
+//----------------------------------------------------------------
+//
+// 16x16 byte pixel block can be saved using just 1 "send" instruction.
+
+#if defined(_DEBUG)
+ mov (1) EntrySignature:w 0x3535:w
+#endif
+
+// URB write header:
+//mov (8) MSGSRC.0:ud r0.0<8;8,1>:ud // Copy parent R0 header
+
+//shr (1) Temp2_W:uw URBOffset:uw 1:w // divide by 2, because URB entry is counted by 512bits. Offset is counted by 256bits.
+//add (1) MSGSRC.0:uw r0.0:uw Temp2_W:uw
+
+
+shr (1) MSGSRC.0:uw URBOffset:uw 1:w // divide by 2, because URB entry is counted by 512bits. Offset is counted by 256bits.
+//mov (1) MSGSRC.0:uw URBOffset_2:uw
+
+//mov (1) MSGSRC.1:ud 0:ud // Reset Handle 1
+
+send null:uw m0 MSGSRC:uw URBWRITE URBWriteMsgDesc:ud // URB write
+//send null:ud MRF0 null:ud URBWriteMsgDesc:ud // URB write
diff --git a/shaders/h264/ildb/writeURB_UV_Child.asm b/shaders/h264/ildb/writeURB_UV_Child.asm new file mode 100644 index 0000000..27d257a --- /dev/null +++ b/shaders/h264/ildb/writeURB_UV_Child.asm @@ -0,0 +1,39 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: WriteURB_Child.asm
+//
+// General purpose module to write data to URB using the URB handle/offset in r0
+//
+//----------------------------------------------------------------
+// Assume:
+// - a0.0 and a0.1 is meg desc, has been assign with URB offset and msg size
+// - MRFs are alrady assigned with data.
+//----------------------------------------------------------------
+//
+// 16x16 byte pixel block can be saved using just 1 "send" instruction.
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0x3535:w
+#endif
+
+// URB write header:
+//mov (8) MSGSRC.0:ud r0.0<8;8,1>:ud // Copy parent R0 header
+
+//shr (1) Temp2_W:uw URBOffsetC:uw 1:w // divide by 2, because URB entry is counted by 512bits. Offset is counted by 256bits.
+//add (1) MSGSRC.0:uw r0.0:uw Temp2_W:uw
+
+shr (1) MSGSRC.0:uw URBOffsetC:uw 1:w // divide by 2, because URB entry is counted by 512bits. Offset is counted by 256bits.
+
+//mov (1) MSGSRC.1:ud 0:ud // Reset Handle 1
+
+ // URB write 1 MRFs,
+ // Current MB offset is in URBOffset, use it as write origin
+ // Add 2 to offset to store data be be passed to the right MB
+
+send null:uw m0 MSGSRC:uw URBWRITE MSG_LEN(1)+URBWMSGDSC+0x20 // URB write
diff --git a/shaders/h264/ildb/writeURB_Y_Child.asm b/shaders/h264/ildb/writeURB_Y_Child.asm new file mode 100644 index 0000000..e51e4ad --- /dev/null +++ b/shaders/h264/ildb/writeURB_Y_Child.asm @@ -0,0 +1,40 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: WriteURB_Child.asm
+//
+// General purpose module to write data to URB using the URB handle/offset in r0
+//
+//----------------------------------------------------------------
+// Assume:
+// - a0.0 and a0.1 is meg desc, has been assign with URB offset and msg size
+// - MRFs are alrady assigned with data.
+//----------------------------------------------------------------
+//
+// 16x16 byte pixel block can be saved using just 1 "send" instruction.
+
+#if defined(_DEBUG)
+ mov (1) EntrySignatureC:w 0x3535:w
+#endif
+
+// URB write header:
+//mov (8) MSGSRC.0:ud r0.0<8;8,1>:ud // Copy parent R0 header
+
+//shr (1) Temp2_W:uw URBOffsetC:uw 1:w // divide by 2, because URB entry is counted by 512bits. Offset is counted by 256bits.
+//add (1) MSGSRC.0:uw r0.0:uw Temp2_W:uw
+
+shr (1) MSGSRC.0:uw URBOffsetC:uw 1:w // divide by 2, because URB entry is counted by 512bits. Offset is counted by 256bits.
+
+//mov (1) MSGSRC.1:ud 0:ud // Reset Handle 1
+
+ // URB write 2 MRFs,
+ // Current MB offset is in URBOffset, use it as write origin
+ // Add 2 to offset to store data be be passed to the right MB
+ //mov (1) URBWriteMsgDesc:ud 0x06300020:ud
+
+send null:uw m0 MSGSRC:uw URBWRITE MSG_LEN(2)+URBWMSGDSC+0x20 // URB write
diff --git a/shaders/h264/mc/AVCMCInter.asm b/shaders/h264/mc/AVCMCInter.asm new file mode 100644 index 0000000..691fb33 --- /dev/null +++ b/shaders/h264/mc/AVCMCInter.asm @@ -0,0 +1,254 @@ +/*
+ * All inter-prediction macroblock kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: AVCMCInter.asm
+
+#ifdef INTERLABEL
+#undef INTERLABEL
+#endif
+
+#if defined(MBAFF)
+// < MBaff_Motion >
+#define INTERLABEL(x) x##_##MBF
+#elif defined(FIELD)
+// < FieldMB_Motion >
+#define INTERLABEL(x) x##_##FLD
+#else // FRAME
+// < FrameMB_Motion >
+#define INTERLABEL(x) x##_##FRM
+#endif
+//
+// Decoding an inter-prediction macroblock (conditional compile)
+// -DMBAFF : MBAff picture MB
+// -DFRAME : Frame picture MB
+// -DFIELD : Field picture MB
+// -DMBAFF -DMONO : MBAff mono picture MB
+// -DFRAME -DMONO : Frame mono picture MB
+// -DFIELD -DMONO : Field mono picture MB
+
+
+//#if !defined(__AVCMCInter__) // Make sure this is only included once
+//#define __AVCMCInter__
+
+
+// TODO: header files need to be in sync with intra prediction
+#include "header.inc"
+#include "inter_Header.inc"
+
+// TODO: Kernel names for mono cases
+#if defined(MBAFF)
+.kernel MBAff_Motion
+MBAFF_MB:
+#elif defined(FIELD)
+.kernel FieldMB_Motion
+FIELD_MB:
+#else // Frame
+.kernel FrameMB_Motion
+FRAME_MB:
+#endif
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+#if defined(MBAFF)
+mov (1) acc0:ud 0x0aaa55a5:ud
+#elif defined(FIELD)
+mov (1) acc0:ud 0x0baa55a5:ud
+#else // Frame
+mov (1) acc0:ud 0x0caa55a5:ud
+#endif
+#endif
+
+
+#ifdef SW_SCOREBOARD
+ CALL(scoreboard_start_inter,1)
+#endif
+
+ mov (8) gMSGSRC<1>:ud r0.0<8;8,1>:ud // Initialize message header payload with R0
+
+ and (1) gwMBTYPE<1> gMBTYPE:ub nMBTYPE_MASK:w // MB type
+ shl (2) gX<1>:w gORIX<2;2,1>:ub 4:w // Convert MB origin to pixel unit
+
+// #include "process_inter16x16.asm" // Handle B_L0_16x16 case with zero MVs and weighted pred off.
+ // In the case of B_L0_16x16 with zero MVs and weighted pred off, the kernel jumps to INTERLABEL(EXIT_LOOP).
+
+INTERLABEL(INIT_MBPARA):
+ #include "initialize_MBPara.asm"
+
+
+ //========================= BEGIN - LOOP_SUBMB ===========================
+ mov (1) gLOOP_SUBMB:uw 0:uw // 0, 2, 4, 6
+INTERLABEL(LOOP_SUBMB):
+
+ //========================== BEGIN - LOOP_DIR ============================
+ // Prediction flag (gPREDFLAG - 0:Pred_L0, 1:Pred_L1, 2:BiPred)
+ asr (1) gPREDFLAG:w gSUBMB_MODE:ub gLOOP_SUBMB:uw
+ mov (1) gLOOP_DIR:uw 1:uw // 1, 0
+ and (1) gPREDFLAG:w gPREDFLAG:w 0x3:w
+INTERLABEL(LOOP_DIR):
+
+ cmp.e.f0.0 (1) null:w gLOOP_DIR:w gPREDFLAG:w
+ (f0.0) jmpi INTERLABEL(LOOP_DIR_CONTINUE)
+
+ // Get binding table index
+ // & reference picture parity (gREFPARITY - 0:top, 0x100:bottom, x:frame)
+ // & address of interpolation result
+ cmp.e.f0.1 (1) null:w gLOOP_DIR:w 1:w
+ (f0.1) mov (1) gpINTP:ud nOFFSET_INTP0:ud {NoDDClr} //
+ (f0.1) and (1) gBIDX:w r[pBIDX]:ub 0x7f:w {NoDDChk} //
+ (-f0.1) mov (1) gpINTP:ud nOFFSET_INTP1:ud {NoDDClr} //
+ (-f0.1) and (1) gBIDX:w r[pBIDX,4]:ub 0x7f:w {NoDDChk} //
+#if defined(MBAFF) || defined(FIELD)
+ (f0.1) and (1) gREFPARITY:w r[pBIDX]:ub 0x80:w
+ (-f0.1) and (1) gREFPARITY:w r[pBIDX,4]:ub 0x80:w
+ shl (1) gREFPARITY:w gREFPARITY<0;1,0>:w 1:w
+#endif
+
+ // Sub MB shape
+ asr (1) gSHAPETEMP:w gSUBMB_SHAPE:ub gLOOP_SUBMB:w
+
+ // Chroma MV adjustment & Set message descriptor for frame/field read
+#if defined(MBAFF)
+ #include "chromaMVAdjust.asm"
+ and.nz.f0.0 (1) null:uw gFIELDMBFLAG:ub nFIELDMB_MASK:uw
+ (f0.0) add (1) gD0:ud gBIDX:uw nDWBRMSGDSC_SC_TF:ud
+ (-f0.0) add (1) gMSGDSC_R:ud gBIDX:uw nDWBRMSGDSC_SC:ud
+ (f0.0) add (1) gMSGDSC_R:ud gD0:ud gREFPARITY:uw
+#elif defined(FIELD)
+ #include "chromaMVAdjust.asm"
+ add (1) gMSGDSC_R:ud gBIDX:uw nDWBRMSGDSC_SC_TF:ud
+ add (1) gMSGDSC_R:ud gMSGDSC_R:ud gREFPARITY:uw
+#else // FRAME
+ add (1) gMSGDSC_R:ud gBIDX:uw nDWBRMSGDSC_SC:ud
+#endif
+
+ and.nz.f0.1 (1) null:w gSHAPETEMP:w 3:w
+ (f0.1) jmpi INTERLABEL(PROCESS4x4)
+
+ //======================== BEGIN - PROCESS 8x8 ===========================
+
+ // Reference block load
+ #include "loadRef_Y_16x13.asm"
+#ifndef MONO
+#if defined(MBAFF) || defined(FIELD)
+ add (1) r[pMV,2]:w r[pMV,2]:w gCHRMVADJ:w
+#endif
+ #include "loadRef_C_10x5.asm"
+#endif
+
+ // Interpolation
+ //CALL_INTER(INTERLABEL(Interpolate_Y_8x8_Func), 1)
+ #include "interpolate_Y_8x8.asm"
+#ifndef MONO
+ //CALL_INTER(INTERLABEL(Interpolate_C_4x4_Func), 1)
+ #include "interpolate_C_4x4.asm"
+#endif
+
+ jmpi INTERLABEL(ROUND_SHIFT_C)
+ //========================= END - PROCESS 8x8 ============================
+
+ //======================== BEGIN - LOOP_SUBMBPT ==========================
+INTERLABEL(PROCESS4x4):
+
+ mov (1) gLOOP_SUBMBPT:uw 4:uw // 4, 3, 2, 1
+INTERLABEL(LOOP_SUBMBPT):
+
+ // Reference block load
+ #include "loadRef_Y_16x9.asm"
+#ifndef MONO
+#if defined(MBAFF) || defined(FIELD)
+ add (1) r[pMV,2]:w r[pMV,2]:w gCHRMVADJ:w
+#endif
+ #include "loadRef_C_6x3.asm"
+#endif
+
+ // Interpolation
+ #include "interpolate_Y_4x4.asm"
+#ifndef MONO
+ #include "interpolate_C_2x2.asm"
+#endif
+
+ cmp.e.f0.0 (1) null:w gLOOP_SUBMBPT:uw 3:w
+ add.z.f0.1 (1) gLOOP_SUBMBPT:uw gLOOP_SUBMBPT:uw -1:w
+ add (1) pMV:w pMV:w 8:w
+ (-f0.0) add (1) gpINTP:ud gpINTP:ud 0x00080008:ud // 8 & 8
+ (f0.0) add (1) gpINTP:ud gpINTP:ud 0x00180038:ud // 24 & 56
+ (-f0.1) jmpi INTERLABEL(LOOP_SUBMBPT)
+
+ cmp.e.f0.1 null:w gLOOP_DIR:w 1:w
+ add (1) pMV:w pMV:w -32:w
+ (f0.1) mov (1) gpINTP:ud nOFFSET_INTP0:ud
+ (-f0.1) mov (1) gpINTP:ud nOFFSET_INTP1:ud
+
+ mov (1) pRESULT:uw gpINTPC:uw
+
+ //========================= END - LOOP_SUBMBPT ===========================
+
+INTERLABEL(ROUND_SHIFT_C):
+
+#ifndef MONO
+ #include "roundShift_C_4x4.asm"
+#endif
+
+INTERLABEL(LOOP_DIR_CONTINUE):
+
+ add.nz.f0.1 (1) gLOOP_DIR:uw gLOOP_DIR:uw -1:w
+ add (1) pMV:w pMV:w 4:w
+ (-f0.1) jmpi INTERLABEL(LOOP_DIR)
+ //=========================== END - LOOP_DIR =============================
+
+INTERLABEL(Weighted_Prediction):
+ #include "weightedPred.asm"
+
+ and.z.f0.1 (16) null<1>:w gLOOP_SUBMB<0;1,0>:uw 2:w
+
+ #include "recon_Y_8x8.asm"
+#ifndef MONO
+ #include "recon_C_4x4.asm"
+
+ (-f0.1) add (1) pERRORC:w pERRORC:w 48:w
+#endif
+
+ cmp.e.f0.1 (1) null:w gLOOP_SUBMB:uw 6:w
+ add (1) gLOOP_SUBMB:uw gLOOP_SUBMB:uw 2:w
+
+ add (1) pWGT_BIDX:ud pWGT_BIDX:ud 0x00100001:ud // 12 & 1
+ add (1) pMV:w pMV:w gMVSTEP:w
+
+ (-f0.1) jmpi INTERLABEL(LOOP_SUBMB)
+ //========================== END - LOOP_SUBMB ============================
+
+INTERLABEL(EXIT_LOOP):
+ #include "writeRecon_YC.asm"
+
+#ifdef SW_SCOREBOARD
+ wait n0:ud // Now wait for scoreboard to response
+ #include "Soreboard_update.asm" // scorboard update function
+#else
+// Check for write commit first if SW scoreboard is disabled
+ mov (1) gREG_WRITE_COMMIT_Y<1>:ud gREG_WRITE_COMMIT_Y<0;1,0>:ud // Make sure Y write is committed
+ mov (1) gREG_WRITE_COMMIT_UV<1>:ud gREG_WRITE_COMMIT_UV<0;1,0>:ud // Make sure U/V write is committed
+#endif
+
+// Terminate the thread
+//
+ END_THREAD
+
+
+//#include "Interpolate_Y_8x8_Func.asm"
+//#include "Interpolate_C_4x4_Func.asm"
+//#include "WeightedPred_Y_Func.asm"
+//#include "WeightedPred_C_Func.asm"
+
+
+.end_code
+
+.end_kernel
+
+
+//#endif // !defined(__AVCMCInter__)
diff --git a/shaders/h264/mc/AllAVC.asm b/shaders/h264/mc/AllAVC.asm new file mode 100644 index 0000000..a9149ff --- /dev/null +++ b/shaders/h264/mc/AllAVC.asm @@ -0,0 +1,426 @@ +/*
+ * All HWMC kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Kernel name: AllAVC.asm
+//
+// All HWMC kernels merged into this file
+//
+// $Revision: 2 $
+// $Date: 9/10/06 2:02a $
+//
+
+// Note: To enable SW scoreboard for ILK AVC kernels, simply toggle the HW_SCOREBOARD
+// and SW_SCOREBOARD definition as described below.
+//
+// ----------------------------------------------------
+// Main: ALLINTRA
+// ----------------------------------------------------
+
+#define COMBINED_KERNEL
+//#define ENABLE_ILDB
+
+// WA for *Stim tool issue, should be removed later
+
+#ifdef DEV_ILK
+#define INSTFACTOR 2 // 128-bit count as 2 instructions
+#else
+#define INSTFACTOR 1 // 128-bit is 1 instruction
+#endif // DEV_ILK
+
+#ifdef DEV_CTG
+ #define SW_SCOREBOARD // SW Scoreboard should be enabled for CTG and earlier
+ #undef HW_SCOREBOARD // HW Scoreboard should be disabled for CTG and earlier
+#else
+ #define HW_SCOREBOARD // HW Scoreboard should be enabled for ILK and beyond
+ #undef SW_SCOREBOARD // SW Scoreboard should be disabled for ILK and beyond
+#endif // DEV_CTG
+#include "export.inc"
+#if defined(_EXPORT)
+ #include "AllAVC_Export.inc"
+#elif defined(_BUILD)
+ #include "AllAVC.ich" // ISAasm dumped .exports
+ #include "AllAVC_Export.inc" // Keep jumping targets aligned, only for CTG and beyond
+ #include "AllAVC_Build.inc"
+#else
+#endif
+
+.kernel AllAVC
+
+// Build all intra prediction kernels
+//
+#ifdef INTRA_16x16_PAD_NENOP
+ $for(0; <INTRA_16x16_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef INTRA_16x16_PAD_NOP
+ $for(0; <INTRA_16x16_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "Intra_16x16.asm"
+
+#ifdef INTRA_8x8_PAD_NENOP
+ $for(0; <INTRA_8x8_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef INTRA_8x8_PAD_NOP
+ $for(0; <INTRA_8x8_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "Intra_8x8.asm"
+
+#ifdef INTRA_4x4_PAD_NENOP
+ $for(0; <INTRA_4x4_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef INTRA_4x4_PAD_NOP
+ $for(0; <INTRA_4x4_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "Intra_4x4.asm"
+
+#ifdef INTRA_PCM_PAD_NENOP
+ $for(0; <INTRA_PCM_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef INTRA_PCM_PAD_NOP
+ $for(0; <INTRA_PCM_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "Intra_PCM.asm"
+
+// Build FrameMB_Motion kernel
+//
+#define FRAME
+
+ #ifdef FRAME_MB_PAD_NENOP
+ $for(0; <FRAME_MB_PAD_NENOP; 1) {
+ nenop
+ }
+ #endif
+ #ifdef FRAME_MB_PAD_NOP
+ $for(0; <FRAME_MB_PAD_NOP; 1) {
+ nop
+ }
+ #endif
+ #include "AVCMCInter.asm"
+#undef FRAME
+
+// Build FieldMB_Motion kernel
+//
+#define FIELD
+
+ #ifdef FIELD_MB_PAD_NENOP
+ $for(0; <FIELD_MB_PAD_NENOP; 1) {
+ nenop
+ }
+ #endif
+ #ifdef FIELD_MB_PAD_NOP
+ $for(0; <FIELD_MB_PAD_NOP; 1) {
+ nop
+ }
+ #endif
+ #include "AVCMCInter.asm"
+#undef FIELD
+
+// Build MBAff_Motion kernel
+//
+#define MBAFF
+
+ #ifdef MBAFF_MB_PAD_NENOP
+ $for(0; <MBAFF_MB_PAD_NENOP; 1) {
+ nenop
+ }
+ #endif
+ #ifdef MBAFF_MB_PAD_NOP
+ $for(0; <MBAFF_MB_PAD_NOP; 1) {
+ nop
+ }
+ #endif
+ #include "AVCMCInter.asm"
+#undef MBAFF
+
+#ifdef SW_SCOREBOARD
+
+// SW scoreboard kernel for non-MBAFF
+//
+#ifdef SCOREBOARD_PAD_NENOP
+ $for(0; <SCOREBOARD_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef SCOREBOARD_PAD_NOP
+ $for(0; <SCOREBOARD_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "scoreboard.asm"
+
+// SW scoreboard kernel for MBAFF
+
+#ifdef SCOREBOARD_MBAFF_PAD_NENOP
+ $for(0; <SCOREBOARD_MBAFF_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef SCOREBOARD_MBAFF_PAD_NOP
+ $for(0; <SCOREBOARD_MBAFF_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "scoreboard_MBAFF.asm"
+
+#elif defined(HW_SCOREBOARD)
+
+// SetHWscoreboard kernel for non-MBAFF
+//
+#ifdef SETHWSCOREBOARD_PAD_NENOP
+ $for(0; <SETHWSCOREBOARD_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef SETHWSCOREBOARD_PAD_NOP
+ $for(0; <SETHWSCOREBOARD_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "SetHWScoreboard.asm"
+
+// SetHWscoreboard kernel for MBAFF
+
+#ifdef SETHWSCOREBOARD_MBAFF_PAD_NENOP
+ $for(0; <SETHWSCOREBOARD_MBAFF_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef SETHWSCOREBOARD_MBAFF_PAD_NOP
+ $for(0; <SETHWSCOREBOARD_MBAFF_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "SetHWScoreboard_MBAFF.asm"
+
+#endif // SW_SCOREBOARD
+
+#ifdef BSDRESET_PAD_NENOP
+ $for(0; <BSDRESET_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef BSDRESET_PAD_NOP
+ $for(0; <BSDRESET_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "BSDReset.asm"
+
+#ifdef DCRESETDUMMY_PAD_NENOP
+ $for(0; <DCRESETDUMMY_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef DCRESETDUMMY_PAD_NOP
+ $for(0; <DCRESETDUMMY_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "DCResetDummy.asm"
+
+#ifdef ENABLE_ILDB
+
+// Build all ILDB kernels
+//
+// Undefine some previous defined symbols since they will be re-defined/re-declared in ILDB kernels
+#undef A
+#undef B
+#undef p0
+#undef p1
+
+#define MSGPAYLOADB MSGPAYLOADB_ILDB
+#define MSGPAYLOADW MSGPAYLOADW_ILDB
+#define MSGPAYLOADD MSGPAYLOADD_ILDB
+#define MSGPAYLOADF MSGPAYLOADF_ILDB
+
+// < Frame ILDB >
+#define _PROGRESSIVE
+#define ILDB_LABEL(x) x##_ILDB_FRAME
+#ifdef AVC_ILDB_ROOT_Y_ILDB_FRAME_PAD_NENOP
+ $for(0; <AVC_ILDB_ROOT_Y_ILDB_FRAME_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_ROOT_Y_ILDB_FRAME_PAD_NOP
+ $for(0; <AVC_ILDB_ROOT_Y_ILDB_FRAME_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Root_Y.asm"
+
+#ifdef AVC_ILDB_CHILD_Y_ILDB_FRAME_PAD_NENOP
+ $for(0; <AVC_ILDB_CHILD_Y_ILDB_FRAME_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_CHILD_Y_ILDB_FRAME_PAD_NOP
+ $for(0; <AVC_ILDB_CHILD_Y_ILDB_FRAME_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Child_Y.asm"
+
+#ifdef AVC_ILDB_ROOT_UV_ILDB_FRAME_PAD_NENOP
+ $for(0; <AVC_ILDB_ROOT_UV_ILDB_FRAME_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_ROOT_UV_ILDB_FRAME_PAD_NOP
+ $for(0; <AVC_ILDB_ROOT_UV_ILDB_FRAME_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Root_UV.asm"
+
+#ifdef AVC_ILDB_CHILD_UV_ILDB_FRAME_PAD_NENOP
+ $for(0; <AVC_ILDB_CHILD_UV_ILDB_FRAME_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_CHILD_UV_ILDB_FRAME_PAD_NOP
+ $for(0; <AVC_ILDB_CHILD_UV_ILDB_FRAME_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Child_UV.asm"
+#undef ILDB_LABEL
+#undef _PROGRESSIVE
+
+// < Field ILDB >
+#define _FIELD
+#define ILDB_LABEL(x) x##_ILDB_FIELD
+#ifdef AVC_ILDB_ROOT_Y_ILDB_FIELD_PAD_NENOP
+ $for(0; <AVC_ILDB_ROOT_Y_ILDB_FIELD_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_ROOT_Y_ILDB_FIELD_PAD_NOP
+ $for(0; <AVC_ILDB_ROOT_Y_ILDB_FIELD_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Root_Field_Y.asm"
+
+#ifdef AVC_ILDB_CHILD_Y_ILDB_FIELD_PAD_NENOP
+ $for(0; <AVC_ILDB_CHILD_Y_ILDB_FIELD_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_CHILD_Y_ILDB_FIELD_PAD_NOP
+ $for(0; <AVC_ILDB_CHILD_Y_ILDB_FIELD_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Child_Field_Y.asm"
+
+#ifdef AVC_ILDB_ROOT_UV_ILDB_FIELD_PAD_NENOP
+ $for(0; <AVC_ILDB_ROOT_UV_ILDB_FIELD_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_ROOT_UV_ILDB_FIELD_PAD_NOP
+ $for(0; <AVC_ILDB_ROOT_UV_ILDB_FIELD_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Root_Field_UV.asm"
+
+#ifdef AVC_ILDB_CHILD_UV_ILDB_FIELD_PAD_NENOP
+ $for(0; <AVC_ILDB_CHILD_UV_ILDB_FIELD_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_CHILD_UV_ILDB_FIELD_PAD_NOP
+ $for(0; <AVC_ILDB_CHILD_UV_ILDB_FIELD_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Child_Field_UV.asm"
+#undef ILDB_LABEL
+#undef _FIELD
+
+// < MBAFF Frame ILDB >
+#define _MBAFF
+#define ILDB_LABEL(x) x##_ILDB_MBAFF
+#ifdef AVC_ILDB_ROOT_Y_ILDB_MBAFF_PAD_NENOP
+ $for(0; <AVC_ILDB_ROOT_Y_ILDB_MBAFF_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_ROOT_Y_ILDB_MBAFF_PAD_NOP
+ $for(0; <AVC_ILDB_ROOT_Y_ILDB_MBAFF_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Root_Mbaff_Y.asm"
+
+#ifdef AVC_ILDB_CHILD_Y_ILDB_MBAFF_PAD_NENOP
+ $for(0; <AVC_ILDB_CHILD_Y_ILDB_MBAFF_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_CHILD_Y_ILDB_MBAFF_PAD_NOP
+ $for(0; <AVC_ILDB_CHILD_Y_ILDB_MBAFF_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Child_Mbaff_Y.asm"
+
+#ifdef AVC_ILDB_ROOT_UV_ILDB_MBAFF_PAD_NENOP
+ $for(0; <AVC_ILDB_ROOT_UV_ILDB_MBAFF_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_ROOT_UV_ILDB_MBAFF_PAD_NOP
+ $for(0; <AVC_ILDB_ROOT_UV_ILDB_MBAFF_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Root_Mbaff_UV.asm"
+
+#ifdef AVC_ILDB_CHILD_UV_ILDB_MBAFF_PAD_NENOP
+ $for(0; <AVC_ILDB_CHILD_UV_ILDB_MBAFF_PAD_NENOP; 1) {
+ nenop
+ }
+#endif
+#ifdef AVC_ILDB_CHILD_UV_ILDB_MBAFF_PAD_NOP
+ $for(0; <AVC_ILDB_CHILD_UV_ILDB_MBAFF_PAD_NOP; 1) {
+ nop
+ }
+#endif
+ #include "AVC_ILDB_Child_Mbaff_UV.asm"
+#undef ILDB_LABEL
+#undef _MBAFF
+
+#endif // ENABLE_ILDB
+
+AllAVC_END:
+nop
+// End of AllAVC
+
+.end_code
+
+.end_kernel
+
diff --git a/shaders/h264/mc/AllAVCField.asm b/shaders/h264/mc/AllAVCField.asm new file mode 100644 index 0000000..88240c3 --- /dev/null +++ b/shaders/h264/mc/AllAVCField.asm @@ -0,0 +1,70 @@ +/*
+ * All field picture HWMC kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// 2857702934 // 0xAA551616 - GUID for Intra_16x16 luma prediction mode offsets
+// 0 // Offset to Intra_16x16 luma prediction mode 0
+// 9 // Offset to Intra_16x16 luma prediction mode 1
+// 19 // Offset to Intra_16x16 luma prediction mode 2
+// 42 // Offset to Intra_16x16 luma prediction mode 3
+// 2857699336 // 0xAA550808 - GUID for Intra_8x8 luma prediction mode offsets
+// 0 // Offset to Intra_8x8 luma prediction mode 0
+// 5 // Offset to Intra_8x8 luma prediction mode 1
+// 10 // Offset to Intra_8x8 luma prediction mode 2
+// 26 // Offset to Intra_8x8 luma prediction mode 3
+// 36 // Offset to Intra_8x8 luma prediction mode 4
+// 50 // Offset to Intra_8x8 luma prediction mode 5
+// 68 // Offset to Intra_8x8 luma prediction mode 6
+// 85 // Offset to Intra_8x8 luma prediction mode 7
+// 95 // Offset to Intra_8x8 luma prediction mode 8
+// 2857698308 // 0xAA550404 - GUID for Intra_4x4 luma prediction mode offsets
+// 0 // Offset to Intra_4x4 luma prediction mode 0
+// 2 // Offset to Intra_4x4 luma prediction mode 1
+// 4 // Offset to Intra_4x4 luma prediction mode 2
+// 16 // Offset to Intra_4x4 luma prediction mode 3
+// 23 // Offset to Intra_4x4 luma prediction mode 4
+// 32 // Offset to Intra_4x4 luma prediction mode 5
+// 45 // Offset to Intra_4x4 luma prediction mode 6
+// 59 // Offset to Intra_4x4 luma prediction mode 7
+// 66 // Offset to Intra_4x4 luma prediction mode 8
+// 2857700364 // 0xAA550C0C - GUID for intra chroma prediction mode offsets
+// 0 // Offset to intra chroma prediction mode 0
+// 30 // Offset to intra chroma prediction mode 1
+// 36 // Offset to intra chroma prediction mode 2
+// 41 // Offset to intra chroma prediction mode 3
+
+// Kernel name: AllAVCField.asm
+//
+// All field picture HWMC kernels merged into this file
+//
+// $Revision: 1 $
+// $Date: 4/13/06 4:35p $
+//
+
+// ----------------------------------------------------
+// Main: AllAVCField
+// ----------------------------------------------------
+
+#define ALLHWMC
+#define COMBINED_KERNEL
+
+.kernel AllAVCField
+
+ #include "Intra_PCM.asm"
+ #include "Intra_16x16.asm"
+ #include "Intra_8x8.asm"
+ #include "Intra_4x4.asm"
+ #include "scoreboard.asm"
+
+ #define FIELD
+ #include "AVCMCInter.asm"
+
+// End of AllAVCField
+
+.end_kernel
+
diff --git a/shaders/h264/mc/AllAVCFrame.asm b/shaders/h264/mc/AllAVCFrame.asm new file mode 100644 index 0000000..8871627 --- /dev/null +++ b/shaders/h264/mc/AllAVCFrame.asm @@ -0,0 +1,69 @@ +/*
+ * All frame picture HWMC kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// 2857702934 // 0xAA551616 - GUID for Intra_16x16 luma prediction mode offsets
+// 0 // Offset to Intra_16x16 luma prediction mode 0
+// 9 // Offset to Intra_16x16 luma prediction mode 1
+// 19 // Offset to Intra_16x16 luma prediction mode 2
+// 42 // Offset to Intra_16x16 luma prediction mode 3
+// 2857699336 // 0xAA550808 - GUID for Intra_8x8 luma prediction mode offsets
+// 0 // Offset to Intra_8x8 luma prediction mode 0
+// 5 // Offset to Intra_8x8 luma prediction mode 1
+// 10 // Offset to Intra_8x8 luma prediction mode 2
+// 26 // Offset to Intra_8x8 luma prediction mode 3
+// 36 // Offset to Intra_8x8 luma prediction mode 4
+// 50 // Offset to Intra_8x8 luma prediction mode 5
+// 68 // Offset to Intra_8x8 luma prediction mode 6
+// 85 // Offset to Intra_8x8 luma prediction mode 7
+// 95 // Offset to Intra_8x8 luma prediction mode 8
+// 2857698308 // 0xAA550404 - GUID for Intra_4x4 luma prediction mode offsets
+// 0 // Offset to Intra_4x4 luma prediction mode 0
+// 2 // Offset to Intra_4x4 luma prediction mode 1
+// 4 // Offset to Intra_4x4 luma prediction mode 2
+// 16 // Offset to Intra_4x4 luma prediction mode 3
+// 23 // Offset to Intra_4x4 luma prediction mode 4
+// 32 // Offset to Intra_4x4 luma prediction mode 5
+// 45 // Offset to Intra_4x4 luma prediction mode 6
+// 59 // Offset to Intra_4x4 luma prediction mode 7
+// 66 // Offset to Intra_4x4 luma prediction mode 8
+// 2857700364 // 0xAA550C0C - GUID for intra chroma prediction mode offsets
+// 0 // Offset to intra chroma prediction mode 0
+// 30 // Offset to intra chroma prediction mode 1
+// 36 // Offset to intra chroma prediction mode 2
+// 41 // Offset to intra chroma prediction mode 3
+
+// Kernel name: AllAVCFrame.asm
+//
+// All frame picture HWMC kernels merged into this file
+//
+// $Revision: 1 $
+// $Date: 4/13/06 4:35p $
+//
+
+// ----------------------------------------------------
+// Main: AllAVCFrame
+// ----------------------------------------------------
+
+#define ALLHWMC
+#define COMBINED_KERNEL
+
+.kernel AllAVCFrame
+
+ #include "Intra_PCM.asm"
+ #include "Intra_16x16.asm"
+ #include "Intra_8x8.asm"
+ #include "Intra_4x4.asm"
+ #include "scoreboard.asm"
+
+ #include "AVCMCInter.asm"
+
+// End of AllAVCFrame
+
+.end_kernel
+
diff --git a/shaders/h264/mc/AllAVCMBAFF.asm b/shaders/h264/mc/AllAVCMBAFF.asm new file mode 100644 index 0000000..1dd06ed --- /dev/null +++ b/shaders/h264/mc/AllAVCMBAFF.asm @@ -0,0 +1,70 @@ +/*
+ * All MBAFF frame picture HWMC kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// 2857702934 // 0xAA551616 - GUID for Intra_16x16 luma prediction mode offsets
+// 0 // Offset to Intra_16x16 luma prediction mode 0
+// 9 // Offset to Intra_16x16 luma prediction mode 1
+// 19 // Offset to Intra_16x16 luma prediction mode 2
+// 42 // Offset to Intra_16x16 luma prediction mode 3
+// 2857699336 // 0xAA550808 - GUID for Intra_8x8 luma prediction mode offsets
+// 0 // Offset to Intra_8x8 luma prediction mode 0
+// 5 // Offset to Intra_8x8 luma prediction mode 1
+// 10 // Offset to Intra_8x8 luma prediction mode 2
+// 26 // Offset to Intra_8x8 luma prediction mode 3
+// 36 // Offset to Intra_8x8 luma prediction mode 4
+// 50 // Offset to Intra_8x8 luma prediction mode 5
+// 68 // Offset to Intra_8x8 luma prediction mode 6
+// 85 // Offset to Intra_8x8 luma prediction mode 7
+// 95 // Offset to Intra_8x8 luma prediction mode 8
+// 2857698308 // 0xAA550404 - GUID for Intra_4x4 luma prediction mode offsets
+// 0 // Offset to Intra_4x4 luma prediction mode 0
+// 2 // Offset to Intra_4x4 luma prediction mode 1
+// 4 // Offset to Intra_4x4 luma prediction mode 2
+// 16 // Offset to Intra_4x4 luma prediction mode 3
+// 23 // Offset to Intra_4x4 luma prediction mode 4
+// 32 // Offset to Intra_4x4 luma prediction mode 5
+// 45 // Offset to Intra_4x4 luma prediction mode 6
+// 59 // Offset to Intra_4x4 luma prediction mode 7
+// 66 // Offset to Intra_4x4 luma prediction mode 8
+// 2857700364 // 0xAA550C0C - GUID for intra chroma prediction mode offsets
+// 0 // Offset to intra chroma prediction mode 0
+// 30 // Offset to intra chroma prediction mode 1
+// 36 // Offset to intra chroma prediction mode 2
+// 41 // Offset to intra chroma prediction mode 3
+
+// Kernel name: AllAVCMBAFF.asm
+//
+// All MBAFF frame picture HWMC kernels merged into this file
+//
+// $Revision: 1 $
+// $Date: 4/13/06 4:35p $
+//
+
+// ----------------------------------------------------
+// Main: AllAVCMBAFF
+// ----------------------------------------------------
+
+#define ALLHWMC
+#define COMBINED_KERNEL
+
+.kernel AllAVCMBAFF
+
+ #include "Intra_PCM.asm"
+ #include "Intra_16x16.asm"
+ #include "Intra_8x8.asm"
+ #include "Intra_4x4.asm"
+ #include "scoreboard.asm"
+
+ #define MBAFF
+ #include "AVCMCInter.asm"
+
+// End of AllAVCMBAFF
+
+.end_kernel
+
diff --git a/shaders/h264/mc/AllAVC_Build.inc b/shaders/h264/mc/AllAVC_Build.inc new file mode 100644 index 0000000..5bfb753 --- /dev/null +++ b/shaders/h264/mc/AllAVC_Build.inc @@ -0,0 +1,82 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+$table {
+AllAVC_END_IP/INSTFACTOR // Total instruction count
+#if (defined(SW_SCOREBOARD) || defined(HW_SCOREBOARD)) && defined(ENABLE_ILDB)
+// 23 // Total kernel count
+#elif defined(SW_SCOREBOARD) || defined(HW_SCOREBOARD)
+// 11 // Total kernel count
+#elif defined(ENABLE_ILDB)
+// 21 // Total kernel count
+#else
+// 11 // Total kernel count
+#endif
+INTRA_16x16_ENTRY/INSTFACTOR // Instruction offset to 'Intra_16x16'
+INTRA_8x8_ENTRY/INSTFACTOR // Instruction offset to 'Intra_8x8'
+INTRA_4x4_ENTRY/INSTFACTOR // Instruction offset to 'Intra_4x4'
+INTRA_PCM_ENTRY/INSTFACTOR // Instruction offset to 'Intra_PCM'
+FRAME_MB_ENTRY/INSTFACTOR // Instruction offset to 'FrameMB_Motion'
+FIELD_MB_ENTRY/INSTFACTOR // Instruction offset to 'FieldMB_Motion'
+MBAFF_MB_ENTRY/INSTFACTOR // Instruction offset to 'MBAff_Motion'
+#ifdef SW_SCOREBOARD
+SCOREBOARD_ENTRY/INSTFACTOR // Instruction offset to 'scoreboard'
+SCOREBOARD_MBAFF_ENTRY/INSTFACTOR // Instruction offset to 'scoreboard_MBAFF'
+#elif defined(HW_SCOREBOARD)
+SETHWSCOREBOARD_ENTRY/INSTFACTOR // Instruction offset to 'AVC_SetIntraDepend'
+SETHWSCOREBOARD_MBAFF_ENTRY/INSTFACTOR // Instruction offset to 'AVC_SetIntraDependMBAFF'
+#endif // SW_SCOREBOARD
+#ifdef ENABLE_ILDB
+AVC_ILDB_ROOT_Y_ILDB_FRAME_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Root_Y'
+AVC_ILDB_CHILD_Y_ILDB_FRAME_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Child_Y'
+AVC_ILDB_ROOT_UV_ILDB_FRAME_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Root_UV'
+AVC_ILDB_CHILD_UV_ILDB_FRAME_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Child_UV'
+AVC_ILDB_ROOT_Y_ILDB_FIELD_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Root_Field_Y'
+AVC_ILDB_CHILD_Y_ILDB_FIELD_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Child_Field_Y'
+AVC_ILDB_ROOT_UV_ILDB_FIELD_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Root_Field_UV'
+AVC_ILDB_CHILD_UV_ILDB_FIELD_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Child_Field_UV'
+AVC_ILDB_ROOT_Y_ILDB_MBAFF_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Root_Mbaff_Y'
+AVC_ILDB_CHILD_Y_ILDB_MBAFF_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Child_Mbaff_Y'
+AVC_ILDB_ROOT_UV_ILDB_MBAFF_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Root_Mbaff_UV'
+AVC_ILDB_CHILD_UV_ILDB_MBAFF_ENTRY/INSTFACTOR // Instruction offset to 'AVC_ILDB_Child_Mbaff_UV'
+#endif // ENABLE_ILDB
+BSDRESET_ENTRY/INSTFACTOR // Instruction offset to 'BSDReset'
+DCRESETDUMMY_ENTRY/INSTFACTOR // Instruction offset to 'DCResetDummy'
+
+// 0 // Instruction offset to Intra_4x4_luma_prediction_mode_0
+INTRA_4X4_HORIZONTAL_IP-INTRA_4X4_VERTICAL_IP // Instruction offset to Intra_4x4_luma_prediction_mode_1
+INTRA_4X4_DC_IP-INTRA_4X4_VERTICAL_IP // Instruction offset to Intra_4x4_luma_prediction_mode_2
+INTRA_4X4_DIAG_DOWN_LEFT_IP-INTRA_4X4_VERTICAL_IP // Instruction offset to Intra_4x4_luma_prediction_mode_3
+INTRA_4X4_DIAG_DOWN_RIGHT_IP-INTRA_4X4_VERTICAL_IP // Instruction offset to Intra_4x4_luma_prediction_mode_4
+INTRA_4X4_VERT_RIGHT_IP-INTRA_4X4_VERTICAL_IP // Instruction offset to Intra_4x4_luma_prediction_mode_5
+INTRA_4X4_HOR_DOWN_IP-INTRA_4X4_VERTICAL_IP // Instruction offset to Intra_4x4_luma_prediction_mode_6
+INTRA_4X4_VERT_LEFT_IP-INTRA_4X4_VERTICAL_IP // Instruction offset to Intra_4x4_luma_prediction_mode_7
+INTRA_4X4_HOR_UP_IP-INTRA_4X4_VERTICAL_IP // Instruction offset to Intra_4x4_luma_prediction_mode_8
+
+// 0 // Instruction offset to Intra_8x8_luma_prediction_mode_0
+INTRA_8X8_HORIZONTAL_IP-INTRA_8X8_VERTICAL_IP // Instruction offset to Intra_8x8_luma_prediction_mode_1
+INTRA_8X8_DC_IP-INTRA_8X8_VERTICAL_IP // Instruction offset to Intra_8x8_luma_prediction_mode_2
+INTRA_8X8_DIAG_DOWN_LEFT_IP-INTRA_8X8_VERTICAL_IP // Instruction offset to Intra_8x8_luma_prediction_mode_3
+INTRA_8X8_DIAG_DOWN_RIGHT_IP-INTRA_8X8_VERTICAL_IP // Instruction offset to Intra_8x8_luma_prediction_mode_4
+INTRA_8X8_VERT_RIGHT_IP-INTRA_8X8_VERTICAL_IP // Instruction offset to Intra_8x8_luma_prediction_mode_5
+INTRA_8X8_HOR_DOWN_IP-INTRA_8X8_VERTICAL_IP // Instruction offset to Intra_8x8_luma_prediction_mode_6
+INTRA_8X8_VERT_LEFT_IP-INTRA_8X8_VERTICAL_IP // Instruction offset to Intra_8x8_luma_prediction_mode_7
+INTRA_8X8_HOR_UP_IP-INTRA_8X8_VERTICAL_IP // Instruction offset to Intra_8x8_luma_prediction_mode_8
+
+// 0 // Instruction offset to Intra_16x16_luma_prediction_mode_0
+INTRA_16x16_HORIZONTAL_IP-INTRA_16x16_VERTICAL_IP // Instruction offset to Intra_16x16_luma_prediction_mode_1
+INTRA_16x16_DC_IP-INTRA_16x16_VERTICAL_IP // Instruction offset to Intra_16x16_luma_prediction_mode_2
+INTRA_16x16_PLANE_IP-INTRA_16x16_VERTICAL_IP // Instruction offset to Intra_16x16_luma_prediction_mode_3
+
+// 0 // Instruction offset to intra_chroma_prediction_mode_0
+INTRA_CHROMA_HORIZONTAL_IP-INTRA_CHROMA_DC_IP // Instruction offset to intra_chroma_prediction_mode_1
+INTRA_CHROMA_VERTICAL_IP-INTRA_CHROMA_DC_IP // Instruction offset to intra_chroma_prediction_mode_2
+INTRA_Chroma_PLANE_IP-INTRA_CHROMA_DC_IP // Instruction offset to intra_chroma_prediction_mode_3
+
+intra_Pred_4x4_Y_IP-ADD_ERROR_SB3_IP*0x100+intra_Pred_4x4_Y_IP-ADD_ERROR_SB2_IP*0x100+intra_Pred_4x4_Y_IP-ADD_ERROR_SB1_IP*0x100+intra_Pred_4x4_Y_IP-ADD_ERROR_SB0_IP // Instruction offset to intra_4x4_pred_module
+}
diff --git a/shaders/h264/mc/AllAVC_Export.inc b/shaders/h264/mc/AllAVC_Export.inc new file mode 100644 index 0000000..6bb3eff --- /dev/null +++ b/shaders/h264/mc/AllAVC_Export.inc @@ -0,0 +1,172 @@ +.export entry_point INTRA_16x16
+.export entry_point INTRA_8x8
+.export entry_point INTRA_4x4
+.export entry_point INTRA_PCM
+.export entry_point FRAME_MB
+.export entry_point FIELD_MB
+.export entry_point MBAFF_MB
+#ifdef SW_SCOREBOARD
+.export entry_point SCOREBOARD
+.export entry_point SCOREBOARD_MBAFF
+#elif defined(HW_SCOREBOARD)
+.export entry_point SETHWSCOREBOARD
+.export entry_point SETHWSCOREBOARD_MBAFF
+#endif // SW_SCOREBOARD
+
+#ifdef ENABLE_ILDB
+.export entry_point AVC_ILDB_ROOT_Y_ILDB_FRAME
+.export entry_point AVC_ILDB_CHILD_Y_ILDB_FRAME
+.export entry_point AVC_ILDB_ROOT_UV_ILDB_FRAME
+.export entry_point AVC_ILDB_CHILD_UV_ILDB_FRAME
+.export entry_point AVC_ILDB_ROOT_Y_ILDB_FIELD
+.export entry_point AVC_ILDB_CHILD_Y_ILDB_FIELD
+.export entry_point AVC_ILDB_ROOT_UV_ILDB_FIELD
+.export entry_point AVC_ILDB_CHILD_UV_ILDB_FIELD
+.export entry_point AVC_ILDB_ROOT_Y_ILDB_MBAFF
+.export entry_point AVC_ILDB_CHILD_Y_ILDB_MBAFF
+.export entry_point AVC_ILDB_ROOT_UV_ILDB_MBAFF
+.export entry_point AVC_ILDB_CHILD_UV_ILDB_MBAFF
+#endif // ENABLE_ILDB
+
+.export entry_point BSDRESET
+.export entry_point DCRESETDUMMY
+
+.export label INTRA_16x16_VERTICAL
+.export label INTRA_16x16_HORIZONTAL
+.export label INTRA_16x16_DC
+.export label INTRA_16x16_PLANE
+
+.export label INTRA_8X8_VERTICAL
+.export label INTRA_8X8_HORIZONTAL
+.export label INTRA_8X8_DC
+.export label INTRA_8X8_DIAG_DOWN_LEFT
+.export label INTRA_8X8_DIAG_DOWN_RIGHT
+.export label INTRA_8X8_VERT_RIGHT
+.export label INTRA_8X8_HOR_DOWN
+.export label INTRA_8X8_VERT_LEFT
+.export label INTRA_8X8_HOR_UP
+
+.export label INTRA_4X4_VERTICAL
+.export label INTRA_4X4_HORIZONTAL
+.export label INTRA_4X4_DC
+.export label INTRA_4X4_DIAG_DOWN_LEFT
+.export label INTRA_4X4_DIAG_DOWN_RIGHT
+.export label INTRA_4X4_VERT_RIGHT
+.export label INTRA_4X4_HOR_DOWN
+.export label INTRA_4X4_VERT_LEFT
+.export label INTRA_4X4_HOR_UP
+
+.export label INTRA_CHROMA_DC
+.export label INTRA_CHROMA_HORIZONTAL
+.export label INTRA_CHROMA_VERTICAL
+.export label INTRA_Chroma_PLANE
+
+.export label intra_Pred_4x4_Y
+.export label ADD_ERROR_SB0
+.export label ADD_ERROR_SB1
+.export label ADD_ERROR_SB2
+.export label ADD_ERROR_SB3
+
+.export label AllAVC_END
+
+#ifdef SW_SCOREBOARD
+.export label MB_Loop
+.export label No_Message
+.export label Dependency_Check
+.export label Notify_MSG
+.export label Update_CurMB
+.export label MBAFF_MB_Loop
+.export label MBAFF_No_Message
+.export label MBAFF_Dependency_Check
+.export label MBAFF_Notify_MSG
+.export label MBAFF_Update_CurMB
+
+//.export label
+
+// Definitions for first pass MC kernel building
+#ifndef No_Message_IP
+#define No_Message_IP 0
+#endif
+
+#ifndef Dependency_Check_IP
+#define Dependency_Check_IP 0
+#endif
+
+#ifndef Notify_MSG_IP
+#define Notify_MSG_IP 0
+#endif
+
+#ifndef Update_CurMB_IP
+#define Update_CurMB_IP 0
+#endif
+
+#ifndef MBAFF_No_Message_IP
+#define MBAFF_No_Message_IP 0
+#endif
+
+#ifndef MBAFF_Dependency_Check_IP
+#define MBAFF_Dependency_Check_IP 0
+#endif
+
+#ifndef MBAFF_Notify_MSG_IP
+#define MBAFF_Notify_MSG_IP 0
+#endif
+
+#ifndef AS_ENABLED
+ #ifndef MBAFF_MB_Loop_IP
+ #define MBAFF_MB_Loop_IP 0
+ #endif
+
+ #ifndef MB_Loop_IP
+ #define MB_Loop_IP 0
+ #endif
+#endif // End AS_ENABLED
+
+#ifndef MBAFF_Update_CurMB_IP
+#define MBAFF_Update_CurMB_IP 0
+#endif
+
+#endif // SW_SCOREBOARD
+/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#ifdef ENABLE_ILDB
+.export label ALL_SPAWNED_UV_ILDB_FRAME
+.export label SLEEP_ENTRY_UV_ILDB_FRAME
+.export label POST_SLEEP_UV_ILDB_FRAME
+.export label ALL_SPAWNED_Y_ILDB_FRAME
+.export label SLEEP_ENTRY_Y_ILDB_FRAME
+.export label POST_SLEEP_Y_ILDB_FRAME
+
+// Definitions for first pass ILDB kernel building
+#ifndef ALL_SPAWNED_UV_ILDB_FRAME_IP
+#define ALL_SPAWNED_UV_ILDB_FRAME_IP 0
+#endif
+
+#ifndef SLEEP_ENTRY_UV_ILDB_FRAME_IP
+#define SLEEP_ENTRY_UV_ILDB_FRAME_IP 0
+#endif
+
+#ifndef POST_SLEEP_UV_ILDB_FRAME_IP
+#define POST_SLEEP_UV_ILDB_FRAME_IP 0
+#endif
+
+#ifndef ALL_SPAWNED_Y_ILDB_FRAME_IP
+#define ALL_SPAWNED_Y_ILDB_FRAME_IP 0
+#endif
+
+#ifndef SLEEP_ENTRY_Y_ILDB_FRAME_IP
+#define SLEEP_ENTRY_Y_ILDB_FRAME_IP 0
+#endif
+
+#ifndef POST_SLEEP_Y_ILDB_FRAME_IP
+#define POST_SLEEP_Y_ILDB_FRAME_IP 0
+#endif
+
+#endif // ENABLE_ILDB
diff --git a/shaders/h264/mc/AllIntra.asm b/shaders/h264/mc/AllIntra.asm new file mode 100644 index 0000000..1cc895a --- /dev/null +++ b/shaders/h264/mc/AllIntra.asm @@ -0,0 +1,68 @@ +/*
+ * All intra-prediction macroblock kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// 2857702934 // 0xAA551616 - GUID for Intra_16x16 luma prediction mode offsets
+// 0 // Offset to Intra_16x16 luma prediction mode 0
+// 9 // Offset to Intra_16x16 luma prediction mode 1
+// 19 // Offset to Intra_16x16 luma prediction mode 2
+// 42 // Offset to Intra_16x16 luma prediction mode 3
+// 2857699336 // 0xAA550808 - GUID for Intra_8x8 luma prediction mode offsets
+// 0 // Offset to Intra_8x8 luma prediction mode 0
+// 5 // Offset to Intra_8x8 luma prediction mode 1
+// 10 // Offset to Intra_8x8 luma prediction mode 2
+// 26 // Offset to Intra_8x8 luma prediction mode 3
+// 36 // Offset to Intra_8x8 luma prediction mode 4
+// 50 // Offset to Intra_8x8 luma prediction mode 5
+// 68 // Offset to Intra_8x8 luma prediction mode 6
+// 85 // Offset to Intra_8x8 luma prediction mode 7
+// 95 // Offset to Intra_8x8 luma prediction mode 8
+// 2857698308 // 0xAA550404 - GUID for Intra_4x4 luma prediction mode offsets
+// 0 // Offset to Intra_4x4 luma prediction mode 0
+// 2 // Offset to Intra_4x4 luma prediction mode 1
+// 4 // Offset to Intra_4x4 luma prediction mode 2
+// 16 // Offset to Intra_4x4 luma prediction mode 3
+// 23 // Offset to Intra_4x4 luma prediction mode 4
+// 32 // Offset to Intra_4x4 luma prediction mode 5
+// 45 // Offset to Intra_4x4 luma prediction mode 6
+// 59 // Offset to Intra_4x4 luma prediction mode 7
+// 66 // Offset to Intra_4x4 luma prediction mode 8
+// 2857700364 // 0xAA550C0C - GUID for intra chroma prediction mode offsets
+// 0 // Offset to intra chroma prediction mode 0
+// 30 // Offset to intra chroma prediction mode 1
+// 36 // Offset to intra chroma prediction mode 2
+// 41 // Offset to intra chroma prediction mode 3
+
+// Kernel name: AllIntra.asm
+//
+// All HWMC kernels merged into this file
+//
+// $Revision: 1 $
+// $Date: 4/13/06 4:35p $
+//
+
+// ----------------------------------------------------
+// Main: ALLINTRA
+// ----------------------------------------------------
+
+#define ALLHWMC
+#define COMBINED_KERNEL
+
+.kernel ALLINTRA
+
+ // All frame destination HWMC kernels
+ //
+ #include "Intra_PCM.asm"
+ #include "Intra_16x16.asm"
+ #include "Intra_8x8.asm"
+ #include "Intra_4x4.asm"
+
+// End of ALLINTRA
+
+.end_kernel
+
diff --git a/shaders/h264/mc/BSDReset.asm b/shaders/h264/mc/BSDReset.asm new file mode 100644 index 0000000..5c6e5df --- /dev/null +++ b/shaders/h264/mc/BSDReset.asm @@ -0,0 +1,43 @@ +/*
+ * Initial kernel for filling initial BSD command buffer
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: BSDReset.asm
+//
+// Initial kernel for filling initial BSD command buffer
+//
+
+// ----------------------------------------------------
+// Main: BSDReset
+// ----------------------------------------------------
+
+.kernel BSDReset
+BSDRESET:
+
+#include "header.inc"
+
+.code
+#ifdef SW_SCOREBOARD
+ CALL(scoreboard_start_inter,1)
+ wait n0:ud // Now wait for scoreboard to response
+
+#define BSDRESET_ENABLE
+ #include "scoreboard_update.asm" // scorboard update function
+#undef BSDRESET_ENABLE
+
+#endif // defined(SW_SCOREBOARD)
+
+// Terminate the thread
+//
+ END_THREAD
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif // !defined(COMBINED_KERNEL)
diff --git a/shaders/h264/mc/DCResetDummy.asm b/shaders/h264/mc/DCResetDummy.asm new file mode 100644 index 0000000..d4e52a9 --- /dev/null +++ b/shaders/h264/mc/DCResetDummy.asm @@ -0,0 +1,34 @@ +/*
+ * Dummy kernel
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: DCResetDummy.asm
+//
+// Dummy kernel used by driver for debug-counter reset SW WA
+//
+
+// ----------------------------------------------------
+// Main: DCResetDummy
+// ----------------------------------------------------
+
+.kernel DCResetDummy
+DCRESETDUMMY:
+
+#include "header.inc"
+
+.code
+
+// Terminate the thread
+//
+ END_THREAD
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif // !defined(COMBINED_KERNEL)
diff --git a/shaders/h264/mc/Decode_Chroma_Intra.asm b/shaders/h264/mc/Decode_Chroma_Intra.asm new file mode 100644 index 0000000..7799825 --- /dev/null +++ b/shaders/h264/mc/Decode_Chroma_Intra.asm @@ -0,0 +1,29 @@ +/*
+ * Decode both intra chroma blocks
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__DECODE_CHROMA_INTRA__) // Make sure this is only included once
+#define __DECODE_CHROMA_INTRA__
+
+// Module name: Decode_Chroma_Intra.asm
+//
+// Decode both intra chroma blocks
+//
+
+decode_Chroma_Intra:
+#ifndef MONO
+ #include "load_Intra_Ref_UV.asm" // Load intra U/V reference data
+ #include "intra_Pred_Chroma.asm" // Intra predict chroma blocks
+ #include "add_Error_UV.asm" // Add error data to predicted U/V data blocks
+#endif // !defined(MONO)
+ #include "save_8x8_UV.asm" // Save to destination U/V frame surface
+
+ RETURN
+// End of Decode_Chroma_Intra
+
+#endif // !defined(__DECODE_CHROMA_INTRA__)
diff --git a/shaders/h264/mc/EndIntraThread.asm b/shaders/h264/mc/EndIntraThread.asm new file mode 100644 index 0000000..2c78b62 --- /dev/null +++ b/shaders/h264/mc/EndIntraThread.asm @@ -0,0 +1,30 @@ +/*
+ * Common module to end current intra thread
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: EndIntraThread.asm
+//
+// Common module to end current intra thread
+//
+#ifndef SW_SCOREBOARD
+// Check for write commit first if SW scoreboard is disabled
+ mov (1) REG_WRITE_COMMIT_Y<1>:ud REG_WRITE_COMMIT_Y<0;1,0>:ud // Make sure Y write is committed
+ mov (1) REG_WRITE_COMMIT_UV<1>:ud REG_WRITE_COMMIT_UV<0;1,0>:ud // Make sure U/V write is committed
+#endif
+
+ END_THREAD
+
+ #include "Intra_funcLib.asm"
+
+#ifndef COMBINED_KERNEL // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif // COMBINED_KERNEL
+
+// End of EndIntraThread
diff --git a/shaders/h264/mc/HwmcOnlyHeader.inc b/shaders/h264/mc/HwmcOnlyHeader.inc new file mode 100644 index 0000000..514cb78 --- /dev/null +++ b/shaders/h264/mc/HwmcOnlyHeader.inc @@ -0,0 +1,29 @@ +/*
+ * Header file used only in HWMC_ONLY mode
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: HwmcOnlyHeader.inc
+//
+// Header file used only in HWMC_ONLY mode
+//
+
+#include "header.inc"
+
+#if !defined(__HWMCONLYHEADER__) // Make sure the following are only included once
+#define __HWMCONLYHEADER__
+
+.reg_count_total 64
+.reg_count_payload 2
+
+//
+// Now, begin source code....
+//
+
+.code
+#endif // !defined(__HWMCONLYHEADER__)
+
diff --git a/shaders/h264/mc/Intra_16x16.asm b/shaders/h264/mc/Intra_16x16.asm new file mode 100644 index 0000000..e40e6a3 --- /dev/null +++ b/shaders/h264/mc/Intra_16x16.asm @@ -0,0 +1,71 @@ +/*
+ * Decode Intra_16x16 macroblock
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Intra_16x16.asm
+//
+// Decoding of Intra_16x16 macroblock
+//
+// $Revision: 8 $
+// $Date: 10/18/06 4:10p $
+//
+
+// ----------------------------------------------------
+// Main: Intra_16x16
+// ----------------------------------------------------
+
+#define INTRA_16X16
+
+.kernel Intra_16x16
+INTRA_16x16:
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+mov (1) acc0:ud 0x00aa55a5:ud
+#endif
+
+#include "SetupForHWMC.asm"
+
+#ifdef SW_SCOREBOARD
+ CALL(scoreboard_start_intra,1)
+#endif
+
+#ifdef SW_SCOREBOARD
+ wait n0:ud // Now wait for scoreboard to response
+#endif
+
+//
+// Decode Y blocks
+//
+// Load reference data from neighboring macroblocks
+ CALL(load_Intra_Ref_Y,1)
+
+// Intra predict Intra_16x16 luma block
+ #include "intra_pred_16x16_Y.asm"
+
+// Add error data to predicted intra data
+ #include "add_Error_16x16_Y.asm"
+
+// Save decoded Y picture
+ CALL(save_16x16_Y,1)
+//
+// Decode U/V blocks
+//
+// Note: The decoding for chroma blocks will be the same for all intra prediction mode
+//
+ CALL(decode_Chroma_Intra,1)
+
+#ifdef SW_SCOREBOARD
+ #include "scoreboard_update.asm"
+#endif
+
+// Terminate the thread
+//
+ #include "EndIntraThread.asm"
+
+// End of Intra_16x16
diff --git a/shaders/h264/mc/Intra_4x4.asm b/shaders/h264/mc/Intra_4x4.asm new file mode 100644 index 0000000..1169983 --- /dev/null +++ b/shaders/h264/mc/Intra_4x4.asm @@ -0,0 +1,175 @@ +/*
+ * Decode Intra_4x4 macroblock
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Intra_4x4.asm
+//
+// Decoding of Intra_4x4 macroblock
+//
+// $Revision: 12 $
+// $Date: 10/18/06 4:10p $
+//
+
+// ----------------------------------------------------
+// Main: Intra_4x4
+// ----------------------------------------------------
+
+#define INTRA_4X4
+
+.kernel Intra_4x4
+INTRA_4x4:
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+mov (1) acc0:ud 0x02aa55a5:ud
+#endif
+
+#include "SetupForHWMC.asm"
+
+#undef PPREDBUF_Y
+#define PPREDBUF_Y a0.3 // Pointer to predicted Y picture
+
+#define REG_INTRA_PRED_AVAIL REG_INTRA_TEMP_4
+#define REG_INTRA_4X4_PRED REG_INTRA_TEMP_7 // Store predicted Intra_4x4 data
+
+// Offset where 4x4 predicted data blocks are stored
+#define PREDSUBBLK0 0*GRFWIB
+#define PREDSUBBLK1 1*GRFWIB
+#define PREDSUBBLK2 2*GRFWIB
+#define PREDSUBBLK3 3*GRFWIB
+#define PREDSUBBLK4 4*GRFWIB
+#define PREDSUBBLK5 5*GRFWIB
+#define PREDSUBBLK6 6*GRFWIB
+#define PREDSUBBLK7 7*GRFWIB
+#define PREDSUBBLK8 8*GRFWIB
+#define PREDSUBBLK9 9*GRFWIB
+#define PREDSUBBLK10 10*GRFWIB
+#define PREDSUBBLK11 11*GRFWIB
+#define PREDSUBBLK12 12*GRFWIB
+#define PREDSUBBLK13 13*GRFWIB
+#define PREDSUBBLK14 14*GRFWIB
+#define PREDSUBBLK15 15*GRFWIB
+
+// 4x4 error block byte offset within the 8x8 error block
+#define ERRBLK0 0
+#define ERRBLK1 8
+#define ERRBLK2 64
+#define ERRBLK3 72
+
+#ifdef SW_SCOREBOARD
+ CALL(scoreboard_start_intra,1)
+#endif
+
+#ifdef SW_SCOREBOARD
+ wait n0:ud // Now wait for scoreboard to response
+#endif
+
+//
+// Decode Y blocks
+//
+// Load reference data from neighboring macroblocks
+ CALL(load_Intra_Ref_Y,1)
+
+ mov (1) PERROR<1>:w ERRBUF*GRFWIB:w // Pointer to macroblock error data
+ mov (1) PPREDBUF_Y<1>:w PREDBUF*GRFWIB:w // Pointer to predicted data
+ shr (2) REG_INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL_FLAG_BYTE<0;1,0>:ub 0x40:v
+ and.nz.f0.0 (8) NULLREG REG_INTRA_PRED_AVAIL_FLAG_BYTE<0;1,0>:ub 4:w // Top-right macroblock available for intra prediction?
+ (-f0.0.any8h) mov (8) INTRA_REF_TOP(0,16)<1> INTRA_REF_TOP(0,15)REGION(1,0) // Extend right boundary of MB B to C
+
+// Intra predict Intra_4x4 luma blocks
+//
+// Sub-macroblock 0 *****************
+ mov (16) REF_TOP0(0)<1> INTRA_REF_TOP0(0) // Top reference data
+ mov (8) REF_LEFT(0)<1> INTRA_REF_LEFT(0)REGION(8,4) // Left reference data
+ shr (4) PRED_MODE<1>:w INTRA_PRED_MODE(0)<1;2,0> 0x4040:v // Expand IntraPredMode to 1 byte/block
+ CALL(intra_Pred_4x4_Y_4,1)
+ add (1) PERROR<1>:w PERROR<0;1,0>:w 0x0080:w // Pointers to next error block
+
+ or (1) REG_INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w 0x1:w // Left neighbor is available now
+
+// Sub-macroblock 1 *****************
+
+ mov (16) REF_TOP0(0)<1> INTRA_REF_TOP0(0,8) // Top reference data
+ mov (4) REF_LEFT(0)<1> r[PPREDBUF_Y,PREDSUBBLK1+6]<8;1,0>:ub // Left reference data (top half)
+ mov (4) REF_LEFT(0,4)<1> r[PPREDBUF_Y,PREDSUBBLK3+6]<8;1,0>:ub // Left reference data (bottom half)
+ shr (4) PRED_MODE<1>:w INTRA_PRED_MODE(0,2)<1;2,0> 0x4040:v // Expand IntraPredMode to 1 byte/block
+ add (1) PPREDBUF_Y<1>:w PPREDBUF_Y<0;1,0>:w 4*GRFWIB:w // Pointer to predicted sub-macroblock 1
+ CALL(intra_Pred_4x4_Y_4,1)
+ add (1) PERROR<1>:w PERROR<0;1,0>:w 0x0080:w // Pointers to next error block
+
+ or (1) REG_INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL.1<0;1,0>:w 0x2:w // Top neighbor is available now
+
+// Pack constructed data from word-aligned to byte-aligned format
+// to speed up save_4x4_Y module later
+// PPREDBUF_Y now points to sub-block #4
+ mov (16) r[PPREDBUF_Y,-PREDSUBBLK4]<1>:ub r[PPREDBUF_Y,-PREDSUBBLK4]<32;16,2>:ub // Sub-block 0
+ mov (16) r[PPREDBUF_Y,0-PREDSUBBLK4+16]<1>:ub r[PPREDBUF_Y,-PREDSUBBLK3]<32;16,2>:ub // Sub-block 1
+ mov (16) r[PPREDBUF_Y,-PREDSUBBLK2]<1>:ub r[PPREDBUF_Y,-PREDSUBBLK2]<32;16,2>:ub // Sub-block 2
+ mov (16) r[PPREDBUF_Y,0-PREDSUBBLK2+16]<1>:ub r[PPREDBUF_Y,-PREDSUBBLK1]<32;16,2>:ub // Sub-block 3
+
+ mov (16) r[PPREDBUF_Y,-PREDSUBBLK3]<1>:ub r[PPREDBUF_Y]<32;16,2>:ub // Sub-block 4
+ mov (16) r[PPREDBUF_Y,0-PREDSUBBLK3+16]<1>:ub r[PPREDBUF_Y,PREDSUBBLK1]<32;16,2>:ub // Sub-block 5
+ mov (16) r[PPREDBUF_Y,-PREDSUBBLK1]<1>:ub r[PPREDBUF_Y,PREDSUBBLK2]<32;16,2>:ub // Sub-block 6
+ mov (16) r[PPREDBUF_Y,0-PREDSUBBLK1+16]<1>:ub r[PPREDBUF_Y,PREDSUBBLK3]<32;16,2>:ub // Sub-block 7
+
+// Sub-macroblock 2 *****************
+
+ mov (4) REF_TOP0(0)<1> INTRA_REF_LEFT0(0,28)REGION(4,1) // Top-left reference data
+ mov (8) REF_TOP0(0,4)<1> r[PPREDBUF_Y,0-2*GRFWIB+12]<16;4,1>:ub // Top reference data from SB 2,3
+ mov (8) REF_TOP0(0,12)<1> r[PPREDBUF_Y,0-GRFWIB+12]<16;4,1>:ub // Top reference data from SB 6,7
+ mov (8) REF_TOP0(0,20)<1> r[PPREDBUF_Y,0-GRFWIB+31]<0;1,0>:ub // Top-right reference data
+ mov (16) REG_INTRA_REF_TOP<1>:w REF_TOP_W(0) // Store top reference data for SubMB #2 and #3.
+ mov (8) REF_LEFT(0)<1> INTRA_REF_LEFT(1)REGION(8,4) // Left reference data
+ shr (4) PRED_MODE<1>:w INTRA_PRED_MODE(0,4)<1;2,0> 0x4040:v // Expand IntraPredMode to 1 byte/block
+ CALL(intra_Pred_4x4_Y_4,1)
+ add (1) PERROR<1>:w PERROR<0;1,0>:w 0x0080:w // Pointers to next error block
+
+ or (1) REG_INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w 0x1:w // Left neighbor is available now
+
+// Sub-macroblock 3 *****************
+
+ mov (16) REF_TOP0(0)<1> INTRA_REF_TOP0(0,8) // Top reference data
+ mov (8) REF_TOP0(0,16)<1> INTRA_REF_TOP0(0,24)REGION(8,1) // Top reference data
+ mov (4) REF_LEFT(0)<1> r[PPREDBUF_Y,PREDSUBBLK1+6]<8;1,0>:ub // Left reference data (top half)
+ mov (4) REF_LEFT(0,4)<1> r[PPREDBUF_Y,PREDSUBBLK3+6]<8;1,0>:ub // Left reference data (bottom half)
+ shr (4) PRED_MODE<1>:w INTRA_PRED_MODE(0,6)<1;2,0> 0x4040:v // Expand IntraPredMode to 1 byte/block
+ add (1) PPREDBUF_Y<1>:w PPREDBUF_Y<0;1,0>:w 4*GRFWIB:w // Pointer to predicted sub-macroblock 3
+ CALL(intra_Pred_4x4_Y_4,1)
+
+// Pack constructed data from word-aligned to byte-aligned format
+// to speed up save_4x4_Y module later
+// PPREDBUF_Y now points to sub-block #12
+ mov (16) r[PPREDBUF_Y,-PREDSUBBLK4]<1>:ub r[PPREDBUF_Y,-PREDSUBBLK4]<32;16,2>:ub // Sub-block 8
+ mov (16) r[PPREDBUF_Y,0-PREDSUBBLK4+16]<1>:ub r[PPREDBUF_Y,-PREDSUBBLK3]<32;16,2>:ub // Sub-block 9
+ mov (16) r[PPREDBUF_Y,-PREDSUBBLK2]<1>:ub r[PPREDBUF_Y,-PREDSUBBLK2]<32;16,2>:ub // Sub-block 10
+ mov (16) r[PPREDBUF_Y,0-PREDSUBBLK2+16]<1>:ub r[PPREDBUF_Y,-PREDSUBBLK1]<32;16,2>:ub // Sub-block 11
+
+ mov (16) r[PPREDBUF_Y,-PREDSUBBLK3]<1>:ub r[PPREDBUF_Y]<32;16,2>:ub // Sub-block 12
+ mov (16) r[PPREDBUF_Y,0-PREDSUBBLK3+16]<1>:ub r[PPREDBUF_Y,PREDSUBBLK1]<32;16,2>:ub // Sub-block 13
+ mov (16) r[PPREDBUF_Y,-PREDSUBBLK1]<1>:ub r[PPREDBUF_Y,PREDSUBBLK2]<32;16,2>:ub // Sub-block 14
+ mov (16) r[PPREDBUF_Y,0-PREDSUBBLK1+16]<1>:ub r[PPREDBUF_Y,PREDSUBBLK3]<32;16,2>:ub // Sub-block 15
+
+// All 4 sub-macroblock (containing 4 intra_4x4 blocks) have be constructed
+// Save constructed Y picture
+ CALL(save_4x4_Y,1) // Save Intra_4x4 predicted luma data.
+//
+// Decode U/V blocks
+//
+// Note: The decoding for chroma blocks will be the same for all intra prediction mode
+//
+ CALL(decode_Chroma_Intra,1)
+
+#ifdef SW_SCOREBOARD
+ #include "scoreboard_update.asm"
+#endif
+
+// Terminate the thread
+//
+ #include "EndIntraThread.asm"
+
+// End of Intra_4x4
diff --git a/shaders/h264/mc/Intra_8x8.asm b/shaders/h264/mc/Intra_8x8.asm new file mode 100644 index 0000000..05a0be5 --- /dev/null +++ b/shaders/h264/mc/Intra_8x8.asm @@ -0,0 +1,192 @@ +/*
+ * Decode Intra_8x8 macroblock
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Intra_8x8.asm
+//
+// Decoding of Intra_8x8 macroblock
+//
+// $Revision: 12 $
+// $Date: 10/18/06 4:10p $
+//
+
+// ----------------------------------------------------
+// Main: Intra_8x8
+// ----------------------------------------------------
+
+#define INTRA_8X8
+
+.kernel Intra_8x8
+INTRA_8x8:
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+mov (1) acc0:ud 0x01aa55a5:ud
+#endif
+
+#include "SetupForHWMC.asm"
+
+#define REG_INTRA_PRED_AVAIL REG_INTRA_TEMP_4
+#define INTRA_PRED_AVAIL REG_INTRA_TEMP_4.4
+
+// Offset where 8x8 predicted data blocks are stored
+#define PREDBLK0 0*GRFWIB
+#define PREDBLK1 4*GRFWIB
+#define PREDBLK2 8*GRFWIB
+#define PREDBLK3 12*GRFWIB
+
+#ifdef SW_SCOREBOARD
+
+// Update "E" flag with "F" flag information
+ mov (1) REG_INTRA_TEMP_0<1>:w REG_INTRA_PRED_AVAIL_FLAG_WORD<0;1,0>:w // Store original Intra_Pred_Avail_Flag
+ and.nz.f0.0 (1) NULLREG REG_MBAFF_PIC MBAFF_PIC // Is current MBAFF picture
+ and.z.f0.1 (1) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_TH_AVAIL_FLAG // Is "A" not available?
+ (f0.0) and.z.f0.0 (1) NULLREG REG_FIELD_MACROBLOCK_FLAG FIELD_MACROBLOCK_FLAG // Is current frame MB?
+ (f0.1) and.nz.f0.1 (1) NULLREG REG_INTRA_PRED_8X8_BLK2_AVAIL_FLAG INTRA_PRED_8X8_BLK2_AVAIL_FLAG // Is "F" flag set?
+ (f0.0.allv) or (1) REG_INTRA_PRED_AVAIL_FLAG_WORD<1>:w REG_INTRA_PRED_AVAIL_FLAG_WORD<0;1,0>:w INTRA_PRED_LEFT_BH_AVAIL_FLAG // Set "E" to 1 if all conditions meet
+
+ CALL(scoreboard_start_intra,1)
+ mov (1) REG_INTRA_PRED_AVAIL_FLAG_WORD<1>:w REG_INTRA_TEMP_0<0;1,0>:w // Restore original Intra_Pred_Avail_Flag
+#endif
+
+#ifdef SW_SCOREBOARD
+ wait n0:ud // Now wait for scoreboard to response
+#endif
+
+//
+// Decode Y blocks
+//
+// Load reference data from neighboring macroblocks
+ CALL(load_Intra_Ref_Y,1)
+
+ mov (1) PERROR<1>:w ERRBUF*GRFWIB:w // Pointer to macroblock error data
+ mov (1) PDECBUF_UD<1>:ud 0x00010001*PREDBUF*GRFWIB+0x00100000:ud // Pointers to predicted data
+ shr (2) REG_INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL_FLAG_BYTE<0;1,0>:ub 0x40:v
+
+#if 1
+ mov (4) REF_LEFT_D(0,0)<1> 0:ud // This is to make validation easier. Without it, DRAM mismatch occurs.
+#endif
+
+// Intra predict Intra_8x8 luma blocks
+//
+// Sub-macroblock 0 *****************
+ mov (16) REF_TOP_W(0)<1> REG_INTRA_REF_TOP<16;16,1>:w // Copy entire top reference data
+ and.nz.f0.0 (8) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_UP_LEFT_AVAIL_FLAG // Is "D" available?
+ (-f0.0) mov (1) REF_TOP(0,-1)<1> INTRA_REF_TOP(0)REGION(1,0) // p[-1,-1] = p[0,-1]
+
+ mov (8) REF_LEFT(0,2)<1> INTRA_REF_LEFT(0) // Left reference data, (leave 2 for reference filtering)
+ (-f0.0) mov (1) REF_LEFT(0,1)<1> INTRA_REF_LEFT(0)REGION(1,0) // p[-1,-1]=p[-1,0]
+ (f0.0.any2h) mov (2) REF_LEFT(0,0)<1> INTRA_REF_TOP(0,-1)REGION(1,0) // p'[-1,y] (y=0,1) = p[-1,-1]
+ and.nz.f0.1 (1) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_UP_AVAIL_FLAG // Is "B" available?
+ (f0.1) mov (1) REF_LEFT(0,0)<1> INTRA_REF_TOP(0,0)REGION(1,0) // p[0,-1] for left filtering
+ and.nz.f0.1 (1) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_TH_AVAIL_FLAG // Is "A" available?
+ (-f0.1) mov (1) REF_LEFT(0,2)<1> INTRA_REF_TOP(0,-1)REGION(1,0) // p'[-1,2] = p[-1,-1]
+
+ and (1) PRED_MODE<1>:w INTRA_PRED_MODE(0)REGION(1,0) 0x0F:w // Intra pred mode for current block
+ mov (1) INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w // Top/Left neighbor available flags
+ CALL(intra_Pred_8x8_Y,1)
+ add (1) PERROR<1>:w PERROR<0;1,0>:w 0x0080:w // Pointers to next error block
+
+// Sub-macroblock 1 *****************
+ mov (16) REF_TOP0(0)<1> INTRA_REF_TOP(0,4) // Top reference data
+ and.nz.f0.1 (8) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_UP_RIGHT_AVAIL_FLAG // Is "C" available?
+ (f0.1.any8h) mov (8) REF_TOP(0,8)<1> INTRA_REF_TOP(0,16)<8;8,1> // Take data from "C"
+ (-f0.1.any8h) mov (8) REF_TOP(0,8)<1> INTRA_REF_TOP(0,15)REGION(1,0) // Repeat last pixel from "B"
+
+ mov (4) REF_LEFT(0,2)<1> DEC_Y(0,14)<16;1,0> // Left reference data (top half) (leave 2 for reference filtering)
+ mov (4) REF_LEFT(0,6)<1> DEC_Y(2,14)<16;1,0> // Left reference data (bottom half)
+ mov (2) REF_LEFT(0,0)<1> INTRA_REF_TOP(0,7)REGION(1,0) // p'[-1,y] (y=0,1) = p[-1,-1]
+ and.nz.f0.1 (1) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_UP_AVAIL_FLAG // Is "B" available?
+ (f0.1) mov (1) REF_LEFT(0,0)<1> INTRA_REF_TOP(0,8)REGION(1,0) // p[-1,-1] for left filtering
+ (-f0.1) mov (1) REF_LEFT(0,1)<1> DEC_Y(0,14)REGION(1,0) // p[-1,-1] = p[-1,0]
+
+ shr (1) PRED_MODE<1>:w INTRA_PRED_MODE(0)REGION(1,0) 4:w // Intra pred mode for current block
+ add (2) PPREDBUF_Y<1>:w PPREDBUF_Y<2;2,1>:w 4*GRFWIB:w // Pointer to predicted sub-macroblock 1
+ or (1) INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w 1:w // Left neighbor is available
+ CALL(intra_Pred_8x8_Y,1)
+ add (1) PERROR<1>:w PERROR<0;1,0>:w 0x0080:w // Pointers to next error block
+
+// Pack constructed data from word-aligned to byte-aligned format and interlace Y0 and Y1(every two Y rows)
+// to speed up save_8x8_Y module later
+// PPREDBUF_Y now points to sub-macroblock Y1
+ mov (32) r[PPREDBUF_Y,-PREDBLK1]<1>:ub DEC_Y(0)<32;16,2> {Compr} // First 4 Y0 rows
+ mov (32) r[PPREDBUF_Y,0-PREDBLK1+32]<1>:ub DEC_Y(4)<32;16,2> {Compr} // First 4 Y1 rows
+ mov (32) r[PPREDBUF_Y,0-PREDBLK1+64]<1>:ub DEC_Y(2)<32;16,2> {Compr} // Second 4 Y0 rows
+ mov (32) r[PPREDBUF_Y,0-PREDBLK1+96]<1>:ub DEC_Y(6)<32;16,2> {Compr} // Second 4 Y1 rows
+
+// Sub-macroblock 2 *****************
+// Intra_8x8 special available flag handling
+ and.nz.f0.0 (1) NULLREG REG_MBAFF_PIC MBAFF_PIC // Is current MBAFF picture
+ and.z.f0.1 (1) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_TH_AVAIL_FLAG // Is "A" not available?
+ (f0.0) and.z.f0.0 (1) NULLREG REG_FIELD_MACROBLOCK_FLAG FIELD_MACROBLOCK_FLAG // Is current frame MB?
+ (f0.1) and.nz.f0.1 (1) NULLREG REG_INTRA_PRED_8X8_BLK2_AVAIL_FLAG INTRA_PRED_8X8_BLK2_AVAIL_FLAG // Is special intra_8x8 available flag set?
+ (f0.0.allv) mov (1) REF_TOP(0,-1)<1> INTRA_REF_LEFT0(0,31)REGION(1,0) // Top-left reference data
+ (f0.0.allv) jmpi (1) INTRA_8x8_BLK2
+// Done intra_8x8 special available flag handling
+
+ and.nz.f0.0 (8) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_TH_AVAIL_FLAG // Is top-half "A" available?
+ (f0.0.any4h) mov (4) REF_TOP0(0)<1> INTRA_REF_LEFT0(0,28)REGION(4,1) // Top-left reference data
+ (-f0.0) mov (1) REF_TOP(0,-1)<1> DEC_Y(2,24)REGION(1,0) // p[-1,-1] = p[0,-1]
+INTRA_8x8_BLK2:
+ mov (8) REF_TOP(0)<1> DEC_Y(2,24)REGION(8,1) // Top reference data
+ mov (8) REF_TOP(0,8)<1> DEC_Y(3,24)REGION(8,1) // Top reference data
+
+ mov (8) REF_LEFT(0,2)<1> INTRA_REF_LEFT(1) // Left reference data, (leave 2 for reference filtering)
+ mov (1) REF_LEFT(0,0)<1> DEC_Y(2,24)REGION(1,0) // p'[-1,0] = p[0,-1] since "B" is always available
+ (f0.0) mov (1) REF_LEFT(0,1)<1> INTRA_REF_LEFT(0,28)REGION(1,0) // p[-1,1] = p[-1,-1] if top-half "A" available
+ (-f0.0) mov (1) REF_LEFT(0,1)<1> INTRA_REF_LEFT(1)REGION(1,0) // p[-1,1] = p[-1,0]
+ and.nz.f0.1 (1) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_BH_AVAIL_FLAG // Is bottom-half "A" available?
+ (-f0.1) mov (1) REF_LEFT(0,2)<1> INTRA_REF_LEFT(0,28)REGION(1,0) // p'[-1,2] = p[-1,-1]
+
+ and (1) PRED_MODE<1>:w INTRA_PRED_MODE(0,1)REGION(1,0) 0x0F:w // Intra pred mode for current block
+ or (1) INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL.1<0;1,0>:w 2:w // Top neighbor is available
+ CALL(intra_Pred_8x8_Y,1)
+ add (1) PERROR<1>:w PERROR<0;1,0>:w 0x0080:w // Pointers to next error block
+
+// Sub-macroblock 3 *****************
+ mov (4) REF_TOP0(0)<1> DEC_Y(2,28)REGION(4,1) // Top-left reference data
+ mov (8) REF_TOP(0)<1> DEC_Y(3,24)REGION(8,1) // Top reference data
+ mov (16) REF_TOP(0,8)<1> DEC_Y(3,31)REGION(1,0) // Top-right reference data
+
+ mov (4) REF_LEFT(0,2)<1> DEC_Y(4,14)<16;1,0> // Left reference data (top half) (leave 2 for reference filtering)
+ mov (4) REF_LEFT(0,6)<1> DEC_Y(6,14)<16;1,0> // Left reference data (bottom half)
+ mov (1) REF_LEFT(0,0)<1> DEC_Y(3,24)REGION(1,0) // p[-1,0] = p[0,-1]
+ mov (1) REF_LEFT(0,1)<1> DEC_Y(2,31)REGION(1,0) // p[-1,1] = p[-1,-1]
+
+ shr (1) PRED_MODE<1>:w INTRA_PRED_MODE(0,1)REGION(1,0) 4:w // Intra pred mode for current block
+ add (2) PPREDBUF_Y<1>:w PPREDBUF_Y<2;2,1>:w 4*GRFWIB:w // Pointer to predicted sub-macroblock 3
+ or (1) INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w 3:w // Top and Left neighbor are available
+ CALL(intra_Pred_8x8_Y,1)
+
+// Pack constructed data from word-aligned to byte-aligned format
+// to speed up save_8x8_Y module later
+// PPREDBUF_Y now points to sub-macroblock Y1
+ mov (32) r[PPREDBUF_Y,-PREDBLK1]<1>:ub DEC_Y(4)<32;16,2> {Compr} // First 4 Y2 rows
+ mov (32) r[PPREDBUF_Y,0-PREDBLK1+32]<1>:ub DEC_Y(8)<32;16,2> {Compr} // First 4 Y3 rows
+ mov (32) r[PPREDBUF_Y,0-PREDBLK1+64]<1>:ub DEC_Y(6)<32;16,2> {Compr} // Second 4 Y2 rows
+ mov (32) r[PPREDBUF_Y,0-PREDBLK1+96]<1>:ub DEC_Y(10)<32;16,2> {Compr} // Second 4 Y3 rows
+
+// All 4 sub-macroblock (containing 4 intra_8x8 blocks) have be constructed
+// Save constructed Y picture
+ CALL(save_8x8_Y,1) // Save Intra_8x8 predicted luma data.
+//
+// Decode U/V blocks
+//
+// Note: The decoding for chroma blocks will be the same for all intra prediction mode
+//
+ CALL(decode_Chroma_Intra,1)
+
+#ifdef SW_SCOREBOARD
+ #include "scoreboard_update.asm"
+#endif
+
+// Terminate the thread
+//
+ #include "EndIntraThread.asm"
+
+// End of Intra_8x8
diff --git a/shaders/h264/mc/Intra_PCM.asm b/shaders/h264/mc/Intra_PCM.asm new file mode 100644 index 0000000..6bc81af --- /dev/null +++ b/shaders/h264/mc/Intra_PCM.asm @@ -0,0 +1,56 @@ +/*
+ * Decode Intra_PCM macroblock
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Intra_PCM.asm
+//
+// Decoding of I_PCM macroblock
+//
+// $Revision: 8 $
+// $Date: 10/18/06 4:10p $
+//
+
+// ----------------------------------------------------
+// Main: Intra_PCM
+// ----------------------------------------------------
+
+.kernel Intra_PCM
+INTRA_PCM:
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+mov (1) acc0:ud 0x03aa55a5:ud
+#endif
+
+#include "SetupForHWMC.asm"
+
+// Not actually needed here but just want to slow down the Intra-PCM to avoid race condition
+//
+#ifdef SW_SCOREBOARD
+ and (1) REG_INTRA_PRED_AVAIL_FLAG_WORD<1>:w REG_INTRA_PRED_AVAIL_FLAG_WORD<0;1,0>:w 0xffe0:w // Ensure all neighbor avail flags are "0"
+ CALL(scoreboard_start_intra,1)
+ wait n0:ud // Now wait for scoreboard to response
+#endif
+
+//
+// Decoding Y blocks
+//
+// In I_PCM mode, the samples are already arranged in raster scan order within the macroblock.
+// We just need to save them to picture buffers
+//
+ #include "save_I_PCM.asm" // Save to destination picture buffers
+
+#ifdef SW_SCOREBOARD
+ #include "scoreboard_update.asm"
+#endif
+
+// Terminate the thread
+//
+ #include "EndIntraThread.asm"
+
+// End of Intra_PCM
diff --git a/shaders/h264/mc/Intra_funcLib.asm b/shaders/h264/mc/Intra_funcLib.asm new file mode 100644 index 0000000..9644f7f --- /dev/null +++ b/shaders/h264/mc/Intra_funcLib.asm @@ -0,0 +1,42 @@ +/*
+ * Library of common modules shared among different intra prediction kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: Intra_funcLib.asm
+//
+// Library of common modules shared among different intra prediction kernels
+//
+// Note: Any sub-modules, if they are #included in more than one kernel,
+// should be moved to this module.
+//
+#if defined(INTRA_16X16)
+#undef INTRA_16X16
+ #include "load_Intra_Ref_Y.asm" // Load intra Y reference data
+ #include "Decode_Chroma_Intra.asm" // Decode chroma blocks
+ #include "save_16x16_Y.asm" // Save to destination Y frame surface
+#elif defined(INTRA_8X8)
+#undef INTRA_8X8
+ #include "load_Intra_Ref_Y.asm" // Load intra Y reference data
+ #include "Decode_Chroma_Intra.asm" // Decode chroma blocks
+ #include "intra_Pred_8x8_Y.asm" // Intra predict Intra_4x4 blocks
+ #include "save_8x8_Y.asm" // Save to destination Y frame surface
+#elif defined(INTRA_4X4)
+#undef INTRA_4X4
+ #include "load_Intra_Ref_Y.asm" // Load intra Y reference data
+ #include "Decode_Chroma_Intra.asm" // Decode chroma blocks
+ #include "intra_Pred_4x4_Y_4.asm" // Intra predict Intra_4x4 blocks
+ #include "save_4x4_Y.asm" // Save to destination Y frame surface
+#else // For all merged kernels
+#endif
+
+#ifdef SW_SCOREBOARD
+ #include "scoreboard_start_intra.asm" // scorboard intra start function
+ #include "scoreboard_start_inter.asm" // scorboard inter start function
+#endif // SW_SCOREBOARD
+
+// End of Intra_funcLib
diff --git a/shaders/h264/mc/Makefile.am b/shaders/h264/mc/Makefile.am new file mode 100644 index 0000000..9f97eb0 --- /dev/null +++ b/shaders/h264/mc/Makefile.am @@ -0,0 +1,28 @@ + +INTEL_G4I = + +INTEL_G4A = null.g4a + +INTEL_G4B = null.g4b + +INTEL_G4B_GEN5 = null.g4b.gen5 + +EXTRA_DIST = $(INTEL_G4I) \ + $(INTEL_G4A) \ + $(INTEL_G4B) \ + $(INTEL_G4B_GEN5) + +if HAVE_GEN4ASM + +SUFFIXES = .g4a .g4b +.g4a.g4b: + m4 $*.g4a > $*.g4m && intel-gen4asm -o $@ $*.g4m && intel-gen4asm -g 5 -o $@.gen5 $*.g4m && rm $*.g4m + +$(INTEL_G4B): $(INTEL_G4I) + +BUILT_SOURCES= $(INTEL_G4B) + +clean-local: + -rm -f $(INTEL_G4B) + -rm -f $(INTEL_G4B_GEN5) +endif diff --git a/shaders/h264/mc/Scoreboard_header.inc b/shaders/h264/mc/Scoreboard_header.inc new file mode 100644 index 0000000..5e87275 --- /dev/null +++ b/shaders/h264/mc/Scoreboard_header.inc @@ -0,0 +1,85 @@ +/*
+ * Common header file for both scoreboard and scoreboard_MBAFF kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__SCOREBOARD_HEADER__) // Make sure this file is only included once
+#define __SCOREBOARD_HEADER__
+
+// Module name: scoreboard_header.inc
+//
+// Common header file for both scoreboard and scoreboard_MBAFF kernels
+//
+
+#define ONE_MB_WA // Enable WA for 1-MB wide pictures. To disable WA, simply comment out this line.
+
+#define INLINE_REG_OFF 1
+#define INLINE_REG r1
+#define INLINE_REG1 r2
+
+#define DONEFLAG 0x40 // Bit mask of "completed" thread flag
+
+// GRF r1 map
+//
+#define WIDTHINMB_1 INLINE_REG.0 // :uw type. Picture width in MB - 1
+#define HEIGHTINMB_1 INLINE_REG.1 // :uw type. Picture height in MB - 1
+#define TotalMB INLINE_REG.2 // :uw type. Total number of macroblocks
+#define WFLen_B INLINE_REG.3 // :uw type. Bottom MB Wavefront length (Reserved for MBAFF scoreboard)
+#define WFLen INLINE_REG.4 // :uw type. Wavefront length (used as loop counter)
+#define WFLenY INLINE_REG.5 // :uw type. Wavefront length (vertical component)
+#define StartX INLINE_REG.6 // :uw type. Start X of current wavefront
+#define StartY INLINE_REG.7 // :uw type. Start Y of current wavefront
+#define StartXD INLINE_REG.3 // :ud type. Start (X,Y) of current wavefront
+#define CASE00PTR INLINE_REG.4 // :ud type. Pointer to "inter start" handler
+#define WFLen_Save INLINE_REG.10 // :uw type. Saved Wavefront length (Reserved for MBAFF scoreboard)
+#define CASE10PTR INLINE_REG.6 // :ud type. Pointer to "intra start" handler
+#define CASE11PTR INLINE_REG.7 // :ud type. Pointer to "inter complete" handler
+
+// GRF r2 map
+//
+.declare WFStart Base=GRF(2) ElementSize=2 SrcRegion=REGION(4,1) Type=w // Start MB of recent 4 wavefronts, actually use 5 WORDs
+.declare WFStart_T Base=GRF(2) ElementSize=2 SrcRegion=REGION(4,1) Type=w // Start MB of recent 4 wavefronts
+.declare WFStart_B Base=GRF(2).4 ElementSize=2 SrcRegion=REGION(4,1) Type=w // Start MB of recent 4 wavefronts
+
+#define NewWFOffsetD INLINE_REG1.5 // :d type. Offsets used for new wavefront = 0x01ffff00 (0, -1, -1, 1)
+#define NewWFOffset INLINE_REG1.20 // :b type. Offsets used for new wavefront = 0x01ffff00 (0, -1, -1, 1)
+
+#define AVAILFLAGD INLINE_REG1.6 // :ud type. Neighbor available flags = 0x08020401 (in ACBD order)
+#define AVAILFLAG INLINE_REG1.24 // :ub type. Neighbor available flags as above
+#define AVAILFLAG1D INLINE_REG1.7 // :ud type. Top-half neighbor available flags = 0x80402010 (in A_Bxxx order)
+
+.declare MBINDEX Base=GRF(3) ElementSize=2 SrcRegion=REGION(16,1) Type=w // MB order # of current MB group (Cur, ACBD and AC_B_D_)
+#define AR_SAVE r3.8 // :uw type. Saved Address Register information
+
+#define CMDPTR a0.0 // :uw type. DWORD Pointer to the scoreboard
+#define DEPPTR a0.0 // :uw type. Pointer to the dependency scoreboard - Current MB
+#define DEPPTRL a0.1 // :uw type. Pointer to the dependency scoreboard - Left MB
+#define DEPPTRTR a0.2 // :uw type. Pointer to the dependency scoreboard - Top right MB
+#define DEPPTRT a0.3 // :uw type. Pointer to the dependency scoreboard - Top MB
+#define DEPPTRTL a0.4 // :uw type. Pointer to the dependency scoreboard - Top left MB
+#define DEPPTRLB a0.5 // :uw type. Pointer to the dependency scoreboard - Left bottom-half MB
+
+#define PMSGSEL a0.7 // :uw type. Pointer to current message in message handling table
+
+#define CMD_SB_REG_OFF 4
+.declare CMD_SB Base=GRF(4) ElementSize=4 SrcRegion=REGION(8,1) Type=ud // Command scoreboard (64 GRF)
+
+#ifdef AS_ENABLED
+// Definitions for Advanced Scheduler support
+#define AS_INT BIT23 // "Preemption Exception Status" bit in cr0.1:ud control register
+#define AS_INT_EN BIT10 // "Preemption Exception Enable" bit in cr0.1:ud control register
+#define TH_INT BIT2 // "Thread Interrupted" bit in message descriptor
+#define TH_RES BIT0 // "Thread Restart Enable" bit in R0 header r0.2
+
+#define AS_SAVE 34 // Surface state for saving scoreboard contents
+ // Ensure not to conflict with existing binding table entries
+#endif // End AS_ENABLED
+
+// End of scoreboard_header
+
+#endif // !defined(__SCOREBOARD_HEADER__)
+
diff --git a/shaders/h264/mc/SetHWScoreboard.asm b/shaders/h264/mc/SetHWScoreboard.asm new file mode 100644 index 0000000..c2da855 --- /dev/null +++ b/shaders/h264/mc/SetHWScoreboard.asm @@ -0,0 +1,209 @@ +/*
+ * Set dependency control HW scoreboard kernel
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: SetHWScoreboard.asm
+//
+// Set dependency control HW scoreboard kernel
+//
+
+// ----------------------------------------------------
+// Main: SetHWScoreboard
+// ----------------------------------------------------
+
+.kernel SetHWScoreboard
+
+SETHWSCOREBOARD:
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+mov (1) acc0:ud 0xf0aa55a5:ud
+#endif
+
+#include "header.inc"
+#include "SetHWScoreboard_header.inc"
+
+//
+// Now, begin source code....
+//
+
+.code
+
+// Separate the TotalMB so TotalMB will be multiple of 8
+// and RemainderMB will hold the TotalMB%8
+//
+ and.z.f0.1 (1) RemainderMB<1>:uw TotalMB<0;1,0>:uw 0x0007:uw // number of %8 commands
+ and.z.f0.0 (1) TotalMB<1>:uw TotalMB<0;1,0>:uw 0xfff8:uw // Number of 8-command blocks
+
+ mov (1) MB_SHIFT_MASK_W<1>:uw 0x100*16+12:w // Set up shift values (12, 16)
+
+// Initialize common DAP read header
+//
+ mov (8) MRF_READ_HEADER_SRC<1>:ud r0.0<8;8,1>:ud
+ shl (1) MRF_READ_HEADER_SRC.2<1>:ud StartingMB<0;1,0>:uw 6:uw // Byte-aligned offset being read
+
+// Initialize Inter DAP write header
+ mov (8) MRF_INTER_WRITE_HEADER<1>:ud r0.0<8;8,1>:ud
+
+ (f0.0) jmpi (1) SetHWScoreboard_Remainder // Jump if TotalMB < 8
+
+//------------------------------------------------------------------------
+// Command buffer parsing loop
+// Each loop will handle 8 commands
+//------------------------------------------------------------------------
+//
+SetHWScoreboard_Loop:
+// Load block 0 (Commands 0/1)
+ mov (8) MRF_READ_HEADER0.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ send (16) CMD_BUFFER_W(0)<1> MRF_READ_HEADER0 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 1 (Commands 2/3)
+ mov (8) MRF_READ_HEADER1.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER1.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 128:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(4)<1> MRF_READ_HEADER1 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 2 (Commands 4/5)
+ mov (8) MRF_READ_HEADER2.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER2.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 256:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(8)<1> MRF_READ_HEADER2 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 3 (Commands 6/7)
+ mov (8) MRF_READ_HEADER3.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER3.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 384:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(12)<1> MRF_READ_HEADER3 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Start parsing commands
+ $for(0; <16; 2) {
+ and.nz.f0.1 (8) NULLREG CMD_BUFFER_D(%1,4)<0;1,0> IS_INTRA_MB:ud // Is it an "Intra" MB?
+ or (1) CMD_BUFFER_D(%1,2)<1> CMD_BUFFER_D(%1,2)<0;1,0> BIT21:ud // Set "Use Scoreboard" for every MB
+ shl (2) CMD_BUFFER_W(%1,2)<1> CMD_BUFFER_W(%1,14)<0;1,0> MB_SHIFT_MASK_B<2;2,1>:b // Set HW SB masks
+ mov (2) CMD_BUFFER_B(%1,4)<2> CMD_BUFFER_B(%1,20)<2;2,1> // Set scoreboard (X,Y) for intra MB
+ (-f0.1) mov (2) CMD_BUFFER_W(%1,2)<1> CMD_BUFFER_B(%1,20)<2;2,1> // Set scoreboard (X,Y) for inter MB
+ (f0.1) jmpi (1) Parse_8_Loop_%1
+
+// Inter Macroblock
+// Output MEDIA_OBJECT command in raster scan order
+ mul (16) acc0<1>:uw CMD_BUFFER_B(%1,21)<0;1,0> PicWidthMB<0;1,0>:uw // MB offset = Y*W
+ add (16) acc0<1>:uw acc0<8;8,1>:uw CMD_BUFFER_B(%1,20)<0;1,0> // MB offset = Y*W+X
+ shl (1) MRF_INTER_WRITE_HEADER.2<1>:ud acc0.2<0;1,0>:uw 6:uw // Byte-aligned MB offset
+ mov (16) MRF_INTER_WRITE_DATA0<1>:ud CMD_BUFFER_D(%1)<8;8,1> {Compr} // Copy entire command to inter buffer
+ mov (16) CMD_BUFFER_D(%1)<1> 0:ud {Compr} // Clear original command
+ send (16) NULLREGW MRF_INTER_WRITE_HEADER null:uw DAPWRITE MSG_LEN(2)+OWBWMSGDSC+OWORD_4+BI_CMD_BUFFER
+
+Parse_8_Loop_%1:
+ }
+
+ add.z.f0.0 (1) TotalMB<1>:w TotalMB<0;1,0>:w -8:w // Update remaining number of 8-command blocks
+
+// Output modified intra commands
+// Write block 0
+ mov (8) MRF_INTRA_WRITE_HEADER.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ $for(0; <4; 2) {
+ mov (16) MRF_CMD_BUF_D(%1)<1> CMD_BUFFER_D(%1)<8;8,1> {Compr}
+ }
+ send (16) NULLREGW MRF_INTRA_WRITE_HEADER null:uw DAPWRITE MSG_LEN(4)+OWBWMSGDSC+OWORD_8+BI_CMD_BUFFER
+
+// Write block 1
+ mov (8) m1.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) m1.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 128:ud // Point to next 2-command block
+ mov (16) m2<1>:ud CMD_BUFFER_D(4)<8;8,1> {Compr}
+ mov (16) m4<1>:ud CMD_BUFFER_D(6)<8;8,1> {Compr}
+ send (16) NULLREGW m1 null:uw DAPWRITE MSG_LEN(4)+OWBWMSGDSC+OWORD_8+BI_CMD_BUFFER
+
+// Write block 2
+ add (1) MRF_INTRA_WRITE_HEADER.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 256:ud // Point to next 2-command block
+ $for(0; <4; 2) {
+ mov (16) MRF_CMD_BUF_D(%1)<1> CMD_BUFFER_D(%1+8)<8;8,1> {Compr}
+ }
+ send (16) NULLREGW MRF_INTRA_WRITE_HEADER null:uw DAPWRITE MSG_LEN(4)+OWBWMSGDSC+OWORD_8+BI_CMD_BUFFER
+
+// Write block 3
+ add (1) m1.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 384:ud // Point to next 2-command block
+ mov (16) m2<1>:ud CMD_BUFFER_D(12)<8;8,1> {Compr}
+ mov (16) m4<1>:ud CMD_BUFFER_D(14)<8;8,1> {Compr}
+ send (16) NULLREGW m1 null:uw DAPWRITE MSG_LEN(4)+OWBWMSGDSC+OWORD_8+BI_CMD_BUFFER
+
+// Update message header for next DAP read
+ add (1) MRF_READ_HEADER_SRC.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 512:ud // Point to next block of 8-commands
+
+ cmp.z.f0.1 (1) NULLREG RemainderMB<0;1,0>:w 0:uw // Check if remainder MB = 0
+ (-f0.0) jmpi (1) SetHWScoreboard_Loop // Continue if more command blocks remain
+
+SetHWScoreboard_Remainder:
+// f0.1 should have been set to indicate if RemainderMB = 0
+//
+ (f0.1) jmpi (1) SetHWScoreboard_Done // Stop if all commands have been updated
+
+// Blindly load next 8 commands anyway
+//
+// Load block 0 (Commands 0/1)
+ mov (8) MRF_READ_HEADER0.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ send (16) CMD_BUFFER_W(0)<1> MRF_READ_HEADER0 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 1 (Commands 2/3)
+ mov (8) MRF_READ_HEADER1.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER1.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 128:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(4)<1> MRF_READ_HEADER1 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 2 (Commands 4/5)
+ mov (8) MRF_READ_HEADER2.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER2.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 256:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(8)<1> MRF_READ_HEADER2 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 3 (Commands 6/7)
+ mov (8) MRF_READ_HEADER3.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER3.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 384:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(12)<1> MRF_READ_HEADER3 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Initialize necessary pointers
+ mov (1) a0.1<1>:ud ((CMD_BUFFER_REG_OFF+1)*0x10000+CMD_BUFFER_REG_OFF)*32 // a0.2:w points to command buffer (first half)
+ // a0.3:w points to command buffer (second half)
+// Initialize Inter DAP write header
+ mov (8) MRF_INTER_WRITE_HEADER<1>:ud r0.0<8;8,1>:ud
+
+SetHWScoreboard_Remainder_Loop:
+ and.nz.f0.1 (8) NULLREG r[a0.2,4*4]<0;1,0>:ud IS_INTRA_MB:ud // Is it an "Intra" MB?
+ add.z.f0.0 (1) RemainderMB<1>:w RemainderMB<0;1,0>:w -1:w // Decrement MB #
+ or (1) r[a0.2,2*4]<1>:ud r[a0.2,2*4]<0;1,0>:ud BIT21:ud // Set "Use Scoreboard" for every MB
+ shl (2) r[a0.2,2*2]<1>:uw r[a0.2,14*2]<0;1,0>:uw MB_SHIFT_MASK_B<2;2,1>:b // Set HW SB masks
+ mov (2) r[a0.2,4*1]<2>:ub r[a0.2,5*4]<2;2,1>:ub // Set scoreboard (X,Y) for intra MB
+
+ (-f0.1) mov (2) r[a0.2,4*1]<1>:uw r[a0.2,5*4]<2;2,1>:ub // Set scoreboard (X,Y) for inter MB
+ (f0.1) jmpi (1) Output_Remainder_Intra
+
+// Inter Macroblock
+// Output MEDIA_OBJECT command in raster scan order
+ mul (16) acc0<1>:uw r[a0.2,21]<0;1,0>:ub PicWidthMB<0;1,0>:uw // MB offset = Y*W
+ add (16) acc0<1>:uw acc0<8;8,1>:uw r[a0.2,20]<0;1,0>:ub // MB offset = Y*W+X
+ shl (1) MRF_INTER_WRITE_HEADER.2<1>:ud acc0.2<0;1,0>:uw 6:uw // Byte-aligned MB offset
+ mov (16) MRF_INTER_WRITE_DATA0<1>:ud r[a0.2]<8;8,1>:ud {Compr} // Copy entire command to inter buffer
+ mov (16) r[a0.2]<1>:ud 0:ud {Compr} // Clear original command
+ send (16) NULLREGW MRF_INTER_WRITE_HEADER null:uw DAPWRITE MSG_LEN(2)+OWBWMSGDSC+OWORD_4+BI_CMD_BUFFER
+
+Output_Remainder_Intra:
+// Intra MB command always output
+ mov (8) MRF_INTRA_WRITE_HEADER.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ mov (16) MRF_CMD_BUF_D(0)<1> r[a0.2]<8;8,1>:ud {Compr} // Copy entire command to intra buffer
+ send (16) NULLREGW MRF_INTRA_WRITE_HEADER null:uw DAPWRITE MSG_LEN(2)+OWBWMSGDSC+OWORD_4+BI_CMD_BUFFER
+
+ add (1) MRF_READ_HEADER_SRC.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 64:ud // Point to next command
+ add (1) a0.1<1>:ud a0.1<0;1,0>:ud 0x00400040:ud // Update pointers
+ (-f0.0) jmpi (1) SetHWScoreboard_Remainder_Loop
+
+// All MBs have been decoded. Terminate the thread now
+//
+SetHWScoreboard_Done:
+ END_THREAD
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
+
+// End of SetHWScoreboard
diff --git a/shaders/h264/mc/SetHWScoreboard_MBAFF.asm b/shaders/h264/mc/SetHWScoreboard_MBAFF.asm new file mode 100644 index 0000000..c5cfeb3 --- /dev/null +++ b/shaders/h264/mc/SetHWScoreboard_MBAFF.asm @@ -0,0 +1,279 @@ +/*
+ * Set dependency control HW scoreboard kernel for MBAFF picture
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: SetHWScoreboard_MBAFF.asm
+//
+// Set dependency control HW scoreboard kernel for MBAFF picture
+//
+
+// ----------------------------------------------------
+// Main: SetHWScoreboard_MBAFF
+// ----------------------------------------------------
+
+.kernel SetHWScoreboard_MBAFF
+
+SETHWSCOREBOARD_MBAFF:
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+mov (1) acc0:ud 0xf1aa55a5:ud
+#endif
+
+#include "header.inc"
+#include "SetHWScoreboard_header.inc"
+
+//
+// Now, begin source code....
+//
+
+.code
+
+// Separate the TotalMB so TotalMB will be multiple of 8
+// and RemainderMB will hold the TotalMB%8
+//
+ and.z.f0.1 (1) RemainderMB<1>:uw TotalMB<0;1,0>:uw 0x0007:uw // number of %8 commands
+ and.z.f0.0 (1) TotalMB<1>:uw TotalMB<0;1,0>:uw 0xfff8:uw // Number of 8-command blocks
+
+// Initialize common DAP read header
+//
+ mov (8) MRF_READ_HEADER_SRC<1>:ud r0.0<8;8,1>:ud
+ shl (1) MRF_READ_HEADER_SRC.2<1>:ud StartingMB<0;1,0>:uw 6:uw // Byte-aligned offset being read
+
+// Initialize Inter DAP write header
+ mov (8) MRF_INTER_WRITE_HEADER<1>:ud r0.0<8;8,1>:ud
+
+ (f0.0) jmpi (1) SetHWScoreboard_MBAFF_Remainder // Jump if TatalMB < 8
+
+//------------------------------------------------------------------------
+// Command buffer parsing loop
+// Each loop will handle 8 commands
+//------------------------------------------------------------------------
+//
+SetHWScoreboard_MBAFF_Loop:
+// Load block 0 (Commands 0/1)
+ mov (8) MRF_READ_HEADER0.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ send (16) CMD_BUFFER_W(0)<1> MRF_READ_HEADER0 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 1 (Commands 2/3)
+ mov (8) MRF_READ_HEADER1.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER1.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 128:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(4)<1> MRF_READ_HEADER1 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 2 (Commands 4/5)
+ mov (8) MRF_READ_HEADER2.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER2.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 256:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(8)<1> MRF_READ_HEADER2 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 3 (Commands 6/7)
+ mov (8) MRF_READ_HEADER3.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER3.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 384:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(12)<1> MRF_READ_HEADER3 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Start parsing commands
+ $for(0; <16; 2) {
+// Adjust MB Y origin for field MBs
+//
+ mov (2) TEMP_FD_X_W<1>:uw CMD_BUFFER_B(%1,20)<2;2,1> // Initialize temp (X,Y) location
+ and.nz.f0.1 (8) NULLREG CMD_BUFFER_D(%1,4)<0;1,0> IS_BOT_FD:ud // Is it a "Bottom Field MB"?
+ and.nz.f0.0 (8) NULLREG CMD_BUFFER_D(%1,4)<0;1,0> IS_FIELD_MB:ud // Is it a "Field MB"?
+ mul (8) acc0<1>:w CMD_BUFFER_B(%1,21)<0;1,0> 2:w
+ (-f0.1) mov (1) TEMP_FD_Y_W<1>:w acc0<0;1,0>:w
+ (f0.1) add (1) TEMP_FD_Y_W<1>:w acc0<0;1,0>:w 1:w
+ (-f0.0) mov (1) TEMP_FD_Y_W<1>:w CMD_BUFFER_B(%1,21)<0;1,0> // Discard field MB Y origin handling
+
+ and.nz.f0.0 (8) NULLREG CMD_BUFFER_D(%1,4)<0;1,0> IS_INTRA_MB:ud // Is it an "Intra" MB?
+ and.nz.f0.1 (8) NULLREG TEMP_FD_Y_W<0;1,0>:uw BIT0 // Is it "Bottom MB"?
+ or (1) CMD_BUFFER_D(%1,2)<1> CMD_BUFFER_D(%1,2)<0;1,0> BIT21 // Set "Use Scoreboard"
+ mov (2) CMD_BUFFER_W(%1,2)<1> TEMP_FD_X_W<2;2,1>:uw // Set scoreboard (X,Y) for inter MB
+ (f0.0) jmpi (1) SET_SB_MBAFF_INTRA_%1 // Jump if intra MB.
+
+// Inter Macroblock
+// Output MEDIA_OBJECT command in raster scan order
+ mul (16) acc0<1>:uw TEMP_FD_Y_W<0;1,0>:uw PicWidthMB<0;1,0>:uw // MB offset = Y*W
+ add (16) acc0<1>:uw acc0<8;8,1>:uw TEMP_FD_X_W<0;1,0>:uw // MB offset = Y*W+X
+ shl (1) MRF_INTER_WRITE_HEADER.2<1>:ud acc0.2<0;1,0>:uw 6:uw // Byte-aligned MB offset
+ mov (16) MRF_INTER_WRITE_DATA0<1>:ud CMD_BUFFER_D(%1)<8;8,1> {Compr} // Copy entire command to inter buffer
+ mov (16) CMD_BUFFER_D(%1)<1> 0:ud {Compr} // Clear original command
+ send (16) NULLREGW MRF_INTER_WRITE_HEADER null:uw DAPWRITE MSG_LEN(2)+OWBWMSGDSC+OWORD_4+BI_CMD_BUFFER
+ jmpi (1) NEXT_MB_MBAFF_%1 // Done for inter MB. Move to next MB.
+
+SET_SB_MBAFF_INTRA_%1:
+// Intra MB
+//
+ and.nz.f0.0 (8) NULLREG CMD_BUFFER_D(%1,4)<0;1,0> IS_FIELD_MB:ud // Is it an "Field" MB?
+ (f0.1) sel (2) MB_MASK_D<1>:ud BOT_FD_MASK1_D<2;2,1>:ud TOP_FD_MASK1_D<2;2,1>:ud // Assume field MB
+ mov (1) TEMP_INTRA_FLAG_W<1>:uw CMD_BUFFER_W(%1,14)<0;1,0> // Don't want to alter original in-line data
+ (f0.0) jmpi (1) SET_SB_MBAFF_%1 // Jump if it's really field MB
+
+// Frame MB
+//
+// Derive E'
+ and.nz.f0.0 (8) NULLREG CMD_BUFFER_W(%1,14)<0;1,0> E_FLAG // Is "E" = 1
+ (f0.1) sel (2) MB_MASK_D<1>:ud BOT_FM_MASK1_D<2;2,1>:ud TOP_FM_MASK1_D<2;2,1>:ud
+ and.z.f0.1 (8) NULLREG CMD_BUFFER_W(%1,14)<0;1,0> A_FLAG // "A" = 0?
+ (f0.0) jmpi (1) SET_SB_MBAFF_%1 // If "E" flag = 1, skip the rest of derivation
+ (f0.1) and.nz.f0.1 (8) NULLREG CMD_BUFFER_D(%1,4)<0;1,0> IS_INTRA8X8
+ (f0.1) and.nz.f0.1 (8) NULLREG CMD_BUFFER_W(%1,14)<0;1,0> F_FLAG
+ (f0.1) or (1) TEMP_INTRA_FLAG_W<1>:uw CMD_BUFFER_W(%1,14)<0;1,0> E_FLAG
+
+SET_SB_MBAFF_%1:
+ and.nz.f0.1 (16) NULLREGW TEMP_INTRA_FLAG_W<0;1,0>:uw MB_MASK_B<0;8,1>:ub
+ shl (1) CMD_BUFFER_W(%1,2)<1> f0.1<0;1,0>:uw 12:w // Masks 0-3
+ and (1) CMD_BUFFER_W(%1,3)<1> f0.1<0;1,0>:uw 0xf000:uw // Masks 4-7
+
+ mov (2) CMD_BUFFER_B(%1,4)<2> TEMP_FD_X_B<4;2,2>:ub // Set scoreboard (X,Y) for intra MB
+
+NEXT_MB_MBAFF_%1:
+ }
+
+ add.z.f0.0 (1) TotalMB<1>:w TotalMB<0;1,0>:w -8:w // Update remaining number of 8-command blocks
+
+// Output modified intra commands
+// Write block 0
+ mov (8) MRF_INTRA_WRITE_HEADER.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ $for(0; <4; 2) {
+ mov (16) MRF_CMD_BUF_D(%1)<1> CMD_BUFFER_D(%1)<8;8,1> {Compr}
+ }
+ send (16) NULLREGW MRF_INTRA_WRITE_HEADER null:uw DAPWRITE MSG_LEN(4)+OWBWMSGDSC+OWORD_8+BI_CMD_BUFFER
+
+// Write block 1
+ mov (8) m1.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) m1.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 128:ud // Point to next 2-command block
+ mov (16) m2<1>:ud CMD_BUFFER_D(4)<8;8,1> {Compr}
+ mov (16) m4<1>:ud CMD_BUFFER_D(6)<8;8,1> {Compr}
+ send (16) NULLREGW m1 null:uw DAPWRITE MSG_LEN(4)+OWBWMSGDSC+OWORD_8+BI_CMD_BUFFER
+
+// Write block 2
+ add (1) MRF_INTRA_WRITE_HEADER.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 256:ud // Point to next 2-command block
+ $for(0; <4; 2) {
+ mov (16) MRF_CMD_BUF_D(%1)<1> CMD_BUFFER_D(%1+8)<8;8,1> {Compr}
+ }
+ send (16) NULLREGW MRF_INTRA_WRITE_HEADER null:uw DAPWRITE MSG_LEN(4)+OWBWMSGDSC+OWORD_8+BI_CMD_BUFFER
+
+// Write block 3
+ add (1) m1.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 384:ud // Point to next 2-command block
+ mov (16) m2<1>:ud CMD_BUFFER_D(12)<8;8,1> {Compr}
+ mov (16) m4<1>:ud CMD_BUFFER_D(14)<8;8,1> {Compr}
+ send (16) NULLREGW m1 null:uw DAPWRITE MSG_LEN(4)+OWBWMSGDSC+OWORD_8+BI_CMD_BUFFER
+
+// Update message header for next DAP read
+ add (1) MRF_READ_HEADER_SRC.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 512:ud // Point to next block of 8-commands
+
+ cmp.z.f0.1 (1) NULLREG RemainderMB<0;1,0>:w 0:uw // Check if remaining MB = 0
+ (-f0.0) jmpi (1) SetHWScoreboard_MBAFF_Loop // Continue if more command blocks remain
+
+SetHWScoreboard_MBAFF_Remainder:
+// f0.1 should have been set to indicate if RemainderMB = 0
+//
+ (f0.1) jmpi (1) SetHWScoreboard_MBAFF_Done // Stop if all commands have been updated
+
+// Blindly load next 8 commands anyway
+//
+// Load block 0 (Commands 0/1)
+ mov (8) MRF_READ_HEADER0.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ send (16) CMD_BUFFER_W(0)<1> MRF_READ_HEADER0 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 1 (Commands 2/3)
+ mov (8) MRF_READ_HEADER1.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER1.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 128:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(4)<1> MRF_READ_HEADER1 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 2 (Commands 4/5)
+ mov (8) MRF_READ_HEADER2.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER2.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 256:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(8)<1> MRF_READ_HEADER2 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Load block 3 (Commands 6/7)
+ mov (8) MRF_READ_HEADER3.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ add (1) MRF_READ_HEADER3.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 384:ud // Point to next 2-command block
+ send (16) CMD_BUFFER_W(12)<1> MRF_READ_HEADER3 null:uw DAPREAD RESP_LEN(4)+OWBRMSGDSC_SC+OWORD_8+BI_CMD_BUFFER
+
+// Initialize necessary pointers
+ mov (1) a0.1<1>:ud ((CMD_BUFFER_REG_OFF+1)*0x10000+CMD_BUFFER_REG_OFF)*32 // a0.2:w points to command buffer (first half)
+ // a0.3:w points to command buffer (second half)
+// Initialize Inter DAP write header
+ mov (8) MRF_INTER_WRITE_HEADER<1>:ud r0.0<8;8,1>:ud
+
+SetHWScoreboard_MBAFF_Remainder_Loop:
+// Adjust MB Y origin for field MBs
+//
+ mov (2) TEMP_FD_X_W<1>:uw r[a0.2,5*4]<2;2,1>:ub // Initialize temp (X,Y) location
+ and.nz.f0.1 (8) NULLREG r[a0.2,4*4]<0;1,0>:ud IS_BOT_FD:ud // Is it a "Bottom Field MB"?
+ and.nz.f0.0 (8) NULLREG r[a0.2,4*4]<0;1,0>:ud IS_FIELD_MB:ud // Is it a "Field MB"?
+ mul (8) acc0<1>:w r[a0.2,21]<0;1,0>:ub 2:w
+ (-f0.1) mov (1) TEMP_FD_Y_W<1>:w acc0<0;1,0>:w
+ (f0.1) add (1) TEMP_FD_Y_W<1>:w acc0<0;1,0>:w 1:w
+ (-f0.0) mov (1) TEMP_FD_Y_W<1>:w r[a0.2,5*4+1]<0;1,0>:ub // Discard field MB Y origin handling
+
+ and.nz.f0.0 (8) NULLREG r[a0.2,4*4]<0;1,0>:ud IS_INTRA_MB:ud // Is it an "Intra" MB?
+ add.z.f0.1 (1) RemainderMB<1>:w RemainderMB<0;1,0>:w -1:w // Decrement MB #
+ or (1) r[a0.2,2*4]<1>:ud r[a0.2,2*4]<0;1,0>:ud BIT21:ud // Set "Use Scoreboard"
+ mov (2) r[a0.2,2*2]<1>:uw TEMP_FD_X_W<2;2,1>:uw // Set scoreboard (X,Y) for inter MB
+ (f0.0) jmpi (1) SET_SB_MBAFF_REM_INTRA // Jump if intra MB.
+
+// Inter Macroblock
+// Output MEDIA_OBJECT command in raster scan order
+ mul (16) acc0<1>:uw TEMP_FD_Y_W<0;1,0>:uw PicWidthMB<0;1,0>:uw // MB offset = Y*W
+ add (16) acc0<1>:uw acc0<8;8,1>:uw TEMP_FD_X_W<0;1,0>:uw // MB offset = Y*W+X
+ shl (1) MRF_INTER_WRITE_HEADER.2<1>:ud acc0.2<0;1,0>:uw 6:uw // Byte-aligned MB offset
+ mov (16) MRF_INTER_WRITE_DATA0<1>:ud r[a0.2]<8;8,1>:ud {Compr} // Copy entire command to inter buffer
+ mov (16) r[a0.2]<1>:ud 0:ud {Compr} // Clear original command
+ send (16) NULLREGW MRF_INTER_WRITE_HEADER null:uw DAPWRITE MSG_LEN(2)+OWBWMSGDSC+OWORD_4+BI_CMD_BUFFER
+ jmpi (1) Output_MBAFF_Remainder_Intra // Done for inter MB. Move to dump intra MB.
+
+SET_SB_MBAFF_REM_INTRA:
+// Intra MB
+//
+ and.nz.f0.1 (8) NULLREG TEMP_FD_Y_W<0;1,0>:uw BIT0:ud // Is it "Bottom MB"?
+ and.nz.f0.0 (8) NULLREG r[a0.2,4*4]<0;1,0>:ud IS_FIELD_MB:ud // Is it "Field MB"?
+ mov (1) TEMP_INTRA_FLAG_W<1>:uw r[a0.2,14*2]<0;1,0>:uw // Don't want to alter original in-line data
+ (f0.1) sel (2) MB_MASK_D<1>:ud BOT_FD_MASK1_D<2;2,1>:ud TOP_FD_MASK1_D<2;2,1>:ud // Assume field MB
+ (f0.0) jmpi (1) SET_SB_MBAFF_REM // Jump if it's really field MB
+
+// Frame MB
+//
+// Derive E'
+ and.nz.f0.0 (8) NULLREG r[a0.2,14*2]<0;1,0>:uw E_FLAG // Is "E" = 1
+ (f0.1) sel (2) MB_MASK_D<1>:ud BOT_FM_MASK1_D<2;2,1>:ud TOP_FM_MASK1_D<2;2,1>:ud
+ and.z.f0.1 (8) NULLREG r[a0.2,14*2]<0;1,0>:uw A_FLAG // "A" = 0?
+ (f0.0) jmpi (1) SET_SB_MBAFF_REM // If "E" flag = 1, skip the rest of derivation
+ (f0.1) and.nz.f0.1 (8) NULLREG r[a0.2,4*4]<0;1,0>:ud IS_INTRA8X8
+ (f0.1) and.nz.f0.1 (8) NULLREG r[a0.2,14*2]<0;1,0>:uw F_FLAG
+ (f0.1) or (1) TEMP_INTRA_FLAG_W<1>:uw r[a0.2,14*2]<0;1,0>:uw E_FLAG
+
+SET_SB_MBAFF_REM:
+ and.nz.f0.0 (16) NULLREGW TEMP_INTRA_FLAG_W<0;1,0>:uw MB_MASK_B<0;8,1>:ub
+ add.z.f0.1 (1) RemainderMB<1>:w RemainderMB<0;1,0>:w 0:w // Check remaining MB #
+ shl (1) r[a0.2,2*2]<1>:uw f0.0<0;1,0>:uw 12:w // Masks 0-3
+ and (1) r[a0.2,3*2]<1>:uw f0.0<0;1,0>:uw 0xf000:uw // Masks 4-7
+
+ mov (2) r[a0.2,4*1]<2>:ub TEMP_FD_X_B<4;2,2>:ub // Set scoreboard (X,Y) for intra MB
+
+Output_MBAFF_Remainder_Intra:
+// Intra MB command always output
+ mov (8) MRF_INTRA_WRITE_HEADER.0<1>:ud MRF_READ_HEADER_SRC.0<8;8,1>:ud
+ mov (16) MRF_CMD_BUF_D(0)<1> r[a0.2]<8;8,1>:ud {Compr} // Copy entire command to intra buffer
+ send (16) NULLREGW MRF_INTRA_WRITE_HEADER null:uw DAPWRITE MSG_LEN(2)+OWBWMSGDSC+OWORD_4+BI_CMD_BUFFER
+
+ add (1) MRF_READ_HEADER_SRC.2<1>:ud MRF_READ_HEADER_SRC.2<0;1,0>:ud 64:ud // Point to next command
+ add (1) a0.1<1>:ud a0.1<0;1,0>:ud 0x00400040:ud // Update pointers
+ (-f0.1) jmpi (1) SetHWScoreboard_MBAFF_Remainder_Loop
+
+// All MBs have been decoded. Terminate the thread now
+//
+SetHWScoreboard_MBAFF_Done:
+ END_THREAD
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
+
+// End of SetHWScoreboard_MBAFF
diff --git a/shaders/h264/mc/SetHWScoreboard_header.inc b/shaders/h264/mc/SetHWScoreboard_header.inc new file mode 100644 index 0000000..1df91f9 --- /dev/null +++ b/shaders/h264/mc/SetHWScoreboard_header.inc @@ -0,0 +1,134 @@ +/*
+ * Common header file for both SetHWScoreboard and SetHWScoreboard_MBAFF kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__SETHWSCOREBOARD_HEADER__) // Make sure this file is only included once
+#define __SETHWSCOREBOARD_HEADER__
+
+// Module name: SetHWScoreboard_header.inc
+//
+// Common header file for both SetHWScoreboard and SetHWScoreboard_MBAFF kernels
+//
+
+#define BI_CMD_BUF 0 // Binding table index for command buffer
+
+// GRF r1 map
+//
+// For use by setting HW scoreboard kernel for MBAFF picture
+//
+// CURBE data
+#define TOP_FM_MASK1_D r1.0 // Bit mask for first half of top frame MB SB mask
+#define TOP_FM_MASK1_B r1.0 // Bit mask for first half of top frame MB SB mask
+#define TOP_FM_MASK2_D r1.1 // Bit mask for second half of top frame MB SB mask
+#define TOP_FM_MASK2_B r1.4 // Bit mask for second half of top frame MB SB mask
+#define BOT_FM_MASK1_D r1.2 // Bit mask for first half of bottom frame MB SB mask
+#define BOT_FM_MASK1_B r1.8 // Bit mask for first half of bottom frame MB SB mask
+#define BOT_FM_MASK2_D r1.3 // Bit mask for second half of bottom frame MB SB mask
+#define BOT_FM_MASK2_B r1.12 // Bit mask for second half of bottom frame MB SB mask
+#define TOP_FD_MASK1_D r1.4 // Bit mask for first half of top field MB SB mask
+#define TOP_FD_MASK1_B r1.16 // Bit mask for first half of top field MB SB mask
+#define TOP_FD_MASK2_D r1.5 // Bit mask for second half of top field MB SB mask
+#define TOP_FD_MASK2_B r1.20 // Bit mask for second half of top field MB SB mask
+#define BOT_FD_MASK1_D r1.6 // Bit mask for first half of bottom field MB SB mask
+#define BOT_FD_MASK1_B r1.24 // Bit mask for first half of bottom field MB SB mask
+#define BOT_FD_MASK2_D r1.7 // Bit mask for second half of bottom field MB SB mask
+#define BOT_FD_MASK2_B r1.28 // Bit mask for second half of bottom field MB SB mask
+
+// For use by setting HW scoreboard kernel for non-MBAFF picture
+#define MB_SHIFT_MASK_W r1.0 // :w type. Shift values for two parts of the MB SB mask
+#define MB_SHIFT_MASK_B r1.0 // :b type. Shift values for two parts of the MB SB mask
+
+// GRF r2 map
+//
+// In-line data
+//
+#define INLINE_REG_OFFSET 1
+#define INLINE_REG r2
+
+#define StartingMB INLINE_REG.0 // :uw type. Starting MB number
+#define TotalMB INLINE_REG.1 // :uw type. Total number of MB to be processed
+#define PicWidthMB INLINE_REG.2 // :uw type. Picture width in MB
+
+// GRF r3 map
+//
+// Temporary variables
+//
+#define RemainderMB r3.0 // :uw type. Remainder of MB (<16) to be processed
+
+#define TEMP_FD_X_W r3.2 // :w type. Temporary variable for field MB X origin in MBAFF picture
+#define TEMP_FD_X_B r3.4 // :b type. Temporary variable for field MB X origin in MBAFF picture
+#define TEMP_FD_Y_W r3.3 // :w type. Temporary variable for field MB Y origin in MBAFF picture
+#define TEMP_FD_Y_B r3.6 // :b type. Temporary variable for field MB Y origin in MBAFF picture
+
+#define TEMP_INTRA_FLAG_W r3.4 // :uw type. Temporary intra available flag
+
+#define MB_MASK_D r3.4 // :ud type. Bit masks for MBAFF MB
+#define MB_MASK_B r3.16 // :ub type. Bit masks for MBAFF MB
+
+#define MRF_READ_HEADER_SRC r63
+
+// MEDIA_OBJECT_EX Command map
+//
+// In DW1 of each MEDIA_OBJECT_EX command (VFE DWORD)
+#define CUR_X 0 // Byte 0
+#define CUR_Y 0 // Byte 2
+
+// In DW2 of each MEDIA_OBJECT_EX command
+#define USE_SCOREBOARD BIT21
+
+// In DW4 of each MEDIA_OBJECT_EX command
+#define F_FLAG BIT4
+#define IS_INTRA_MB BIT13
+#define IS_FIELD_MB BIT14
+#define IS_INTRA8X8 BIT15
+#define IS_BOT_FD BIT24
+
+// In DW7 of each MEDIA_OBJECT_EX command
+#define A_FLAG BIT0
+#define B_FLAG BIT1
+#define C_FLAG BIT2
+#define D_FLAG BIT3
+#define E_FLAG BIT4
+
+#define CMD_BUFFER_REG_OFF 4
+.declare CMD_BUFFER_D Base=GRF(4) ElementSize=4 SrcRegion=REGION(8,1) Type=ud // Command buffer (32 GRF)
+.declare CMD_BUFFER_W Base=GRF(4) ElementSize=2 SrcRegion=REGION(16,1) Type=uw // Command buffer (32 GRF)
+.declare CMD_BUFFER_B Base=GRF(4) ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Command buffer (32 GRF)
+
+#define MRF_READ_HEADER m1
+#define MRF_READ_HEADER0 m1
+#define MRF_READ_HEADER1 m2
+#define MRF_READ_HEADER2 m3
+#define MRF_READ_HEADER3 m4
+
+#define MRF_INTER_WRITE_HEADER m5
+#define MRF_INTER_WRITE_DATA0 m6
+#define MRF_INTER_WRITE_DATA1 m7
+
+#define MRF_WRITE_HEADER m11
+#define MRF_INTRA_WRITE_HEADER m11
+
+#define MRF_CMD_BUF_REG_OFF 12
+.declare MRF_CMD_BUF_D Base=m12 ElementSize=4 SrcRegion=REGION(8,1) Type=ud // Command buffer stored in MRF
+.declare MRF_CMD_BUF_W Base=m12 ElementSize=2 SrcRegion=REGION(16,1) Type=uw // Command buffer stored in MRF
+.declare MRF_CMD_BUF_B Base=m12 ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Command buffer stored in MRF
+
+#define BI_CMD_BUFFER 0
+
+#define OWBRMSGDSC_SC 0x02088000 // OWORD Block Read Message Descriptor, reading from sampler cache = A.
+#define OWBWMSGDSC 0x02080000 // OWORD Block Write Message Descriptor
+
+#define OWORD_1 0x000
+#define OWORD_2 0x200
+#define OWORD_4 0x300
+#define OWORD_8 0x400
+
+// End of SETHWSCOREBOARD_HEADER
+
+#endif // !defined(__SETHWSCOREBOARD_HEADER__)
+
diff --git a/shaders/h264/mc/SetupForHWMC.asm b/shaders/h264/mc/SetupForHWMC.asm new file mode 100644 index 0000000..b6dc595 --- /dev/null +++ b/shaders/h264/mc/SetupForHWMC.asm @@ -0,0 +1,33 @@ +/*
+ * Initial setup for running HWMC kernels in HWMC-Only decoding mode
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: SetupForHWMC.asm
+//
+// Initial setup for running HWMC kernels in HWMC-Only decoding mode
+//
+#include "header.inc"
+#include "intra_Header.inc"
+
+#if !defined(__SETUPFORHWMC__) // Make sure the following are only included once
+#define __SETUPFORHWMC__
+
+.reg_count_total 64
+.reg_count_payload 2
+
+//
+// Now, begin source code....
+//
+
+.code
+#endif // !defined(__SETUPFORHWMC__)
+
+ mov (8) MSGSRC<1>:ud r0.0<8;8,1>:ud // Initialize message header payload with R0
+ shl (2) I_ORIX<1>:uw ORIX<2;2,1>:ub 4:w // Convert MB origin to pixel unit
+
+// End of SetupForHWMC
diff --git a/shaders/h264/mc/add_Error_16x16_Y.asm b/shaders/h264/mc/add_Error_16x16_Y.asm new file mode 100644 index 0000000..df01b99 --- /dev/null +++ b/shaders/h264/mc/add_Error_16x16_Y.asm @@ -0,0 +1,51 @@ +/*
+ * Add macroblock correction Y data blocks to predicted picture
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+// Module name: add_Error_16x16_Y.asm
+//
+// Add macroblock correction Y data blocks to predicted picture
+//
+
+// Every line of predicted Y data is added to Y error data if CBP bit is set
+
+ mov (1) PERROR_UD<1>:ud 0x10001*ERRBUF*GRFWIB+0x00100000:ud // Pointers to first and second row of error block
+
+ and.z.f0.1 (1) NULLREG REG_CBPCY CBP_Y_MASK
+ (f0.1) jmpi (1) End_add_Error_16x16_Y // Skip all blocks
+
+// Block Y0
+//
+ $for(0,0; <8; 2,1) {
+ add.sat (16) DEC_Y(%1)<2> r[PERROR,%2*GRFWIB]REGION(8,1):w PRED_Y(%1)REGION(8,2) {Compr}
+ }
+
+// Block Y1
+//
+ $for(0,0; <8; 2,1) {
+ add.sat (16) DEC_Y(%1,16)<2> r[PERROR,%2*GRFWIB+0x80]REGION(8,1):w PRED_Y(%1,16)REGION(8,2) {Compr}
+ }
+
+// Block Y2
+//
+ $for(8,0; <16; 2,1) {
+ add.sat (16) DEC_Y(%1)<2> r[PERROR,%2*GRFWIB+0x100]REGION(8,1):w PRED_Y(%1)REGION(8,2) {Compr}
+ }
+
+// Block Y3
+//
+ $for(8,0; <16; 2,1) {
+ add.sat (16) DEC_Y(%1,16)<2> r[PERROR,%2*GRFWIB+0x180]REGION(8,1):w PRED_Y(%1,16)REGION(8,2) {Compr}
+ }
+
+End_add_Error_16x16_Y:
+ add (1) PERROR_UD<1>:ud PERROR_UD:ud 0x01800180:ud // Pointers to Y3 error block
+
+// End of add_Error_16x16_Y
+
diff --git a/shaders/h264/mc/add_Error_UV.asm b/shaders/h264/mc/add_Error_UV.asm new file mode 100644 index 0000000..e2c0dea --- /dev/null +++ b/shaders/h264/mc/add_Error_UV.asm @@ -0,0 +1,38 @@ +/*
+ * Add macroblock correction UV data blocks to predicted picture
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+
+#if !defined(__ADD_ERROR_UV__) // Make sure this is only included once
+#define __ADD_ERROR_UV__
+
+// Module name: add_Error_UV.asm
+//
+// Add macroblock correction UV data blocks to predicted picture
+
+// PERROR points to error block Y3 after decoding Y component
+
+// Update address register used in instruction compression
+//
+
+// U component
+//
+ add (1) PERROR1<1>:w PERROR:w 0x00010:w // Pointers to next error row
+ $for(0,0; <8; 2,1) {
+ add.sat (16) DEC_UV(%1)<4> r[PERROR,%2*GRFWIB+0x80]REGION(8,1):w PRED_UV(%1)REGION(8,4) {Compr}
+ }
+
+// V component
+//
+ $for(0,0; <8; 2,1) {
+ add.sat (16) DEC_UV(%1,2)<4> r[PERROR,%2*GRFWIB+0x100]REGION(8,1):w PRED_UV(%1,2)REGION(8,4) {Compr}
+ }
+
+// End of add_Error_UV
+
+#endif // !defined(__ADD_ERROR_UV__)
diff --git a/shaders/h264/mc/avc_mc.g4b b/shaders/h264/mc/avc_mc.g4b new file mode 100644 index 0000000..5a91f32 --- /dev/null +++ b/shaders/h264/mc/avc_mc.g4b @@ -0,0 +1,2938 @@ + { 0x00600001, 0x27c00021, 0x008d0000, 0x00000000 }, + { 0x00200009, 0x27fa3e29, 0x00450064, 0x00040004 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x000000d4 }, + { 0x00000005, 0x220e3e2c, 0x00000070, 0x000f000f }, + { 0x00000001, 0x26a00221, 0x00009c38, 0x00000000 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a0 }, + { 0x00a02001, 0x24000229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24400229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24800229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24c00229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x25000229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x25400229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x25800229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x25c00229, 0x00b10624, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000086 }, + { 0x00000001, 0x22080060, 0x00000000, 0x06440640 }, + { 0x00a02001, 0x24000229, 0x00009003, 0x00000000 }, + { 0x00a02001, 0x24400229, 0x0000900b, 0x00000000 }, + { 0x00a02001, 0x24800229, 0x00009013, 0x00000000 }, + { 0x00a02001, 0x24c00229, 0x0000901b, 0x00000000 }, + { 0x00a02001, 0x25000229, 0x00009023, 0x00000000 }, + { 0x00a02001, 0x25400229, 0x0000902b, 0x00000000 }, + { 0x00a02001, 0x25800229, 0x00009033, 0x00000000 }, + { 0x00a02001, 0x25c00229, 0x0000903b, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000072 }, + { 0x02600005, 0x20000c20, 0x0000006c, 0x00000002 }, + { 0x00600005, 0x24000c20, 0x0000006c, 0x00000011 }, + { 0x01600007, 0x20000c00, 0x028d0400, 0x00000011 }, + { 0x00780001, 0x26240169, 0x00000000, 0x80808080 }, + { 0x00780001, 0x66430231, 0x028d0624, 0x00000000 }, + { 0x00780001, 0x66630231, 0x028d062c, 0x00000000 }, + { 0x00780001, 0x26240231, 0x00cf0643, 0x00000000 }, + { 0x00780001, 0x262c0231, 0x00cf0663, 0x00000000 }, + { 0x00800040, 0x25e04629, 0x00cf0643, 0x00b10624 }, + { 0x00600040, 0x25e02529, 0x008d05e0, 0x008d05f0 }, + { 0x00400040, 0x25e02529, 0x006905e0, 0x006905e8 }, + { 0x00200040, 0x25e02529, 0x004505e0, 0x004505e4 }, + { 0x00a02040, 0x2400252c, 0x000005e0, 0x000005e2 }, + { 0x00a02040, 0x24003d8c, 0x00b10400, 0x00100010 }, + { 0x00a02008, 0x24003d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x24403d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x24803d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x24c03d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x25003d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x25403d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x25803d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x25c03d89, 0x00b10400, 0x00050005 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000044 }, + { 0x00600041, 0x26806e2d, 0x008d062c, 0x89abcdef }, + { 0x00600041, 0x26906e2d, 0x008d0623, 0xfedcba98 }, + { 0x00600041, 0x26a06e2d, 0x00cf0663, 0x89abcdef }, + { 0x00600041, 0x26b06e2d, 0x00cf0643, 0x0fedcba9 }, + { 0x00000041, 0x26be3e2d, 0x00000623, 0xfff8fff8 }, + { 0x00802040, 0x268035ad, 0x008d4680, 0x008d0690 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0688 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0684 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0682 }, + { 0x00800001, 0x240001ec, 0x00000000, 0x00200020 }, + { 0x00200048, 0x24003dac, 0x00a00680, 0x00050005 }, + { 0x00200008, 0x26e03d8d, 0x00450400, 0x00060006 }, + { 0x00800001, 0x240001ec, 0x00000000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x00000633, 0x00100010 }, + { 0x00800048, 0x26c03e2d, 0x0000067f, 0x00100010 }, + { 0x00800048, 0x272055ad, 0x000006e0, 0x00b10040 }, + { 0x00600041, 0x268055ad, 0x000006e2, 0x00ae0040 }, + { 0x00600041, 0x26a055ad, 0x000006e2, 0x00ae0041 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00680 }, + { 0x80a02008, 0x44003d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00682 }, + { 0x80a02008, 0x44403d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00684 }, + { 0x80a02008, 0x44803d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00686 }, + { 0x80a02008, 0x44c03d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00688 }, + { 0x80a02008, 0x45003d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b0068a }, + { 0x80a02008, 0x45403d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b0068c }, + { 0x80a02008, 0x45803d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b0068e }, + { 0x80a02008, 0x45c03d91, 0x00b10400, 0x00050005 }, + { 0x00000001, 0x22040060, 0x00000000, 0x00900080 }, + { 0x01000005, 0x20000c20, 0x02000068, 0x00003c00 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000020 }, + { 0x80802040, 0x440045b1, 0x008d8800, 0x00ae0400 }, + { 0x80802040, 0x444045b1, 0x008d8820, 0x00ae0440 }, + { 0x80802040, 0x448045b1, 0x008d8840, 0x00ae0480 }, + { 0x80802040, 0x44c045b1, 0x008d8860, 0x00ae04c0 }, + { 0x80802040, 0x441045b1, 0x008d8880, 0x00ae0410 }, + { 0x80802040, 0x445045b1, 0x008d88a0, 0x00ae0450 }, + { 0x80802040, 0x449045b1, 0x008d88c0, 0x00ae0490 }, + { 0x80802040, 0x44d045b1, 0x008d88e0, 0x00ae04d0 }, + { 0x80802040, 0x450045b1, 0x008d8900, 0x00ae0500 }, + { 0x80802040, 0x454045b1, 0x008d8920, 0x00ae0540 }, + { 0x80802040, 0x458045b1, 0x008d8940, 0x00ae0580 }, + { 0x80802040, 0x45c045b1, 0x008d8960, 0x00ae05c0 }, + { 0x80802040, 0x451045b1, 0x008d8980, 0x00ae0510 }, + { 0x80802040, 0x455045b1, 0x008d89a0, 0x00ae0550 }, + { 0x80802040, 0x459045b1, 0x008d89c0, 0x00ae0590 }, + { 0x80802040, 0x45d045b1, 0x008d89e0, 0x00ae05d0 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x01800180 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x000000e2 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000018 }, + { 0x00000001, 0x27800021, 0x00000780, 0x00000000 }, + { 0x00000001, 0x27a00021, 0x000007a0, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00200440, 0x27c055a5, 0x004507fa, 0x0045002a }, + { 0x00000801, 0x27c80061, 0x00000000, 0x0000001b }, + { 0x00000040, 0x22000d20, 0x00000062, 0x02186000 }, + { 0x01600031, 0x26200021, 0x408d07c0, 0x00000200 }, + { 0x00000440, 0x27c43ca5, 0x000007c4, 0x00010001 }, + { 0x00000801, 0x27c80061, 0x00000000, 0x000f0003 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x00100000 }, + { 0x02600031, 0x26400021, 0x408d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00000008, 0x27fc3dad, 0x000007fc, 0x00010001 }, + { 0x00200440, 0x27c055a5, 0x004507fa, 0x0045002a }, + { 0x00000801, 0x27c80061, 0x00000000, 0x00000013 }, + { 0x00000040, 0x22000c00, 0x00000200, 0xefffc001 }, + { 0x01600031, 0x26200021, 0x408d07c0, 0x00000200 }, + { 0x00000440, 0x27c43ca5, 0x000007c4, 0x00010001 }, + { 0x00000801, 0x27c80061, 0x00000000, 0x00070003 }, + { 0x05600031, 0x26400021, 0x408d07c0, 0x00000200 }, + { 0x00000008, 0x220e3e2c, 0x0000006c, 0x00060006 }, + { 0x00000001, 0x26a002a5, 0x00009c3c, 0x00000000 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a0 }, + { 0x02600005, 0x20000c20, 0x0000006c, 0x00000002 }, + { 0x00780001, 0x26240169, 0x00000000, 0x80808080 }, + { 0x02400005, 0x20000c20, 0x0200006c, 0x00000001 }, + { 0x00560001, 0x46420129, 0x02690624, 0x00000000 }, + { 0x02400005, 0x20000c20, 0x0200006c, 0x00000010 }, + { 0x00560001, 0x46520129, 0x0269062c, 0x00000000 }, + { 0x00780001, 0x26240129, 0x00ae0642, 0x00000000 }, + { 0x00800040, 0x24004629, 0x00b10624, 0x00650642 }, + { 0x00600040, 0x24002529, 0x00650400, 0x00650404 }, + { 0x00600040, 0x25202529, 0x00050400, 0x00050404 }, + { 0x00600040, 0x25702529, 0x00050408, 0x0005040c }, + { 0x00560001, 0x26240169, 0x00000000, 0x80808080 }, + { 0x00460001, 0x26240129, 0x028a0652, 0x00000000 }, + { 0x00560001, 0x46520129, 0x02690624, 0x00000000 }, + { 0x02400005, 0x20000c20, 0x0200006c, 0x00000001 }, + { 0x00560001, 0x46420169, 0x02000000, 0x80808080 }, + { 0x00460001, 0x46420129, 0x0069062c, 0x00000000 }, + { 0x00560001, 0x262c0129, 0x008a0642, 0x00000000 }, + { 0x00600040, 0x24004629, 0x008d0624, 0x00650652 }, + { 0x00600040, 0x24104629, 0x00650642, 0x008d062c }, + { 0x00600040, 0x24002529, 0x00650400, 0x00650404 }, + { 0x00600040, 0x25302529, 0x00050408, 0x0005040c }, + { 0x00600040, 0x25602529, 0x00050400, 0x00050404 }, + { 0x00a02040, 0x24003d2c, 0x00b10520, 0x00040004 }, + { 0x00a02008, 0x24003d89, 0x00b10400, 0x00030003 }, + { 0x00a02008, 0x24403d89, 0x00b10400, 0x00030003 }, + { 0x00a02040, 0x24003d2c, 0x00b10560, 0x00040004 }, + { 0x00a02008, 0x24803d89, 0x00b10400, 0x00030003 }, + { 0x00a02008, 0x24c03d89, 0x00b10400, 0x00030003 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000048 }, + { 0x00000001, 0x22080060, 0x00000000, 0x06440640 }, + { 0x00a02001, 0x24000229, 0x00059002, 0x00000000 }, + { 0x00a02001, 0x24400229, 0x0005900a, 0x00000000 }, + { 0x00a02001, 0x24800229, 0x00059012, 0x00000000 }, + { 0x00a02001, 0x24c00229, 0x0005901a, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000003c }, + { 0x00a02001, 0x24000229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24400229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24800229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24c00229, 0x00b10624, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000032 }, + { 0x00600041, 0x26806e2d, 0x008d062c, 0x44332211 }, + { 0x00600041, 0x26906e2d, 0x008d0622, 0xffeeddcc }, + { 0x00600041, 0x26a06e2d, 0x00650652, 0x44332211 }, + { 0x00600041, 0x26b06e2d, 0x00650642, 0x00ffeedd }, + { 0x00200041, 0x26bc3e2d, 0x00450622, 0xfffcfffc }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0690 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0688 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0684 }, + { 0x00800001, 0x240001ec, 0x00000000, 0x00200020 }, + { 0x00400048, 0x24003dac, 0x00a50680, 0x00220022 }, + { 0x00400008, 0x26e03d8d, 0x00690400, 0x00060006 }, + { 0x00800001, 0x240001ec, 0x00000000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x00050632, 0x00100010 }, + { 0x00800048, 0x26c03e2d, 0x0005065e, 0x00100010 }, + { 0x00800048, 0x272055ad, 0x000506e0, 0x00240044 }, + { 0x00600041, 0x268055ad, 0x000506e4, 0x00440044 }, + { 0x00600041, 0x26a055ad, 0x000506e4, 0x00440045 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00050680 }, + { 0x80a02008, 0x44003d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00050684 }, + { 0x80a02008, 0x44403d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00050688 }, + { 0x80a02008, 0x44803d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x0005068c }, + { 0x80a02008, 0x44c03d91, 0x00b10400, 0x00050005 }, + { 0x00000040, 0x22063d8c, 0x00000204, 0x00100010 }, + { 0x80802040, 0x640045b1, 0x008d8880, 0x00cf0400 }, + { 0x80802040, 0x644045b1, 0x008d88a0, 0x00cf0440 }, + { 0x80802040, 0x648045b1, 0x008d88c0, 0x00cf0480 }, + { 0x80802040, 0x64c045b1, 0x008d88e0, 0x00cf04c0 }, + { 0x80802040, 0x640245b1, 0x008d8900, 0x00cf0402 }, + { 0x80802040, 0x644245b1, 0x008d8920, 0x00cf0442 }, + { 0x80802040, 0x648245b1, 0x008d8940, 0x00cf0482 }, + { 0x80802040, 0x64c245b1, 0x008d8960, 0x00cf04c2 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x0007000f }, + { 0x00200801, 0x27c001a1, 0x004507fa, 0x00000000 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x08004000 }, + { 0x00800001, 0x20400232, 0x00d20400, 0x00000000 }, + { 0x00800001, 0x20500232, 0x00d20420, 0x00000000 }, + { 0x00800001, 0x20600232, 0x00d20440, 0x00000000 }, + { 0x00800001, 0x20700232, 0x00d20460, 0x00000000 }, + { 0x00800001, 0x20800232, 0x00d20480, 0x00000000 }, + { 0x00800001, 0x20900232, 0x00d204a0, 0x00000000 }, + { 0x00800001, 0x20a00232, 0x00d204c0, 0x00000000 }, + { 0x00800001, 0x20b00232, 0x00d204e0, 0x00000000 }, + { 0x01600031, 0x27a00021, 0x508d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00200801, 0x27c001a1, 0x004507fa, 0x00000000 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x0ff04000 }, + { 0x00000001, 0x22080060, 0x00000000, 0x04400400 }, + { 0x00a02001, 0x20400232, 0x00d29000, 0x00000000 }, + { 0x00a02001, 0x20500232, 0x00d29020, 0x00000000 }, + { 0x00a02001, 0x20800232, 0x00d29080, 0x00000000 }, + { 0x00a02001, 0x20900232, 0x00d290a0, 0x00000000 }, + { 0x00a02001, 0x20c00232, 0x00d29100, 0x00000000 }, + { 0x00a02001, 0x20d00232, 0x00d29120, 0x00000000 }, + { 0x00a02001, 0x21000232, 0x00d29180, 0x00000000 }, + { 0x00a02001, 0x21100232, 0x00d291a0, 0x00000000 }, + { 0x01600031, 0x27800021, 0x508d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00600001, 0x27c00021, 0x008d0000, 0x00000000 }, + { 0x00200009, 0x27fa3e29, 0x00450064, 0x00040004 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0xffffff04 }, + { 0x00000001, 0x220401ec, 0x00000000, 0x00800080 }, + { 0x00000001, 0x22080060, 0x00000000, 0x04100400 }, + { 0x00200008, 0x27006e2d, 0x0000006c, 0x00000040 }, + { 0x00400001, 0x27400061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x272001a9, 0x00b10620, 0x00000000 }, + { 0x02600005, 0x20001c20, 0x0000006c, 0x00000008 }, + { 0x00110001, 0x27230231, 0x00000624, 0x00000000 }, + { 0x00600001, 0x27420231, 0x00cf0643, 0x00000000 }, + { 0x00110001, 0x27410231, 0x00000643, 0x00000000 }, + { 0x00240001, 0x27400231, 0x00000623, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x0200006c, 0x00000002 }, + { 0x00010001, 0x27400231, 0x02000624, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x0200006c, 0x00000001 }, + { 0x00110001, 0x27420231, 0x02000623, 0x00000000 }, + { 0x00000005, 0x26803e2d, 0x00000070, 0x000f000f }, + { 0x00000001, 0x270801ad, 0x00000700, 0x00000000 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000084 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00800001, 0x27200231, 0x00b10628, 0x00000000 }, + { 0x02600005, 0x20001c20, 0x0200006c, 0x00000004 }, + { 0x00680001, 0x272c0231, 0x028d0634, 0x00000000 }, + { 0x00780001, 0x272c0231, 0x02000633, 0x00000000 }, + { 0x00400001, 0x27420231, 0x00a0040e, 0x00000000 }, + { 0x00400001, 0x27460231, 0x00a0044e, 0x00000000 }, + { 0x00200001, 0x27400231, 0x0000062b, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x0200006c, 0x00000002 }, + { 0x00010001, 0x27400231, 0x0200062c, 0x00000000 }, + { 0x00110001, 0x27410231, 0x0200040e, 0x00000000 }, + { 0x00000008, 0x26803e2d, 0x00000070, 0x00040004 }, + { 0x00200040, 0x22083d8c, 0x00450208, 0x00800080 }, + { 0x00000006, 0x27083dad, 0x00000700, 0x00010001 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000064 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00a02001, 0xb3800231, 0x00d20400, 0x00000000 }, + { 0x00a02001, 0xb3a00231, 0x00d20480, 0x00000000 }, + { 0x00a02001, 0xb3c00231, 0x00d20440, 0x00000000 }, + { 0x00a02001, 0xb3e00231, 0x00d204c0, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x00000060, 0x00000002 }, + { 0x01000005, 0x20001c20, 0x0200006c, 0x00000001 }, + { 0x01010005, 0x20001c20, 0x00000060, 0x00004000 }, + { 0x02010005, 0x20001c20, 0x02000060, 0x00000010 }, + { 0x00030001, 0x27230231, 0x0000065f, 0x00000000 }, + { 0x00030220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02600005, 0x20001c20, 0x0000006c, 0x00000001 }, + { 0x00460001, 0x27200231, 0x0069065c, 0x00000000 }, + { 0x00110001, 0x27230231, 0x00000458, 0x00000000 }, + { 0x00600001, 0x27240231, 0x008d0458, 0x00000000 }, + { 0x00600001, 0x272c0231, 0x008d0478, 0x00000000 }, + { 0x00600001, 0x27420231, 0x00cf0663, 0x00000000 }, + { 0x00000001, 0x27400231, 0x00000458, 0x00000000 }, + { 0x00010001, 0x27410231, 0x0000065f, 0x00000000 }, + { 0x00110001, 0x27410231, 0x00000663, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x0200006c, 0x00000010 }, + { 0x00110001, 0x27420231, 0x0200065f, 0x00000000 }, + { 0x00000005, 0x26803e2d, 0x00000071, 0x000f000f }, + { 0x00000006, 0x27083dad, 0x00000702, 0x00020002 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000030 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00400001, 0x27200231, 0x0069045c, 0x00000000 }, + { 0x00600001, 0x27240231, 0x008d0478, 0x00000000 }, + { 0x00800001, 0x272c0231, 0x0000047f, 0x00000000 }, + { 0x00400001, 0x27420231, 0x00a0048e, 0x00000000 }, + { 0x00400001, 0x27460231, 0x00a004ce, 0x00000000 }, + { 0x00000001, 0x27400231, 0x00000478, 0x00000000 }, + { 0x00000001, 0x27410231, 0x0000045f, 0x00000000 }, + { 0x00000008, 0x26803e2d, 0x00000071, 0x00040004 }, + { 0x00200040, 0x22083d8c, 0x00450208, 0x00800080 }, + { 0x00000006, 0x27083dad, 0x00000700, 0x00030003 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00a02001, 0xb3800231, 0x00d20480, 0x00000000 }, + { 0x00a02001, 0xb3a00231, 0x00d20500, 0x00000000 }, + { 0x00a02001, 0xb3c00231, 0x00d204c0, 0x00000000 }, + { 0x00a02001, 0xb3e00231, 0x00d20540, 0x00000000 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000100 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0xfffffe74 }, + { 0x00000001, 0x27800021, 0x00000780, 0x00000000 }, + { 0x00000001, 0x27a00021, 0x000007a0, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000001, 0x27340231, 0x00000733, 0x00000000 }, + { 0x00600001, 0x274a0231, 0x00000749, 0x00000000 }, + { 0x00800040, 0x24003e2c, 0x00b10723, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10724, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10725, 0x00010001 }, + { 0x00800008, 0x26c03d8d, 0x008d0400, 0x00020002 }, + { 0x00800040, 0x24003e2c, 0x00b10740, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10741, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10742, 0x00010001 }, + { 0x00800008, 0x26a03d8d, 0x008d0400, 0x00020002 }, + { 0x00800001, 0x27240231, 0x00d206c0, 0x00000000 }, + { 0x00600001, 0x27400231, 0x00ae06a2, 0x00000000 }, + { 0x00000001, 0x27230231, 0x000006a0, 0x00000000 }, + { 0x00000005, 0x220e3dac, 0x00000680, 0x000f000f }, + { 0x00000001, 0x26a00221, 0x00009c2c, 0x00000000 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a0 }, + { 0x80800040, 0xd00045b1, 0x00b18800, 0x000d0724 }, + { 0x80800040, 0xd02045b1, 0x00b18820, 0x000d0724 }, + { 0x80800040, 0xd04045b1, 0x00b18840, 0x000d0724 }, + { 0x80800040, 0xd06045b1, 0x00b18860, 0x000d0724 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x80800040, 0xd00045b1, 0x00b18800, 0x002c0740 }, + { 0x80800040, 0xd02045b1, 0x00b18820, 0x002c0742 }, + { 0x80800040, 0xd04045b1, 0x00b18840, 0x002c0744 }, + { 0x80800040, 0xd06045b1, 0x00b18860, 0x002c0746 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x02800005, 0x20003da0, 0x00000708, 0x00020002 }, + { 0x02600005, 0x20003da0, 0x02000708, 0x00010001 }, + { 0x009a0001, 0x27200169, 0x00000000, 0x80808080 }, + { 0x00780001, 0x27400231, 0x028d0724, 0x00000000 }, + { 0x00780001, 0x27240231, 0x008d0740, 0x00000000 }, + { 0x00600040, 0x25e04629, 0x008d0724, 0x008d0740 }, + { 0x00400040, 0x25e02529, 0x006905e0, 0x006905e8 }, + { 0x00200040, 0x25e02529, 0x004505e0, 0x004505e4 }, + { 0x00800040, 0x2400252c, 0x000005e0, 0x000005e2 }, + { 0x00800040, 0x24003d8c, 0x008d0400, 0x00080008 }, + { 0x00800008, 0x26803d8d, 0x008d0400, 0x00040004 }, + { 0x80800040, 0xd00035b1, 0x00b18800, 0x00b10680 }, + { 0x80800040, 0xd02035b1, 0x00b18820, 0x00b10680 }, + { 0x80800040, 0xd04035b1, 0x00b18840, 0x00b10680 }, + { 0x80800040, 0xd06035b1, 0x00b18860, 0x00b10680 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00600001, 0x27340231, 0x008d0733, 0x00000000 }, + { 0x00800040, 0x24003e2c, 0x00b10726, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10725, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10724, 0x00010001 }, + { 0x00800008, 0x26803d8d, 0x00b10400, 0x00020002 }, + { 0x80800040, 0xd00035b1, 0x00b18800, 0x002d0680 }, + { 0x80800040, 0xd02035b1, 0x00b18820, 0x002d0684 }, + { 0x80800040, 0xd04035b1, 0x00b18840, 0x002d0688 }, + { 0x80800040, 0xd06035b1, 0x00b18860, 0x002d068c }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00400009, 0x26c05421, 0x00000744, 0x00690050 }, + { 0x00400009, 0x26d05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26a00231, 0x00cf06c3, 0x00000000 }, + { 0x00800001, 0x26a80231, 0x00b10723, 0x00000000 }, + { 0x00800040, 0x24003e2c, 0x00b106a2, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a1, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a0, 0x00010001 }, + { 0x00800008, 0x26a03d8d, 0x00b10400, 0x00020002 }, + { 0x00200040, 0x220c3eac, 0x00450056, 0x06a006a0 }, + { 0x80800040, 0xd06035b1, 0x01ed9800, 0x00b18860 }, + { 0x80800040, 0xd04035b1, 0x01ed9804, 0x00b18840 }, + { 0x80800040, 0xd02035b1, 0x01ed9808, 0x00b18820 }, + { 0x80800040, 0xd00035b1, 0x01ed980c, 0x00b18800 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00400009, 0x26c05421, 0x00000744, 0x00690050 }, + { 0x00400009, 0x26d05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26a00231, 0x00cf06c3, 0x00000000 }, + { 0x00800001, 0x26a80231, 0x00b10723, 0x00000000 }, + { 0x00800042, 0x25c04629, 0x00b106a8, 0x00b106a9 }, + { 0x00800040, 0x24003e2c, 0x00b106a3, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a2, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a1, 0x00010001 }, + { 0x00800008, 0x26c03d8d, 0x008d0400, 0x00020002 }, + { 0x00600001, 0x26a00231, 0x00ae06c0, 0x00000000 }, + { 0x00600001, 0x46a60231, 0x00ae06cc, 0x00000000 }, + { 0x00600001, 0x46a70231, 0x00ae05c0, 0x00000000 }, + { 0x00200040, 0x220c3eac, 0x00450036, 0x06a006a0 }, + { 0x80800040, 0xd0603631, 0x01ee9800, 0x00b18860 }, + { 0x80800040, 0xd0403631, 0x01ee9802, 0x00b18840 }, + { 0x80800040, 0xd0203631, 0x01ee9804, 0x00b18820 }, + { 0x80800040, 0xd0003631, 0x01ee9806, 0x00b18800 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00400009, 0x26c05421, 0x00000744, 0x00690050 }, + { 0x00400009, 0x26d05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26a00231, 0x00ab06c3, 0x00000000 }, + { 0x00800001, 0x26a80231, 0x00b10723, 0x00000000 }, + { 0x00800040, 0x24003e2c, 0x00b106a2, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a1, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a0, 0x00010001 }, + { 0x00800008, 0x25c03d89, 0x008d0400, 0x00020002 }, + { 0x00800042, 0x26a0462d, 0x00b106a0, 0x00b106a1 }, + { 0x00600001, 0x46a10231, 0x00ae05c0, 0x00000000 }, + { 0x00600001, 0x26b00231, 0x00ae05d0, 0x00000000 }, + { 0x00200040, 0x220c3eac, 0x00450056, 0x06a006a0 }, + { 0x80800040, 0xd0603631, 0x01ed9800, 0x00b18860 }, + { 0x80800040, 0xd0403631, 0x01ed9804, 0x00b18840 }, + { 0x80800040, 0xd0203631, 0x01ed9808, 0x00b18820 }, + { 0x80800040, 0xd0003631, 0x01ed980c, 0x00b18800 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00800042, 0x25c04629, 0x00b10724, 0x00b10725 }, + { 0x00800040, 0x24003e2c, 0x00b10726, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10725, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10724, 0x00010001 }, + { 0x00800008, 0x25e03d89, 0x00240400, 0x00020002 }, + { 0x80800040, 0xd0003531, 0x00ad05c0, 0x00b18800 }, + { 0x80800040, 0xd0203531, 0x00ad05c2, 0x00b18820 }, + { 0x80800040, 0xd0403531, 0x00ad05c4, 0x00b18840 }, + { 0x80800040, 0xd0603531, 0x00ad05c6, 0x00b18860 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00600001, 0x27480231, 0x00000747, 0x00000000 }, + { 0x00800042, 0x25c04629, 0x00b10740, 0x00b10741 }, + { 0x00800040, 0x24003e2c, 0x00b10742, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10741, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10740, 0x00010001 }, + { 0x00800008, 0x25e03d89, 0x00240400, 0x00020002 }, + { 0x00800001, 0x45c10231, 0x00d205e0, 0x00000000 }, + { 0x80800040, 0xd0003631, 0x004d05c0, 0x00b18800 }, + { 0x80800040, 0xd0203631, 0x004d05c4, 0x00b18820 }, + { 0x80800040, 0xd0403631, 0x004d05c8, 0x00b18840 }, + { 0x80800040, 0xd0603631, 0x004d05cc, 0x00b18860 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00200801, 0x27c001a1, 0x004507fa, 0x00000000 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x0ff04000 }, + { 0x00800001, 0x20400232, 0x00cd0400, 0x00000000 }, + { 0x00800001, 0x20500232, 0x00cd0408, 0x00000000 }, + { 0x00800001, 0x20600232, 0x00cd0410, 0x00000000 }, + { 0x00800001, 0x20700232, 0x00cd0418, 0x00000000 }, + { 0x00800001, 0x20800232, 0x00cd0440, 0x00000000 }, + { 0x00800001, 0x20900232, 0x00cd0448, 0x00000000 }, + { 0x00800001, 0x20a00232, 0x00cd0450, 0x00000000 }, + { 0x00800001, 0x20b00232, 0x00cd0458, 0x00000000 }, + { 0x00800001, 0x20c00232, 0x00cd0480, 0x00000000 }, + { 0x00800001, 0x20d00232, 0x00cd0488, 0x00000000 }, + { 0x00800001, 0x20e00232, 0x00cd0490, 0x00000000 }, + { 0x00800001, 0x20f00232, 0x00cd0498, 0x00000000 }, + { 0x00800001, 0x21000232, 0x00cd04c0, 0x00000000 }, + { 0x00800001, 0x21100232, 0x00cd04c8, 0x00000000 }, + { 0x00800001, 0x21200232, 0x00cd04d0, 0x00000000 }, + { 0x00800001, 0x21300232, 0x00cd04d8, 0x00000000 }, + { 0x01600031, 0x27800021, 0x508d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00600001, 0x27c00021, 0x008d0000, 0x00000000 }, + { 0x00200009, 0x27fa3e29, 0x00450064, 0x00040004 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0xfffffd34 }, + { 0x00000001, 0x220401ec, 0x00000000, 0x00800080 }, + { 0x00000001, 0x220601ec, 0x00000000, 0x04000400 }, + { 0x00200008, 0x27006e2d, 0x0000006c, 0x00000040 }, + { 0x02600005, 0x20003e20, 0x0000006c, 0x00040004 }, + { 0x00780001, 0x26340231, 0x00000633, 0x00000000 }, + { 0x00800001, 0x27200231, 0x00b10620, 0x00000000 }, + { 0x00600001, 0x27400231, 0x00cf0643, 0x00000000 }, + { 0x00400008, 0x26806e2d, 0x00240070, 0x00004040 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000006a }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00000006, 0x27003dad, 0x00000700, 0x00010001 }, + { 0x00800001, 0x27200231, 0x00b10628, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00808c26, 0x00000000 }, + { 0x00400001, 0x27440231, 0x00808c66, 0x00000000 }, + { 0x00400008, 0x26806e2d, 0x00240072, 0x00004040 }, + { 0x00000040, 0x22063d8c, 0x00000206, 0x00800080 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000058 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00000006, 0x27003dad, 0x00000702, 0x00020002 }, + { 0x00800001, 0xaf800231, 0x00d28f80, 0x00000000 }, + { 0x00800001, 0xaf900231, 0x00d28fa0, 0x00000000 }, + { 0x00800001, 0xafc00231, 0x00d28fc0, 0x00000000 }, + { 0x00800001, 0xafd00231, 0x00d28fe0, 0x00000000 }, + { 0x00800001, 0xafa00231, 0x00d28c00, 0x00000000 }, + { 0x00800001, 0xafb00231, 0x00d28c20, 0x00000000 }, + { 0x00800001, 0xafe00231, 0x00d28c40, 0x00000000 }, + { 0x00800001, 0xaff00231, 0x00d28c60, 0x00000000 }, + { 0x00400001, 0x27200231, 0x0069065c, 0x00000000 }, + { 0x00600001, 0x27240231, 0x00a98fcc, 0x00000000 }, + { 0x00600001, 0x272c0231, 0x00a98fec, 0x00000000 }, + { 0x00600001, 0x27340231, 0x00008fff, 0x00000000 }, + { 0x00800001, 0x2620012d, 0x00b10720, 0x00000000 }, + { 0x00600001, 0x27400231, 0x00cf0663, 0x00000000 }, + { 0x00400008, 0x26806e2d, 0x00240074, 0x00004040 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000032 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00000006, 0x27003dad, 0x00000700, 0x00010001 }, + { 0x00800001, 0x27200231, 0x00b10628, 0x00000000 }, + { 0x00600001, 0x27300231, 0x008d0638, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00808c26, 0x00000000 }, + { 0x00400001, 0x27440231, 0x00808c66, 0x00000000 }, + { 0x00400008, 0x26806e2d, 0x00240076, 0x00004040 }, + { 0x00000040, 0x22063d8c, 0x00000206, 0x00800080 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x00800001, 0xaf800231, 0x00d28f80, 0x00000000 }, + { 0x00800001, 0xaf900231, 0x00d28fa0, 0x00000000 }, + { 0x00800001, 0xafc00231, 0x00d28fc0, 0x00000000 }, + { 0x00800001, 0xafd00231, 0x00d28fe0, 0x00000000 }, + { 0x00800001, 0xafa00231, 0x00d28c00, 0x00000000 }, + { 0x00800001, 0xafb00231, 0x00d28c20, 0x00000000 }, + { 0x00800001, 0xafe00231, 0x00d28c40, 0x00000000 }, + { 0x00800001, 0xaff00231, 0x00d28c60, 0x00000000 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x000000e0 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0xfffffcce }, + { 0x00000001, 0x27800021, 0x00000780, 0x00000000 }, + { 0x00000001, 0x27a00021, 0x000007a0, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00600001, 0x27800231, 0x008d0740, 0x00000000 }, + { 0x00400005, 0x22083dac, 0x00690680, 0x000f000f }, + { 0x00400040, 0x26a04625, 0x01e09020, 0x00690058 }, + { 0x00000001, 0x26d001ad, 0x00000700, 0x00000000 }, + { 0x00000040, 0x27e41c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a0 }, + { 0x80600040, 0xcc0035b1, 0x00898800, 0x008d0760 }, + { 0x80600040, 0xcc1035b1, 0x00898820, 0x008d0770 }, + { 0x00800001, 0x27200231, 0x008d0724, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00808c06, 0x00000000 }, + { 0x00000006, 0x26d03dad, 0x00000700, 0x00010001 }, + { 0x00000040, 0x27e41c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a4 }, + { 0x80600040, 0xcc2035b1, 0x00898808, 0x008d0760 }, + { 0x80600040, 0xcc3035b1, 0x00898828, 0x008d0770 }, + { 0x00000001, 0x27230231, 0x00000783, 0x00000000 }, + { 0x00400001, 0x27240231, 0x008a8c18, 0x00000000 }, + { 0x00400001, 0x27280231, 0x008a8c38, 0x00000000 }, + { 0x00400001, 0x272c0231, 0x00008c3e, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00690784, 0x00000000 }, + { 0x00000006, 0x26d03dad, 0x00000700, 0x00020002 }, + { 0x00000040, 0x27e41c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a8 }, + { 0x80600040, 0xcc4035b1, 0x00898840, 0x008d0760 }, + { 0x80600040, 0xcc5035b1, 0x00898860, 0x008d0770 }, + { 0x00800001, 0x27200231, 0x008d0724, 0x00000000 }, + { 0x00600001, 0x27280231, 0x00000727, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00808c46, 0x00000000 }, + { 0x00000006, 0x26d03dad, 0x00000700, 0x00030003 }, + { 0x00000040, 0x27e41c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006ac }, + { 0x80600040, 0xcc6035b1, 0x00898848, 0x008d0760 }, + { 0x80600040, 0xcc7035b1, 0x00898868, 0x008d0770 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00800001, 0x2760022d, 0x00090724, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00800001, 0x2760022d, 0x00280740, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x02802005, 0x20003da0, 0x000006d0, 0x00020002 }, + { 0x02802005, 0x20003da0, 0x020006d0, 0x00010001 }, + { 0x009a0001, 0x27200169, 0x00000000, 0x80808080 }, + { 0x00780001, 0x27400231, 0x028d0724, 0x00000000 }, + { 0x00780001, 0x27240231, 0x008d0740, 0x00000000 }, + { 0x00400040, 0x25e04629, 0x00690724, 0x00690740 }, + { 0x00200040, 0x25e02529, 0x004505e0, 0x004505e4 }, + { 0x00800040, 0x2400252c, 0x000005e0, 0x000005e2 }, + { 0x00800040, 0x24003d8c, 0x008d0400, 0x00040004 }, + { 0x00800008, 0x27603d8d, 0x008d0400, 0x00030003 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00600001, 0x26c00231, 0x008d0724, 0x00000000 }, + { 0x00400001, 0x26c80231, 0x0069072b, 0x00000000 }, + { 0x00600040, 0x24003e2c, 0x008d06c2, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c1, 0x00020002 }, + { 0x00600048, 0x25e03e29, 0x008d06c0, 0x00010001 }, + { 0x00800008, 0x27603d2d, 0x002905e0, 0x00020002 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00400009, 0x26e05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26c40231, 0x008d0723, 0x00000000 }, + { 0x00400001, 0x26c00231, 0x00ab06e3, 0x00000000 }, + { 0x00600040, 0x24003e2c, 0x008d06c2, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c1, 0x00020002 }, + { 0x00600048, 0x26c03e2d, 0x008d06c0, 0x00010001 }, + { 0x00400040, 0x22083eac, 0x00690054, 0x06c006c0 }, + { 0x00800008, 0x27603dad, 0x01e99000, 0x00020002 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00400009, 0x26e05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26c40231, 0x008d0723, 0x00000000 }, + { 0x00400001, 0x26c00231, 0x00ab06e3, 0x00000000 }, + { 0x00600042, 0x25c04629, 0x008d06c4, 0x008d06c5 }, + { 0x00600040, 0x24003e2c, 0x008d06c3, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c2, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c1, 0x00010001 }, + { 0x00600008, 0x26c03d8d, 0x008d0400, 0x00020002 }, + { 0x00400001, 0x46c401ad, 0x006906c4, 0x00000000 }, + { 0x00400001, 0x46c6012d, 0x006905c0, 0x00000000 }, + { 0x00400040, 0x22083eac, 0x00690054, 0x06c006c0 }, + { 0x00800001, 0x276001ad, 0x01ea9000, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00400009, 0x26e05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26c40231, 0x008d0723, 0x00000000 }, + { 0x00400001, 0x26c00231, 0x00ab06e3, 0x00000000 }, + { 0x00600042, 0x25c04629, 0x008d06c0, 0x008d06c1 }, + { 0x00600040, 0x24003e2c, 0x008d06c2, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c1, 0x00020002 }, + { 0x00600048, 0x26e03e2d, 0x008d06c0, 0x00010001 }, + { 0x00400008, 0x46c23dad, 0x006906e0, 0x00020002 }, + { 0x00200008, 0x26d03dad, 0x004506e8, 0x00020002 }, + { 0x00400001, 0x46c0012d, 0x006905c0, 0x00000000 }, + { 0x00400009, 0x22083eac, 0x00690054, 0x00010001 }, + { 0x00400040, 0x22083d8c, 0x00690208, 0x06c006c0 }, + { 0x00800001, 0x276001ad, 0x01e99000, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00600042, 0x45c04629, 0x008d0724, 0x008d0725 }, + { 0x00600040, 0x24003e2c, 0x008d0726, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d0725, 0x00020002 }, + { 0x00600048, 0x25e03e29, 0x008d0724, 0x00010001 }, + { 0x00600008, 0x45c23d29, 0x008d05e0, 0x00020002 }, + { 0x00800001, 0x2760012d, 0x002a05c0, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00600001, 0x27440231, 0x00000743, 0x00000000 }, + { 0x00600042, 0x45c04629, 0x008d0740, 0x008d0741 }, + { 0x00600040, 0x24003e2c, 0x008d0742, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d0741, 0x00020002 }, + { 0x00600048, 0x25e03e29, 0x008d0740, 0x00010001 }, + { 0x00600008, 0x45c23d29, 0x008d05e0, 0x00020002 }, + { 0x00800001, 0x2760012d, 0x004905c0, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00200801, 0x27c001a1, 0x004507fa, 0x00000000 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x0ff04000 }, + { 0x00800001, 0x20400232, 0x00a90400, 0x00000000 }, + { 0x00800001, 0x20500232, 0x00a90404, 0x00000000 }, + { 0x00800001, 0x20600232, 0x00a90408, 0x00000000 }, + { 0x00800001, 0x20700232, 0x00a9040c, 0x00000000 }, + { 0x00800001, 0x20800232, 0x00a90440, 0x00000000 }, + { 0x00800001, 0x20900232, 0x00a90444, 0x00000000 }, + { 0x00800001, 0x20a00232, 0x00a90448, 0x00000000 }, + { 0x00800001, 0x20b00232, 0x00a9044c, 0x00000000 }, + { 0x00800001, 0x20c00232, 0x00a90480, 0x00000000 }, + { 0x00800001, 0x20d00232, 0x00a90484, 0x00000000 }, + { 0x00800001, 0x20e00232, 0x00a90488, 0x00000000 }, + { 0x00800001, 0x20f00232, 0x00a9048c, 0x00000000 }, + { 0x00800001, 0x21000232, 0x00a904c0, 0x00000000 }, + { 0x00800001, 0x21100232, 0x00a904c4, 0x00000000 }, + { 0x00800001, 0x21200232, 0x00a904c8, 0x00000000 }, + { 0x00800001, 0x21300232, 0x00a904cc, 0x00000000 }, + { 0x01600031, 0x27800021, 0x508d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00600001, 0x27c00021, 0x008d0000, 0x00000000 }, + { 0x00200009, 0x27fa3e29, 0x00450064, 0x00040004 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00200809, 0x27c03e21, 0x00450064, 0x00040004 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x1218a000 }, + { 0x00a02401, 0x20400232, 0x00b10080, 0x00000000 }, + { 0x00a02801, 0x20500232, 0x00b10090, 0x00000000 }, + { 0x00a02401, 0x20800232, 0x00b100c0, 0x00000000 }, + { 0x00a02801, 0x20900232, 0x00b100d0, 0x00000000 }, + { 0x00a02401, 0x20c00232, 0x00b10100, 0x00000000 }, + { 0x00a02801, 0x20d00232, 0x00b10110, 0x00000000 }, + { 0x00a02401, 0x21000232, 0x00b10140, 0x00000000 }, + { 0x00a02801, 0x21100232, 0x00b10150, 0x00000000 }, + { 0x01600031, 0x27800021, 0x508d07c0, 0x00000200 }, + { 0x00000401, 0x20280062, 0x00000000, 0x0007000f }, + { 0x0000080c, 0x20243c22, 0x000007c4, 0x00010001 }, + { 0x00000040, 0x22001c00, 0x00000200, 0xf8000001 }, + { 0x00800001, 0x40400232, 0x00b10180, 0x00000000 }, + { 0x00800001, 0x40410232, 0x00b101c0, 0x00000000 }, + { 0x00800001, 0x40600232, 0x00b10190, 0x00000000 }, + { 0x00800001, 0x40610232, 0x00b101d0, 0x00000000 }, + { 0x00800001, 0x40800232, 0x00b101a0, 0x00000000 }, + { 0x00800001, 0x40810232, 0x00b101e0, 0x00000000 }, + { 0x00800001, 0x40a00232, 0x00b101b0, 0x00000000 }, + { 0x00800001, 0x40a10232, 0x00b101f0, 0x00000000 }, + { 0x01600031, 0x27a00001, 0x508d0000, 0x00000200 }, + { 0x00000001, 0x27800021, 0x00000780, 0x00000000 }, + { 0x00000001, 0x27a00021, 0x000007a0, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00600001, 0x20400021, 0x008d0000, 0x00000000 }, + { 0x00000005, 0x20203e2d, 0x00000061, 0x001f001f }, + { 0x00200009, 0x20643e2d, 0x00450064, 0x00040004 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00000000 }, + { 0x01000010, 0x20003dac, 0x00000020, 0x00160016 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x01200010, 0x20003e2c, 0x0200006c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000024 }, + { 0x00800009, 0x25403dad, 0x00050064, 0x00020002 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00180018 }, + { 0x00200040, 0x45483dad, 0x00660548, 0x00100010 }, + { 0x00200040, 0x45523dad, 0x00660552, 0x00100010 }, + { 0x00400040, 0x25583dad, 0x00690558, 0x00100010 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0x00200020 }, + { 0x00800040, 0x20a035ad, 0x00b100a0, 0x00b10540 }, + { 0x00600040, 0x45423dad, 0x00ae0542, 0x00200020 }, + { 0x00800040, 0x20e035ad, 0x00b100e0, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0xffe0ffe0 }, + { 0x00800040, 0x20c035ad, 0x00b100c0, 0x00b10540 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000014 }, + { 0x06600010, 0x20003dac, 0x02000020, 0x00030003 }, + { 0x00000001, 0x206c0171, 0x00000000, 0x00000000 }, + { 0x00610001, 0x208000a5, 0x02050080, 0x00000000 }, + { 0x00610001, 0x212000a5, 0x02090100, 0x00000000 }, + { 0x00410001, 0x211000a5, 0x02690100, 0x00000000 }, + { 0x00200009, 0x25403dad, 0x00450064, 0x00020002 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00050540 }, + { 0x00200040, 0x40883dad, 0x00660088, 0x00200020 }, + { 0x00200040, 0x40923dad, 0x00660092, 0x00200020 }, + { 0x00400040, 0x20983dad, 0x00690098, 0x00200020 }, + { 0x00200401, 0x22080060, 0x00000000, 0x03400140 }, + { 0x00000c01, 0x220c0060, 0x00000000, 0x04400080 }, + { 0x00000801, 0x22040060, 0x00000000, 0x01000070 }, + { 0x00000001, 0x20200169, 0x00000000, 0x00000000 }, + { 0x0000000c, 0x2458262d, 0x0000006d, 0x00000020 }, + { 0x00000001, 0x210e0169, 0x00000000, 0x00010001 }, + { 0x00000005, 0x24583dad, 0x00000458, 0x00030003 }, + { 0x01000010, 0x200035ac, 0x0000010e, 0x00000458 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x000002fe }, + { 0x01000010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00010401, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00010805, 0x245a3e2d, 0x02008800, 0x007f007f }, + { 0x00110401, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00110805, 0x245a3e2d, 0x02008804, 0x007f007f }, + { 0x0000000c, 0x211e362d, 0x0000006c, 0x00000020 }, + { 0x00000040, 0x20780d21, 0x0000045a, 0x0208a002 }, + { 0x02000005, 0x20003dac, 0x0200011e, 0x00030003 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000001be }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000012 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00700000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c000c }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x0007000c }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00110040, 0x20441da5, 0x02000442, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c0007 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00070007 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x01600006, 0x20003dac, 0x0000044a, 0x00000000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00010040, 0x22000c20, 0x00000078, 0x00200010 }, + { 0x00110040, 0x22000c20, 0x00000078, 0x00300010 }, + { 0x00010001, 0x20480061, 0x00000000, 0x00030009 }, + { 0x00110001, 0x20480061, 0x00000000, 0x00040009 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001a }, + { 0x00600001, 0x26220231, 0x008d05c0, 0x00000000 }, + { 0x00600401, 0x26120231, 0x008d05b8, 0x00000000 }, + { 0x00600801, 0x26020231, 0x008d05b0, 0x00000000 }, + { 0x00600401, 0x25f20231, 0x008d05a8, 0x00000000 }, + { 0x00600801, 0x25e20231, 0x008d05a0, 0x00000000 }, + { 0x00600401, 0x25d20231, 0x008d0598, 0x00000000 }, + { 0x00600801, 0x25c20231, 0x008d0590, 0x00000000 }, + { 0x00600401, 0x25b20231, 0x008d0588, 0x00000000 }, + { 0x00600801, 0x25a20231, 0x008d0580, 0x00000000 }, + { 0x00600401, 0x25920231, 0x008d0578, 0x00000000 }, + { 0x00600801, 0x25820231, 0x008d0570, 0x00000000 }, + { 0x00600001, 0x25720231, 0x008d0568, 0x00000000 }, + { 0x00600001, 0x25620231, 0x008d0560, 0x00000000 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00810001, 0xbc000229, 0x02ad8000, 0x00000000 }, + { 0x00810001, 0xbc200229, 0x02ad8020, 0x00000000 }, + { 0x00810001, 0xbc400229, 0x02ad8040, 0x00000000 }, + { 0x00810001, 0xbc600229, 0x02ad8060, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000012a }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000066 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000064 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04a00480 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00a02040, 0x2400462c, 0x00ad8000, 0x00ad8005 }, + { 0x00a02048, 0x24003e2c, 0x00ad8001, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8002, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8003, 0x00140014 }, + { 0x00a02048, 0xb8003e2d, 0x00ad8004, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8040, 0x00ad8045 }, + { 0x00a02048, 0x24003e2c, 0x00ad8041, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8042, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8043, 0x00140014 }, + { 0x00a02048, 0xb8403e2d, 0x00ad8044, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8080, 0x00ad8085 }, + { 0x00a02048, 0x24003e2c, 0x00ad8081, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8082, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8083, 0x00140014 }, + { 0x00a02048, 0xb8803e2d, 0x00ad8084, 0xfffbfffb }, + { 0x00600040, 0x2400462c, 0x008d80c0, 0x008d80c5 }, + { 0x00600048, 0x24003e2c, 0x008d80c1, 0xfffbfffb }, + { 0x00600048, 0x24003e2c, 0x008d80c2, 0x00140014 }, + { 0x00600048, 0x24003e2c, 0x008d80c3, 0x00140014 }, + { 0x00600048, 0xb8c03e2d, 0x008d80c4, 0xfffbfffb }, + { 0x00000401, 0x22000060, 0x00000000, 0x04a00480 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04a00490 }, + { 0x00010802, 0x220c2d28, 0x00000454, 0x04800480 }, + { 0x00a02040, 0x24003dac, 0x00b18000, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8800, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8820, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18020, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8820, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8840, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18040, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x000a000a }, + { 0x00a02040, 0x24003dac, 0x00b18040, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18060, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8860, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8880, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18080, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8880, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed88a0, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x000000b4 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000004c }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000004a }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010002, 0x220c2d28, 0x02000454, 0x04800480 }, + { 0x00800040, 0x24003e2c, 0x01ed8000, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8001, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8021, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8002, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8022, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8003, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8023, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8004, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8024, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8005, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8025, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00800040, 0x24003e2c, 0x01ed8040, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8060, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8041, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8061, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8042, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8062, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8043, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8063, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8044, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8064, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8045, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8065, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000005c }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000062 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000042 }, + { 0x00200401, 0x22000128, 0x0066044c, 0x00000000 }, + { 0x00200c01, 0x22040128, 0x0045044e, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x03400010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220c0168, 0x02000000, 0x04800480 }, + { 0x00a02040, 0x24003e2c, 0x00ad83e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8be0, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8800, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8800, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8820, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00a02040, 0x24003e2c, 0x00ad8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8040, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8840, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8860, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8060, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8860, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8880, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000001c }, + { 0x00200401, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x01200010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01200010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00210040, 0x22002d08, 0x00450200, 0x00010001 }, + { 0x00210040, 0x22002d08, 0x02450200, 0x00100010 }, + { 0x00800001, 0xb8000229, 0x01ed8000, 0x00000000 }, + { 0x00800001, 0xb8200229, 0x01ed8020, 0x00000000 }, + { 0x00800001, 0xb8400229, 0x01ed8040, 0x00000000 }, + { 0x00800001, 0xb8600229, 0x01ed8060, 0x00000000 }, + { 0x80800042, 0xd8004631, 0x00d29800, 0x00d20480 }, + { 0x80800042, 0xd8204631, 0x00d29820, 0x00d204a0 }, + { 0x80800042, 0xd8404631, 0x00d29840, 0x00d204c0 }, + { 0x80800042, 0xd8604631, 0x00d29860, 0x00d204e0 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000c01, 0x220201e8, 0x00000000, 0x07100710 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00200040, 0x244c3d09, 0x00450200, 0x00100010 }, + { 0x00800041, 0x24002628, 0x00ad8000, 0x00000540 }, + { 0x00800041, 0x24202628, 0x00ad8020, 0x00000540 }, + { 0x00800048, 0x24002628, 0x00ad8002, 0x00000542 }, + { 0x00800048, 0x24202628, 0x00ad8022, 0x00000542 }, + { 0x00200001, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00800048, 0x24002628, 0x01ed8000, 0x00000544 }, + { 0x00800048, 0x24202628, 0x01ed8020, 0x00000544 }, + { 0x00800048, 0xbc002629, 0x01ed8002, 0x00000546 }, + { 0x00801048, 0xbc202629, 0x01ed8022, 0x00000546 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000126 }, + { 0x00000001, 0x20220169, 0x00000000, 0x00040004 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00500000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x00080008 }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00030008 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0x22000c20, 0x00000078, 0x00100010 }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00000001, 0x20480061, 0x00000000, 0x00020005 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00410001, 0xbc000229, 0x02698000, 0x00000000 }, + { 0x00410001, 0xbc100229, 0x02698010, 0x00000000 }, + { 0x00410001, 0xbc200229, 0x02698020, 0x00000000 }, + { 0x00410001, 0xbc300229, 0x02698030, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000000be }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000044 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000042 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04d004c0 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00802040, 0x2400462c, 0x00a98000, 0x00a98005 }, + { 0x00802048, 0x24003e2c, 0x00a98001, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98002, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98003, 0x00140014 }, + { 0x00802048, 0xb8003e2d, 0x00a98004, 0xfffbfffb }, + { 0x00802040, 0x2400462c, 0x00a98040, 0x00a98045 }, + { 0x00802048, 0x24003e2c, 0x00a98041, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98042, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98043, 0x00140014 }, + { 0x00802048, 0xb8203e2d, 0x00a98044, 0xfffbfffb }, + { 0x00400040, 0x2400462c, 0x00698080, 0x00698085 }, + { 0x00400048, 0x24003e2c, 0x00698081, 0xfffbfffb }, + { 0x00400048, 0x24003e2c, 0x00698082, 0x00140014 }, + { 0x00400048, 0x24003e2c, 0x00698083, 0x00140014 }, + { 0x00400048, 0xb8403e2d, 0x00698084, 0xfffbfffb }, + { 0x00000401, 0x220c0060, 0x00000000, 0x04e004d0 }, + { 0x00000c01, 0x22000060, 0x00000000, 0x04d004c8 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04e004d8 }, + { 0x00000801, 0x22080060, 0x00000000, 0x04f004e8 }, + { 0x00800040, 0x24003dac, 0x00b104c0, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x00b104e0, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed9800, 0x00140014 }, + { 0x00010001, 0x220c0168, 0x00000000, 0x05400540 }, + { 0x00110001, 0x220c0168, 0x00000000, 0x04c004c0 }, + { 0x00800048, 0x24003dac, 0x01e98000, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01e98020, 0x00010001 }, + { 0x00800048, 0x24003dac, 0x01e98800, 0x00140014 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000060 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000022 }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010001, 0x220e0168, 0x02000000, 0x05400540 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x00800040, 0x24003e2c, 0x01e98000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98001, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98002, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98003, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98004, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98005, 0x00010001 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x03600010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x00400401, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x00400840, 0x22083d2c, 0x0069044c, 0x00100010 }, + { 0x00610040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00800040, 0x24003e2c, 0x01e983e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e993e0, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e99000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e99020, 0x00010001 }, + { 0x00000001, 0x220e0168, 0x00000000, 0x05400540 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x00400001, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x01400010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01400010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00010001 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00100010 }, + { 0x00800001, 0x25400229, 0x01e98000, 0x00000000 }, + { 0x80800042, 0x45404631, 0x00d20540, 0x00d204c0 }, + { 0x00000001, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x00400001, 0xd8000231, 0x008a0540, 0x00000000 }, + { 0x00400001, 0xd8100231, 0x008a0548, 0x00000000 }, + { 0x00400001, 0xd8200231, 0x008a0550, 0x00000000 }, + { 0x00400001, 0xd8300231, 0x008a0558, 0x00000000 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00600041, 0x24002628, 0x00898000, 0x00000540 }, + { 0x00600048, 0x24002628, 0x00898002, 0x00000542 }, + { 0x00600048, 0x24002628, 0x00898008, 0x00000544 }, + { 0x00600048, 0x2540262d, 0x0089800a, 0x00000546 }, + { 0x00400401, 0xbc0001a9, 0x00690540, 0x00000000 }, + { 0x00400801, 0xbc1001a9, 0x00690548, 0x00000000 }, + { 0x01000010, 0x20003d2c, 0x00000022, 0x00030003 }, + { 0x01000040, 0x20223d29, 0x02000022, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00080008 }, + { 0x00110040, 0x24540c21, 0x00000454, 0x00080008 }, + { 0x00010040, 0x24540c21, 0x00000454, 0x00180038 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffee6 }, + { 0x01800010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0xffe0ffe0 }, + { 0x00010001, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00110001, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00000001, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00800040, 0x24003dac, 0x00b19c00, 0x00200020 }, + { 0x00800040, 0x24203dac, 0x00b19c20, 0x00200020 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00060006 }, + { 0x8080000c, 0xdc203d91, 0x00b10420, 0x00060006 }, + { 0x02000040, 0x210e3d29, 0x0200010e, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00040004 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffcf8 }, + { 0x01000005, 0x25643e2d, 0x00000060, 0x00c000c0 }, + { 0x01000010, 0x20003dac, 0x02000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000086 }, + { 0x00a02001, 0x46400231, 0x00d20780, 0x00000000 }, + { 0x00a02001, 0x46800231, 0x00d207c0, 0x00000000 }, + { 0x00a02001, 0x46c00231, 0x00d20480, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000007e }, + { 0x80a02042, 0x46404631, 0x00d20640, 0x00d20780 }, + { 0x80a02042, 0x46804631, 0x00d20680, 0x00d207c0 }, + { 0x80a02042, 0x46c04631, 0x00d206c0, 0x00d20480 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000076 }, + { 0x01000010, 0x20003dac, 0x02000564, 0x00800080 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffffe4 }, + { 0x00200401, 0x256801ed, 0x00000000, 0x00200020 }, + { 0x00200801, 0x256c01ed, 0x00000000, 0x00060006 }, + { 0x00400001, 0x258001ed, 0x00000000, 0x00000000 }, + { 0x00600001, 0x45a001ad, 0x00058c00, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000028 }, + { 0x02600005, 0x20006e28, 0x02008c0c, 0x88848421 }, + { 0x0220000c, 0x25663dad, 0x00000458, 0x00010001 }, + { 0x0000000c, 0x25643dad, 0x00000564, 0x00060006 }, + { 0x00310001, 0x25c401ad, 0x00000458, 0x00000000 }, + { 0x00210001, 0x25c00061, 0x00000000, 0x00010001 }, + { 0x00310040, 0x25c03dad, 0x004545c4, 0x00010001 }, + { 0x00610001, 0x25a00061, 0x02000000, 0x00000080 }, + { 0x00710001, 0x45a002ad, 0x02ae8c00, 0x00000000 }, + { 0x00710001, 0x45a202ad, 0x02ae8c01, 0x00000000 }, + { 0x00800041, 0x25a035ad, 0x00b105a0, 0x000905c0 }, + { 0x01200010, 0x20003e2c, 0x0245006e, 0x00000000 }, + { 0x00310009, 0x244c45ad, 0x02000564, 0x0045006e }, + { 0x00210001, 0x244c01ed, 0x02000000, 0x00000000 }, + { 0x0031000c, 0x244c3dad, 0x0245044c, 0x00010001 }, + { 0x00200009, 0x256835ad, 0x0045044c, 0x00000566 }, + { 0x00210040, 0x256835ad, 0x02450568, 0x00000566 }, + { 0x00200040, 0x256c362d, 0x0045006e, 0x00000566 }, + { 0x00400040, 0x240035ac, 0x00ab05a2, 0x00ab05a6 }, + { 0x00400040, 0x2400358c, 0x00690400, 0x00000566 }, + { 0x0040000c, 0x2580358d, 0x00690400, 0x00000566 }, + { 0x00800041, 0x2400362c, 0x00d20640, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d20660, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d20780, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207a0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x464035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x466035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d20680, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d206a0, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d207c0, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207e0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x468035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x46a035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d206c0, 0x000705a8 }, + { 0x00800041, 0x2420362c, 0x00d206e0, 0x000705a8 }, + { 0x00800048, 0x2400362c, 0x00d20480, 0x000705ac }, + { 0x00800048, 0x2420362c, 0x00d204a0, 0x000705ac }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x0000056a }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x0000056a }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056e }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056e }, + { 0x80800040, 0x46c035b1, 0x00b104c0, 0x00050582 }, + { 0x80800040, 0x46e035b1, 0x00b104e0, 0x00050582 }, + { 0x01800005, 0x20003d2c, 0x02000020, 0x00020002 }, + { 0x80800040, 0xd00045b1, 0x00b19000, 0x00d20640 }, + { 0x80800040, 0xd02045b1, 0x00b19020, 0x00d20660 }, + { 0x80800040, 0xd04045b1, 0x00b19040, 0x00d20680 }, + { 0x80800040, 0xd06045b1, 0x00b19060, 0x00d206a0 }, + { 0x00000040, 0x22083d8c, 0x00000208, 0x00800080 }, + { 0x80400040, 0xd40045b1, 0x00699400, 0x00ab06c0 }, + { 0x80400040, 0xd48045b1, 0x00699480, 0x00ab06c2 }, + { 0x80400040, 0xd42045b1, 0x00699420, 0x00ab06e0 }, + { 0x80400040, 0xd4a045b1, 0x006994a0, 0x00ab06e2 }, + { 0x80400040, 0xd41045b1, 0x00699410, 0x00ab06d0 }, + { 0x80400040, 0xd49045b1, 0x00699490, 0x00ab06d2 }, + { 0x80400040, 0xd43045b1, 0x00699430, 0x00ab06f0 }, + { 0x80400040, 0xd4b045b1, 0x006994b0, 0x00ab06f2 }, + { 0x00000040, 0x220a3d8c, 0x0000020a, 0x00080008 }, + { 0x00110040, 0x220a3d8c, 0x0200020a, 0x00300030 }, + { 0x01000010, 0x20003d2c, 0x02000020, 0x00060006 }, + { 0x00000040, 0x20203d29, 0x00000020, 0x00020002 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00100001 }, + { 0x00000040, 0x220c358c, 0x0000020c, 0x0000003a }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffc36 }, + { 0x00000001, 0x220001ec, 0x00000000, 0x01400140 }, + { 0x00000001, 0x220201ec, 0x00000000, 0x01c001c0 }, + { 0x00800401, 0x20400236, 0x01ee8000, 0x00000000 }, + { 0x00800801, 0x20500236, 0x01ee8010, 0x00000000 }, + { 0x00800401, 0x20600236, 0x01ee8020, 0x00000000 }, + { 0x00800801, 0x20700236, 0x01ee8030, 0x00000000 }, + { 0x00800401, 0x20800236, 0x01ee8040, 0x00000000 }, + { 0x00800801, 0x20900236, 0x01ee8050, 0x00000000 }, + { 0x00800401, 0x20a00236, 0x01ee8060, 0x00000000 }, + { 0x00800801, 0x20b00236, 0x01ee8070, 0x00000000 }, + { 0x00800401, 0x20c00236, 0x01ee8100, 0x00000000 }, + { 0x00800801, 0x20d00236, 0x01ee8110, 0x00000000 }, + { 0x00800401, 0x20e00236, 0x01ee8120, 0x00000000 }, + { 0x00800801, 0x20f00236, 0x01ee8130, 0x00000000 }, + { 0x00800401, 0x21000236, 0x01ee8140, 0x00000000 }, + { 0x00800801, 0x21100236, 0x01ee8150, 0x00000000 }, + { 0x00800401, 0x21200236, 0x01ee8160, 0x00000000 }, + { 0x00800801, 0x21300236, 0x01ee8170, 0x00000000 }, + { 0x00200401, 0x204001a5, 0x00450064, 0x00000000 }, + { 0x00000801, 0x20480061, 0x00000000, 0x000f000f }, + { 0x01600031, 0x21401c21, 0x508d0040, 0x1218a000 }, + { 0x00000001, 0x220001ec, 0x00000000, 0x03400340 }, + { 0x00800401, 0x41600236, 0x00d28000, 0x00000000 }, + { 0x00800801, 0x41610236, 0x00d28080, 0x00000000 }, + { 0x00800401, 0x41800236, 0x00d28020, 0x00000000 }, + { 0x00800801, 0x41810236, 0x00d280a0, 0x00000000 }, + { 0x00800401, 0x41a00236, 0x00d28040, 0x00000000 }, + { 0x00800801, 0x41a10236, 0x00d280c0, 0x00000000 }, + { 0x00800401, 0x41c00236, 0x00d28060, 0x00000000 }, + { 0x00800801, 0x41c10236, 0x00d280e0, 0x00000000 }, + { 0x0000040c, 0x20443ca5, 0x00000044, 0x00010001 }, + { 0x00000801, 0x20480061, 0x00000000, 0x0007000f }, + { 0x0a600031, 0x21601c21, 0x508d0040, 0x0a18a001 }, + { 0x00000001, 0x21400021, 0x00000140, 0x00000000 }, + { 0x00000001, 0x21600021, 0x00000160, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00600001, 0x20400021, 0x008d0000, 0x00000000 }, + { 0x00000005, 0x20203e2d, 0x00000061, 0x001f001f }, + { 0x00200009, 0x20643e2d, 0x00450064, 0x00040004 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00000000 }, + { 0x01000010, 0x20003dac, 0x00000020, 0x00160016 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x01200010, 0x20003e2c, 0x0200006c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000024 }, + { 0x00800009, 0x25403dad, 0x00050064, 0x00020002 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00180018 }, + { 0x00200040, 0x45483dad, 0x00660548, 0x00100010 }, + { 0x00200040, 0x45523dad, 0x00660552, 0x00100010 }, + { 0x00400040, 0x25583dad, 0x00690558, 0x00100010 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0x00200020 }, + { 0x00800040, 0x20a035ad, 0x00b100a0, 0x00b10540 }, + { 0x00600040, 0x45423dad, 0x00ae0542, 0x00200020 }, + { 0x00800040, 0x20e035ad, 0x00b100e0, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0xffe0ffe0 }, + { 0x00800040, 0x20c035ad, 0x00b100c0, 0x00b10540 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000014 }, + { 0x06600010, 0x20003dac, 0x02000020, 0x00030003 }, + { 0x00000001, 0x206c0171, 0x00000000, 0x00000000 }, + { 0x00610001, 0x208000a5, 0x02050080, 0x00000000 }, + { 0x00610001, 0x212000a5, 0x02090100, 0x00000000 }, + { 0x00410001, 0x211000a5, 0x02690100, 0x00000000 }, + { 0x00200009, 0x25403dad, 0x00450064, 0x00020002 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00050540 }, + { 0x00200040, 0x40883dad, 0x00660088, 0x00200020 }, + { 0x00200040, 0x40923dad, 0x00660092, 0x00200020 }, + { 0x00400040, 0x20983dad, 0x00690098, 0x00200020 }, + { 0x00200401, 0x22080060, 0x00000000, 0x03400140 }, + { 0x00000c01, 0x220c0060, 0x00000000, 0x04400080 }, + { 0x00000801, 0x22040060, 0x00000000, 0x01000070 }, + { 0x00000005, 0x203e2e29, 0x00000063, 0x00010001 }, + { 0x00000001, 0x20200169, 0x00000000, 0x00000000 }, + { 0x0000000c, 0x2458262d, 0x0000006d, 0x00000020 }, + { 0x00000001, 0x210e0169, 0x00000000, 0x00010001 }, + { 0x00000005, 0x24583dad, 0x00000458, 0x00030003 }, + { 0x01000010, 0x200035ac, 0x0000010e, 0x00000458 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000316 }, + { 0x01000010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00010401, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00010805, 0x245a3e2d, 0x02008800, 0x007f007f }, + { 0x00110401, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00110805, 0x245a3e2d, 0x02008804, 0x007f007f }, + { 0x00010005, 0x245c3e2d, 0x02008800, 0x00800080 }, + { 0x00110005, 0x245c3e2d, 0x02008804, 0x00800080 }, + { 0x00000009, 0x245c3dad, 0x0000045c, 0x00010001 }, + { 0x0000000c, 0x211e362d, 0x0000006c, 0x00000020 }, + { 0x00000040, 0x240035ac, 0x0000003e, 0x0000045c }, + { 0x01000010, 0x20003d8c, 0x00210400, 0x00010001 }, + { 0x01000010, 0x20003d8c, 0x02210400, 0x01000100 }, + { 0x00000001, 0x203c01ed, 0x00000000, 0x00000000 }, + { 0x00010001, 0x203c01ed, 0x00000000, 0x00020002 }, + { 0x00010001, 0x203c01ed, 0x02000000, 0xfffefffe }, + { 0x00000040, 0x20780d21, 0x0000045a, 0x0208e602 }, + { 0x00000040, 0x20782421, 0x00000078, 0x0000045c }, + { 0x02000005, 0x20003dac, 0x0200011e, 0x00030003 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000001c0 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000012 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00700000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c000c }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x0007000c }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00110040, 0x20441da5, 0x02000442, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c0007 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00070007 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0xb80235ad, 0x00009802, 0x0000003c }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x01600006, 0x20003dac, 0x0000044a, 0x00000000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00010040, 0x22000c20, 0x00000078, 0x00200010 }, + { 0x00110040, 0x22000c20, 0x00000078, 0x00300010 }, + { 0x00010001, 0x20480061, 0x00000000, 0x00030009 }, + { 0x00110001, 0x20480061, 0x00000000, 0x00040009 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001a }, + { 0x00600001, 0x26220231, 0x008d05c0, 0x00000000 }, + { 0x00600401, 0x26120231, 0x008d05b8, 0x00000000 }, + { 0x00600801, 0x26020231, 0x008d05b0, 0x00000000 }, + { 0x00600401, 0x25f20231, 0x008d05a8, 0x00000000 }, + { 0x00600801, 0x25e20231, 0x008d05a0, 0x00000000 }, + { 0x00600401, 0x25d20231, 0x008d0598, 0x00000000 }, + { 0x00600801, 0x25c20231, 0x008d0590, 0x00000000 }, + { 0x00600401, 0x25b20231, 0x008d0588, 0x00000000 }, + { 0x00600801, 0x25a20231, 0x008d0580, 0x00000000 }, + { 0x00600401, 0x25920231, 0x008d0578, 0x00000000 }, + { 0x00600801, 0x25820231, 0x008d0570, 0x00000000 }, + { 0x00600001, 0x25720231, 0x008d0568, 0x00000000 }, + { 0x00600001, 0x25620231, 0x008d0560, 0x00000000 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00810001, 0xbc000229, 0x02ad8000, 0x00000000 }, + { 0x00810001, 0xbc200229, 0x02ad8020, 0x00000000 }, + { 0x00810001, 0xbc400229, 0x02ad8040, 0x00000000 }, + { 0x00810001, 0xbc600229, 0x02ad8060, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000012a }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000066 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000064 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04a00480 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00a02040, 0x2400462c, 0x00ad8000, 0x00ad8005 }, + { 0x00a02048, 0x24003e2c, 0x00ad8001, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8002, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8003, 0x00140014 }, + { 0x00a02048, 0xb8003e2d, 0x00ad8004, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8040, 0x00ad8045 }, + { 0x00a02048, 0x24003e2c, 0x00ad8041, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8042, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8043, 0x00140014 }, + { 0x00a02048, 0xb8403e2d, 0x00ad8044, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8080, 0x00ad8085 }, + { 0x00a02048, 0x24003e2c, 0x00ad8081, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8082, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8083, 0x00140014 }, + { 0x00a02048, 0xb8803e2d, 0x00ad8084, 0xfffbfffb }, + { 0x00600040, 0x2400462c, 0x008d80c0, 0x008d80c5 }, + { 0x00600048, 0x24003e2c, 0x008d80c1, 0xfffbfffb }, + { 0x00600048, 0x24003e2c, 0x008d80c2, 0x00140014 }, + { 0x00600048, 0x24003e2c, 0x008d80c3, 0x00140014 }, + { 0x00600048, 0xb8c03e2d, 0x008d80c4, 0xfffbfffb }, + { 0x00000401, 0x22000060, 0x00000000, 0x04a00480 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04a00490 }, + { 0x00010802, 0x220c2d28, 0x00000454, 0x04800480 }, + { 0x00a02040, 0x24003dac, 0x00b18000, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8800, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8820, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18020, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8820, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8840, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18040, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x000a000a }, + { 0x00a02040, 0x24003dac, 0x00b18040, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18060, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8860, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8880, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18080, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8880, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed88a0, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x000000b4 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000004c }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000004a }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010002, 0x220c2d28, 0x02000454, 0x04800480 }, + { 0x00800040, 0x24003e2c, 0x01ed8000, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8001, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8021, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8002, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8022, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8003, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8023, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8004, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8024, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8005, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8025, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00800040, 0x24003e2c, 0x01ed8040, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8060, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8041, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8061, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8042, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8062, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8043, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8063, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8044, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8064, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8045, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8065, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000005c }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000062 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000042 }, + { 0x00200401, 0x22000128, 0x0066044c, 0x00000000 }, + { 0x00200c01, 0x22040128, 0x0045044e, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x03400010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220c0168, 0x02000000, 0x04800480 }, + { 0x00a02040, 0x24003e2c, 0x00ad83e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8be0, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8800, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8800, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8820, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00a02040, 0x24003e2c, 0x00ad8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8040, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8840, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8860, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8060, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8860, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8880, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000001c }, + { 0x00200401, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x01200010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01200010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00210040, 0x22002d08, 0x00450200, 0x00010001 }, + { 0x00210040, 0x22002d08, 0x02450200, 0x00100010 }, + { 0x00800001, 0xb8000229, 0x01ed8000, 0x00000000 }, + { 0x00800001, 0xb8200229, 0x01ed8020, 0x00000000 }, + { 0x00800001, 0xb8400229, 0x01ed8040, 0x00000000 }, + { 0x00800001, 0xb8600229, 0x01ed8060, 0x00000000 }, + { 0x80800042, 0xd8004631, 0x00d29800, 0x00d20480 }, + { 0x80800042, 0xd8204631, 0x00d29820, 0x00d204a0 }, + { 0x80800042, 0xd8404631, 0x00d29840, 0x00d204c0 }, + { 0x80800042, 0xd8604631, 0x00d29860, 0x00d204e0 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000c01, 0x220201e8, 0x00000000, 0x07100710 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00200040, 0x244c3d09, 0x00450200, 0x00100010 }, + { 0x00800041, 0x24002628, 0x00ad8000, 0x00000540 }, + { 0x00800041, 0x24202628, 0x00ad8020, 0x00000540 }, + { 0x00800048, 0x24002628, 0x00ad8002, 0x00000542 }, + { 0x00800048, 0x24202628, 0x00ad8022, 0x00000542 }, + { 0x00200001, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00800048, 0x24002628, 0x01ed8000, 0x00000544 }, + { 0x00800048, 0x24202628, 0x01ed8020, 0x00000544 }, + { 0x00800048, 0xbc002629, 0x01ed8002, 0x00000546 }, + { 0x00801048, 0xbc202629, 0x01ed8022, 0x00000546 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000128 }, + { 0x00000001, 0x20220169, 0x00000000, 0x00040004 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00500000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x00080008 }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00030008 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0xb80235ad, 0x00009802, 0x0000003c }, + { 0x00000040, 0x22000c20, 0x00000078, 0x00100010 }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00000001, 0x20480061, 0x00000000, 0x00020005 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00410001, 0xbc000229, 0x02698000, 0x00000000 }, + { 0x00410001, 0xbc100229, 0x02698010, 0x00000000 }, + { 0x00410001, 0xbc200229, 0x02698020, 0x00000000 }, + { 0x00410001, 0xbc300229, 0x02698030, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000000be }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000044 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000042 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04d004c0 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00802040, 0x2400462c, 0x00a98000, 0x00a98005 }, + { 0x00802048, 0x24003e2c, 0x00a98001, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98002, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98003, 0x00140014 }, + { 0x00802048, 0xb8003e2d, 0x00a98004, 0xfffbfffb }, + { 0x00802040, 0x2400462c, 0x00a98040, 0x00a98045 }, + { 0x00802048, 0x24003e2c, 0x00a98041, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98042, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98043, 0x00140014 }, + { 0x00802048, 0xb8203e2d, 0x00a98044, 0xfffbfffb }, + { 0x00400040, 0x2400462c, 0x00698080, 0x00698085 }, + { 0x00400048, 0x24003e2c, 0x00698081, 0xfffbfffb }, + { 0x00400048, 0x24003e2c, 0x00698082, 0x00140014 }, + { 0x00400048, 0x24003e2c, 0x00698083, 0x00140014 }, + { 0x00400048, 0xb8403e2d, 0x00698084, 0xfffbfffb }, + { 0x00000401, 0x220c0060, 0x00000000, 0x04e004d0 }, + { 0x00000c01, 0x22000060, 0x00000000, 0x04d004c8 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04e004d8 }, + { 0x00000801, 0x22080060, 0x00000000, 0x04f004e8 }, + { 0x00800040, 0x24003dac, 0x00b104c0, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x00b104e0, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed9800, 0x00140014 }, + { 0x00010001, 0x220c0168, 0x00000000, 0x05400540 }, + { 0x00110001, 0x220c0168, 0x00000000, 0x04c004c0 }, + { 0x00800048, 0x24003dac, 0x01e98000, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01e98020, 0x00010001 }, + { 0x00800048, 0x24003dac, 0x01e98800, 0x00140014 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000060 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000022 }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010001, 0x220e0168, 0x02000000, 0x05400540 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x00800040, 0x24003e2c, 0x01e98000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98001, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98002, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98003, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98004, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98005, 0x00010001 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x03600010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x00400401, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x00400840, 0x22083d2c, 0x0069044c, 0x00100010 }, + { 0x00610040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00800040, 0x24003e2c, 0x01e983e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e993e0, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e99000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e99020, 0x00010001 }, + { 0x00000001, 0x220e0168, 0x00000000, 0x05400540 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x00400001, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x01400010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01400010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00010001 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00100010 }, + { 0x00800001, 0x25400229, 0x01e98000, 0x00000000 }, + { 0x80800042, 0x45404631, 0x00d20540, 0x00d204c0 }, + { 0x00000001, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x00400001, 0xd8000231, 0x008a0540, 0x00000000 }, + { 0x00400001, 0xd8100231, 0x008a0548, 0x00000000 }, + { 0x00400001, 0xd8200231, 0x008a0550, 0x00000000 }, + { 0x00400001, 0xd8300231, 0x008a0558, 0x00000000 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00600041, 0x24002628, 0x00898000, 0x00000540 }, + { 0x00600048, 0x24002628, 0x00898002, 0x00000542 }, + { 0x00600048, 0x24002628, 0x00898008, 0x00000544 }, + { 0x00600048, 0x2540262d, 0x0089800a, 0x00000546 }, + { 0x00400401, 0xbc0001a9, 0x00690540, 0x00000000 }, + { 0x00400801, 0xbc1001a9, 0x00690548, 0x00000000 }, + { 0x01000010, 0x20003d2c, 0x00000022, 0x00030003 }, + { 0x01000040, 0x20223d29, 0x02000022, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00080008 }, + { 0x00110040, 0x24540c21, 0x00000454, 0x00080008 }, + { 0x00010040, 0x24540c21, 0x00000454, 0x00180038 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffee4 }, + { 0x01800010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0xffe0ffe0 }, + { 0x00010001, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00110001, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00000001, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00800040, 0x24003dac, 0x00b19c00, 0x00200020 }, + { 0x00800040, 0x24203dac, 0x00b19c20, 0x00200020 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00060006 }, + { 0x8080000c, 0xdc203d91, 0x00b10420, 0x00060006 }, + { 0x02000040, 0x210e3d29, 0x0200010e, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00040004 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffce0 }, + { 0x01000005, 0x25643e2d, 0x00000060, 0x00c000c0 }, + { 0x01000010, 0x20003dac, 0x02000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000086 }, + { 0x00a02001, 0x46400231, 0x00d20780, 0x00000000 }, + { 0x00a02001, 0x46800231, 0x00d207c0, 0x00000000 }, + { 0x00a02001, 0x46c00231, 0x00d20480, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000007e }, + { 0x80a02042, 0x46404631, 0x00d20640, 0x00d20780 }, + { 0x80a02042, 0x46804631, 0x00d20680, 0x00d207c0 }, + { 0x80a02042, 0x46c04631, 0x00d206c0, 0x00d20480 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000076 }, + { 0x01000010, 0x20003dac, 0x02000564, 0x00800080 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffffe4 }, + { 0x00200401, 0x256801ed, 0x00000000, 0x00200020 }, + { 0x00200801, 0x256c01ed, 0x00000000, 0x00060006 }, + { 0x00400001, 0x258001ed, 0x00000000, 0x00000000 }, + { 0x00600001, 0x45a001ad, 0x00058c00, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000028 }, + { 0x02600005, 0x20006e28, 0x02008c0c, 0x88848421 }, + { 0x0220000c, 0x25663dad, 0x00000458, 0x00010001 }, + { 0x0000000c, 0x25643dad, 0x00000564, 0x00060006 }, + { 0x00310001, 0x25c401ad, 0x00000458, 0x00000000 }, + { 0x00210001, 0x25c00061, 0x00000000, 0x00010001 }, + { 0x00310040, 0x25c03dad, 0x004545c4, 0x00010001 }, + { 0x00610001, 0x25a00061, 0x02000000, 0x00000080 }, + { 0x00710001, 0x45a002ad, 0x02ae8c00, 0x00000000 }, + { 0x00710001, 0x45a202ad, 0x02ae8c01, 0x00000000 }, + { 0x00800041, 0x25a035ad, 0x00b105a0, 0x000905c0 }, + { 0x01200010, 0x20003e2c, 0x0245006e, 0x00000000 }, + { 0x00310009, 0x244c45ad, 0x02000564, 0x0045006e }, + { 0x00210001, 0x244c01ed, 0x02000000, 0x00000000 }, + { 0x0031000c, 0x244c3dad, 0x0245044c, 0x00010001 }, + { 0x00200009, 0x256835ad, 0x0045044c, 0x00000566 }, + { 0x00210040, 0x256835ad, 0x02450568, 0x00000566 }, + { 0x00200040, 0x256c362d, 0x0045006e, 0x00000566 }, + { 0x00400040, 0x240035ac, 0x00ab05a2, 0x00ab05a6 }, + { 0x00400040, 0x2400358c, 0x00690400, 0x00000566 }, + { 0x0040000c, 0x2580358d, 0x00690400, 0x00000566 }, + { 0x00800041, 0x2400362c, 0x00d20640, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d20660, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d20780, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207a0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x464035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x466035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d20680, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d206a0, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d207c0, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207e0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x468035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x46a035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d206c0, 0x000705a8 }, + { 0x00800041, 0x2420362c, 0x00d206e0, 0x000705a8 }, + { 0x00800048, 0x2400362c, 0x00d20480, 0x000705ac }, + { 0x00800048, 0x2420362c, 0x00d204a0, 0x000705ac }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x0000056a }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x0000056a }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056e }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056e }, + { 0x80800040, 0x46c035b1, 0x00b104c0, 0x00050582 }, + { 0x80800040, 0x46e035b1, 0x00b104e0, 0x00050582 }, + { 0x01800005, 0x20003d2c, 0x02000020, 0x00020002 }, + { 0x80800040, 0xd00045b1, 0x00b19000, 0x00d20640 }, + { 0x80800040, 0xd02045b1, 0x00b19020, 0x00d20660 }, + { 0x80800040, 0xd04045b1, 0x00b19040, 0x00d20680 }, + { 0x80800040, 0xd06045b1, 0x00b19060, 0x00d206a0 }, + { 0x00000040, 0x22083d8c, 0x00000208, 0x00800080 }, + { 0x80400040, 0xd40045b1, 0x00699400, 0x00ab06c0 }, + { 0x80400040, 0xd48045b1, 0x00699480, 0x00ab06c2 }, + { 0x80400040, 0xd42045b1, 0x00699420, 0x00ab06e0 }, + { 0x80400040, 0xd4a045b1, 0x006994a0, 0x00ab06e2 }, + { 0x80400040, 0xd41045b1, 0x00699410, 0x00ab06d0 }, + { 0x80400040, 0xd49045b1, 0x00699490, 0x00ab06d2 }, + { 0x80400040, 0xd43045b1, 0x00699430, 0x00ab06f0 }, + { 0x80400040, 0xd4b045b1, 0x006994b0, 0x00ab06f2 }, + { 0x00000040, 0x220a3d8c, 0x0000020a, 0x00080008 }, + { 0x00110040, 0x220a3d8c, 0x0200020a, 0x00300030 }, + { 0x01000010, 0x20003d2c, 0x02000020, 0x00060006 }, + { 0x00000040, 0x20203d29, 0x00000020, 0x00020002 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00100001 }, + { 0x00000040, 0x220c358c, 0x0000020c, 0x0000003a }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffc1e }, + { 0x00000001, 0x220001ec, 0x00000000, 0x01400140 }, + { 0x00000001, 0x220201ec, 0x00000000, 0x01c001c0 }, + { 0x00800401, 0x20400236, 0x01ee8000, 0x00000000 }, + { 0x00800801, 0x20500236, 0x01ee8010, 0x00000000 }, + { 0x00800401, 0x20600236, 0x01ee8020, 0x00000000 }, + { 0x00800801, 0x20700236, 0x01ee8030, 0x00000000 }, + { 0x00800401, 0x20800236, 0x01ee8040, 0x00000000 }, + { 0x00800801, 0x20900236, 0x01ee8050, 0x00000000 }, + { 0x00800401, 0x20a00236, 0x01ee8060, 0x00000000 }, + { 0x00800801, 0x20b00236, 0x01ee8070, 0x00000000 }, + { 0x00800401, 0x20c00236, 0x01ee8100, 0x00000000 }, + { 0x00800801, 0x20d00236, 0x01ee8110, 0x00000000 }, + { 0x00800401, 0x20e00236, 0x01ee8120, 0x00000000 }, + { 0x00800801, 0x20f00236, 0x01ee8130, 0x00000000 }, + { 0x00800401, 0x21000236, 0x01ee8140, 0x00000000 }, + { 0x00800801, 0x21100236, 0x01ee8150, 0x00000000 }, + { 0x00800401, 0x21200236, 0x01ee8160, 0x00000000 }, + { 0x00800801, 0x21300236, 0x01ee8170, 0x00000000 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x1218a600 }, + { 0x00200401, 0x204001a5, 0x00450064, 0x00000000 }, + { 0x00000801, 0x20480061, 0x00000000, 0x000f000f }, + { 0x01600031, 0x21400021, 0x508d0040, 0x00000200 }, + { 0x00000001, 0x220001ec, 0x00000000, 0x03400340 }, + { 0x00800401, 0x41600236, 0x00d28000, 0x00000000 }, + { 0x00800801, 0x41610236, 0x00d28080, 0x00000000 }, + { 0x00800401, 0x41800236, 0x00d28020, 0x00000000 }, + { 0x00800801, 0x41810236, 0x00d280a0, 0x00000000 }, + { 0x00800401, 0x41a00236, 0x00d28040, 0x00000000 }, + { 0x00800801, 0x41a10236, 0x00d280c0, 0x00000000 }, + { 0x00800401, 0x41c00236, 0x00d28060, 0x00000000 }, + { 0x00800801, 0x41c10236, 0x00d280e0, 0x00000000 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x0a18a601 }, + { 0x0000040c, 0x20443ca5, 0x00000044, 0x00010001 }, + { 0x00000801, 0x20480061, 0x00000000, 0x0007000f }, + { 0x0a600031, 0x21600021, 0x508d0040, 0x00000200 }, + { 0x00000001, 0x21400021, 0x00000140, 0x00000000 }, + { 0x00000001, 0x21600021, 0x00000160, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00600001, 0x20400021, 0x008d0000, 0x00000000 }, + { 0x00000005, 0x20203e2d, 0x00000061, 0x001f001f }, + { 0x00200009, 0x20643e2d, 0x00450064, 0x00040004 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00000000 }, + { 0x01000010, 0x20003dac, 0x00000020, 0x00160016 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x01200010, 0x20003e2c, 0x0200006c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000024 }, + { 0x00800009, 0x25403dad, 0x00050064, 0x00020002 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00180018 }, + { 0x00200040, 0x45483dad, 0x00660548, 0x00100010 }, + { 0x00200040, 0x45523dad, 0x00660552, 0x00100010 }, + { 0x00400040, 0x25583dad, 0x00690558, 0x00100010 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0x00200020 }, + { 0x00800040, 0x20a035ad, 0x00b100a0, 0x00b10540 }, + { 0x00600040, 0x45423dad, 0x00ae0542, 0x00200020 }, + { 0x00800040, 0x20e035ad, 0x00b100e0, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0xffe0ffe0 }, + { 0x00800040, 0x20c035ad, 0x00b100c0, 0x00b10540 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000014 }, + { 0x06600010, 0x20003dac, 0x02000020, 0x00030003 }, + { 0x00000001, 0x206c0171, 0x00000000, 0x00000000 }, + { 0x00610001, 0x208000a5, 0x02050080, 0x00000000 }, + { 0x00610001, 0x212000a5, 0x02090100, 0x00000000 }, + { 0x00410001, 0x211000a5, 0x02690100, 0x00000000 }, + { 0x00200009, 0x25403dad, 0x00450064, 0x00020002 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00050540 }, + { 0x00200040, 0x40883dad, 0x00660088, 0x00200020 }, + { 0x00200040, 0x40923dad, 0x00660092, 0x00200020 }, + { 0x00400040, 0x20983dad, 0x00690098, 0x00200020 }, + { 0x00200401, 0x22080060, 0x00000000, 0x03400140 }, + { 0x00000c01, 0x220c0060, 0x00000000, 0x04400080 }, + { 0x00000801, 0x22040060, 0x00000000, 0x01000070 }, + { 0x02000005, 0x20002e28, 0x00000061, 0x00400040 }, + { 0x00010005, 0x203e2e29, 0x00000063, 0x00010001 }, + { 0x00110001, 0x203e0169, 0x00000000, 0x00030003 }, + { 0x00000001, 0x20200169, 0x00000000, 0x00000000 }, + { 0x0000000c, 0x2458262d, 0x0000006d, 0x00000020 }, + { 0x00000001, 0x210e0169, 0x00000000, 0x00010001 }, + { 0x00000005, 0x24583dad, 0x00000458, 0x00030003 }, + { 0x01000010, 0x200035ac, 0x0000010e, 0x00000458 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000031a }, + { 0x01000010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00010401, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00010805, 0x245a3e2d, 0x02008800, 0x007f007f }, + { 0x00110401, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00110805, 0x245a3e2d, 0x02008804, 0x007f007f }, + { 0x00010005, 0x245c3e2d, 0x02008800, 0x00800080 }, + { 0x00110005, 0x245c3e2d, 0x02008804, 0x00800080 }, + { 0x00000009, 0x245c3dad, 0x0000045c, 0x00010001 }, + { 0x0000000c, 0x211e362d, 0x0000006c, 0x00000020 }, + { 0x00000040, 0x240035ac, 0x0000003e, 0x0000045c }, + { 0x01000010, 0x20003d8c, 0x00210400, 0x00010001 }, + { 0x01000010, 0x20003d8c, 0x02210400, 0x01000100 }, + { 0x00000001, 0x203c01ed, 0x00000000, 0x00000000 }, + { 0x00010001, 0x203c01ed, 0x00000000, 0x00020002 }, + { 0x00010001, 0x203c01ed, 0x02000000, 0xfffefffe }, + { 0x02000005, 0x20002e28, 0x00000061, 0x00400040 }, + { 0x00010040, 0x244c0d21, 0x0000045a, 0x0208e602 }, + { 0x00110040, 0x20780d21, 0x0000045a, 0x0208a002 }, + { 0x00010040, 0x20782421, 0x0000044c, 0x0000045c }, + { 0x02000005, 0x20003dac, 0x0200011e, 0x00030003 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000001c0 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000012 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00700000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c000c }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x0007000c }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00110040, 0x20441da5, 0x02000442, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c0007 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00070007 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0xb80235ad, 0x00009802, 0x0000003c }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x01600006, 0x20003dac, 0x0000044a, 0x00000000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00010040, 0x22000c20, 0x00000078, 0x00200010 }, + { 0x00110040, 0x22000c20, 0x00000078, 0x00300010 }, + { 0x00010001, 0x20480061, 0x00000000, 0x00030009 }, + { 0x00110001, 0x20480061, 0x00000000, 0x00040009 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001a }, + { 0x00600001, 0x26220231, 0x008d05c0, 0x00000000 }, + { 0x00600401, 0x26120231, 0x008d05b8, 0x00000000 }, + { 0x00600801, 0x26020231, 0x008d05b0, 0x00000000 }, + { 0x00600401, 0x25f20231, 0x008d05a8, 0x00000000 }, + { 0x00600801, 0x25e20231, 0x008d05a0, 0x00000000 }, + { 0x00600401, 0x25d20231, 0x008d0598, 0x00000000 }, + { 0x00600801, 0x25c20231, 0x008d0590, 0x00000000 }, + { 0x00600401, 0x25b20231, 0x008d0588, 0x00000000 }, + { 0x00600801, 0x25a20231, 0x008d0580, 0x00000000 }, + { 0x00600401, 0x25920231, 0x008d0578, 0x00000000 }, + { 0x00600801, 0x25820231, 0x008d0570, 0x00000000 }, + { 0x00600001, 0x25720231, 0x008d0568, 0x00000000 }, + { 0x00600001, 0x25620231, 0x008d0560, 0x00000000 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00810001, 0xbc000229, 0x02ad8000, 0x00000000 }, + { 0x00810001, 0xbc200229, 0x02ad8020, 0x00000000 }, + { 0x00810001, 0xbc400229, 0x02ad8040, 0x00000000 }, + { 0x00810001, 0xbc600229, 0x02ad8060, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000012a }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000066 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000064 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04a00480 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00a02040, 0x2400462c, 0x00ad8000, 0x00ad8005 }, + { 0x00a02048, 0x24003e2c, 0x00ad8001, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8002, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8003, 0x00140014 }, + { 0x00a02048, 0xb8003e2d, 0x00ad8004, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8040, 0x00ad8045 }, + { 0x00a02048, 0x24003e2c, 0x00ad8041, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8042, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8043, 0x00140014 }, + { 0x00a02048, 0xb8403e2d, 0x00ad8044, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8080, 0x00ad8085 }, + { 0x00a02048, 0x24003e2c, 0x00ad8081, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8082, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8083, 0x00140014 }, + { 0x00a02048, 0xb8803e2d, 0x00ad8084, 0xfffbfffb }, + { 0x00600040, 0x2400462c, 0x008d80c0, 0x008d80c5 }, + { 0x00600048, 0x24003e2c, 0x008d80c1, 0xfffbfffb }, + { 0x00600048, 0x24003e2c, 0x008d80c2, 0x00140014 }, + { 0x00600048, 0x24003e2c, 0x008d80c3, 0x00140014 }, + { 0x00600048, 0xb8c03e2d, 0x008d80c4, 0xfffbfffb }, + { 0x00000401, 0x22000060, 0x00000000, 0x04a00480 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04a00490 }, + { 0x00010802, 0x220c2d28, 0x00000454, 0x04800480 }, + { 0x00a02040, 0x24003dac, 0x00b18000, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8800, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8820, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18020, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8820, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8840, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18040, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x000a000a }, + { 0x00a02040, 0x24003dac, 0x00b18040, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18060, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8860, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8880, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18080, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8880, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed88a0, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x000000b4 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000004c }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000004a }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010002, 0x220c2d28, 0x02000454, 0x04800480 }, + { 0x00800040, 0x24003e2c, 0x01ed8000, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8001, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8021, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8002, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8022, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8003, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8023, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8004, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8024, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8005, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8025, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00800040, 0x24003e2c, 0x01ed8040, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8060, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8041, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8061, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8042, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8062, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8043, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8063, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8044, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8064, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8045, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8065, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000005c }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000062 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000042 }, + { 0x00200401, 0x22000128, 0x0066044c, 0x00000000 }, + { 0x00200c01, 0x22040128, 0x0045044e, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x03400010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220c0168, 0x02000000, 0x04800480 }, + { 0x00a02040, 0x24003e2c, 0x00ad83e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8be0, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8800, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8800, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8820, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00a02040, 0x24003e2c, 0x00ad8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8040, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8840, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8860, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8060, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8860, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8880, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000001c }, + { 0x00200401, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x01200010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01200010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00210040, 0x22002d08, 0x00450200, 0x00010001 }, + { 0x00210040, 0x22002d08, 0x02450200, 0x00100010 }, + { 0x00800001, 0xb8000229, 0x01ed8000, 0x00000000 }, + { 0x00800001, 0xb8200229, 0x01ed8020, 0x00000000 }, + { 0x00800001, 0xb8400229, 0x01ed8040, 0x00000000 }, + { 0x00800001, 0xb8600229, 0x01ed8060, 0x00000000 }, + { 0x80800042, 0xd8004631, 0x00d29800, 0x00d20480 }, + { 0x80800042, 0xd8204631, 0x00d29820, 0x00d204a0 }, + { 0x80800042, 0xd8404631, 0x00d29840, 0x00d204c0 }, + { 0x80800042, 0xd8604631, 0x00d29860, 0x00d204e0 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000c01, 0x220201e8, 0x00000000, 0x07100710 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00200040, 0x244c3d09, 0x00450200, 0x00100010 }, + { 0x00800041, 0x24002628, 0x00ad8000, 0x00000540 }, + { 0x00800041, 0x24202628, 0x00ad8020, 0x00000540 }, + { 0x00800048, 0x24002628, 0x00ad8002, 0x00000542 }, + { 0x00800048, 0x24202628, 0x00ad8022, 0x00000542 }, + { 0x00200001, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00800048, 0x24002628, 0x01ed8000, 0x00000544 }, + { 0x00800048, 0x24202628, 0x01ed8020, 0x00000544 }, + { 0x00800048, 0xbc002629, 0x01ed8002, 0x00000546 }, + { 0x00801048, 0xbc202629, 0x01ed8022, 0x00000546 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000128 }, + { 0x00000001, 0x20220169, 0x00000000, 0x00040004 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00500000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x00080008 }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00030008 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0xb80235ad, 0x00009802, 0x0000003c }, + { 0x00000040, 0x22000c20, 0x00000078, 0x00100010 }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00000001, 0x20480061, 0x00000000, 0x00020005 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00410001, 0xbc000229, 0x02698000, 0x00000000 }, + { 0x00410001, 0xbc100229, 0x02698010, 0x00000000 }, + { 0x00410001, 0xbc200229, 0x02698020, 0x00000000 }, + { 0x00410001, 0xbc300229, 0x02698030, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000000be }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000044 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000042 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04d004c0 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00802040, 0x2400462c, 0x00a98000, 0x00a98005 }, + { 0x00802048, 0x24003e2c, 0x00a98001, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98002, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98003, 0x00140014 }, + { 0x00802048, 0xb8003e2d, 0x00a98004, 0xfffbfffb }, + { 0x00802040, 0x2400462c, 0x00a98040, 0x00a98045 }, + { 0x00802048, 0x24003e2c, 0x00a98041, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98042, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98043, 0x00140014 }, + { 0x00802048, 0xb8203e2d, 0x00a98044, 0xfffbfffb }, + { 0x00400040, 0x2400462c, 0x00698080, 0x00698085 }, + { 0x00400048, 0x24003e2c, 0x00698081, 0xfffbfffb }, + { 0x00400048, 0x24003e2c, 0x00698082, 0x00140014 }, + { 0x00400048, 0x24003e2c, 0x00698083, 0x00140014 }, + { 0x00400048, 0xb8403e2d, 0x00698084, 0xfffbfffb }, + { 0x00000401, 0x220c0060, 0x00000000, 0x04e004d0 }, + { 0x00000c01, 0x22000060, 0x00000000, 0x04d004c8 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04e004d8 }, + { 0x00000801, 0x22080060, 0x00000000, 0x04f004e8 }, + { 0x00800040, 0x24003dac, 0x00b104c0, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x00b104e0, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed9800, 0x00140014 }, + { 0x00010001, 0x220c0168, 0x00000000, 0x05400540 }, + { 0x00110001, 0x220c0168, 0x00000000, 0x04c004c0 }, + { 0x00800048, 0x24003dac, 0x01e98000, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01e98020, 0x00010001 }, + { 0x00800048, 0x24003dac, 0x01e98800, 0x00140014 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000060 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000022 }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010001, 0x220e0168, 0x02000000, 0x05400540 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x00800040, 0x24003e2c, 0x01e98000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98001, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98002, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98003, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98004, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98005, 0x00010001 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x03600010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x00400401, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x00400840, 0x22083d2c, 0x0069044c, 0x00100010 }, + { 0x00610040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00800040, 0x24003e2c, 0x01e983e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e993e0, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e99000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e99020, 0x00010001 }, + { 0x00000001, 0x220e0168, 0x00000000, 0x05400540 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x00400001, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x01400010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01400010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00010001 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00100010 }, + { 0x00800001, 0x25400229, 0x01e98000, 0x00000000 }, + { 0x80800042, 0x45404631, 0x00d20540, 0x00d204c0 }, + { 0x00000001, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x00400001, 0xd8000231, 0x008a0540, 0x00000000 }, + { 0x00400001, 0xd8100231, 0x008a0548, 0x00000000 }, + { 0x00400001, 0xd8200231, 0x008a0550, 0x00000000 }, + { 0x00400001, 0xd8300231, 0x008a0558, 0x00000000 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00600041, 0x24002628, 0x00898000, 0x00000540 }, + { 0x00600048, 0x24002628, 0x00898002, 0x00000542 }, + { 0x00600048, 0x24002628, 0x00898008, 0x00000544 }, + { 0x00600048, 0x2540262d, 0x0089800a, 0x00000546 }, + { 0x00400401, 0xbc0001a9, 0x00690540, 0x00000000 }, + { 0x00400801, 0xbc1001a9, 0x00690548, 0x00000000 }, + { 0x01000010, 0x20003d2c, 0x00000022, 0x00030003 }, + { 0x01000040, 0x20223d29, 0x02000022, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00080008 }, + { 0x00110040, 0x24540c21, 0x00000454, 0x00080008 }, + { 0x00010040, 0x24540c21, 0x00000454, 0x00180038 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffee4 }, + { 0x01800010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0xffe0ffe0 }, + { 0x00010001, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00110001, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00000001, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00800040, 0x24003dac, 0x00b19c00, 0x00200020 }, + { 0x00800040, 0x24203dac, 0x00b19c20, 0x00200020 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00060006 }, + { 0x8080000c, 0xdc203d91, 0x00b10420, 0x00060006 }, + { 0x02000040, 0x210e3d29, 0x0200010e, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00040004 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffcdc }, + { 0x01000005, 0x25643e2d, 0x00000060, 0x00c000c0 }, + { 0x01000010, 0x20003dac, 0x02000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000086 }, + { 0x00a02001, 0x46400231, 0x00d20780, 0x00000000 }, + { 0x00a02001, 0x46800231, 0x00d207c0, 0x00000000 }, + { 0x00a02001, 0x46c00231, 0x00d20480, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000007e }, + { 0x80a02042, 0x46404631, 0x00d20640, 0x00d20780 }, + { 0x80a02042, 0x46804631, 0x00d20680, 0x00d207c0 }, + { 0x80a02042, 0x46c04631, 0x00d206c0, 0x00d20480 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000076 }, + { 0x01000010, 0x20003dac, 0x02000564, 0x00800080 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffffe4 }, + { 0x00200401, 0x256801ed, 0x00000000, 0x00200020 }, + { 0x00200801, 0x256c01ed, 0x00000000, 0x00060006 }, + { 0x00400001, 0x258001ed, 0x00000000, 0x00000000 }, + { 0x00600001, 0x45a001ad, 0x00058c00, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000028 }, + { 0x02600005, 0x20006e28, 0x02008c0c, 0x88848421 }, + { 0x0220000c, 0x25663dad, 0x00000458, 0x00010001 }, + { 0x0000000c, 0x25643dad, 0x00000564, 0x00060006 }, + { 0x00310001, 0x25c401ad, 0x00000458, 0x00000000 }, + { 0x00210001, 0x25c00061, 0x00000000, 0x00010001 }, + { 0x00310040, 0x25c03dad, 0x004545c4, 0x00010001 }, + { 0x00610001, 0x25a00061, 0x02000000, 0x00000080 }, + { 0x00710001, 0x45a002ad, 0x02ae8c00, 0x00000000 }, + { 0x00710001, 0x45a202ad, 0x02ae8c01, 0x00000000 }, + { 0x00800041, 0x25a035ad, 0x00b105a0, 0x000905c0 }, + { 0x01200010, 0x20003e2c, 0x0245006e, 0x00000000 }, + { 0x00310009, 0x244c45ad, 0x02000564, 0x0045006e }, + { 0x00210001, 0x244c01ed, 0x02000000, 0x00000000 }, + { 0x0031000c, 0x244c3dad, 0x0245044c, 0x00010001 }, + { 0x00200009, 0x256835ad, 0x0045044c, 0x00000566 }, + { 0x00210040, 0x256835ad, 0x02450568, 0x00000566 }, + { 0x00200040, 0x256c362d, 0x0045006e, 0x00000566 }, + { 0x00400040, 0x240035ac, 0x00ab05a2, 0x00ab05a6 }, + { 0x00400040, 0x2400358c, 0x00690400, 0x00000566 }, + { 0x0040000c, 0x2580358d, 0x00690400, 0x00000566 }, + { 0x00800041, 0x2400362c, 0x00d20640, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d20660, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d20780, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207a0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x464035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x466035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d20680, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d206a0, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d207c0, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207e0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x468035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x46a035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d206c0, 0x000705a8 }, + { 0x00800041, 0x2420362c, 0x00d206e0, 0x000705a8 }, + { 0x00800048, 0x2400362c, 0x00d20480, 0x000705ac }, + { 0x00800048, 0x2420362c, 0x00d204a0, 0x000705ac }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x0000056a }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x0000056a }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056e }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056e }, + { 0x80800040, 0x46c035b1, 0x00b104c0, 0x00050582 }, + { 0x80800040, 0x46e035b1, 0x00b104e0, 0x00050582 }, + { 0x01800005, 0x20003d2c, 0x02000020, 0x00020002 }, + { 0x80800040, 0xd00045b1, 0x00b19000, 0x00d20640 }, + { 0x80800040, 0xd02045b1, 0x00b19020, 0x00d20660 }, + { 0x80800040, 0xd04045b1, 0x00b19040, 0x00d20680 }, + { 0x80800040, 0xd06045b1, 0x00b19060, 0x00d206a0 }, + { 0x00000040, 0x22083d8c, 0x00000208, 0x00800080 }, + { 0x80400040, 0xd40045b1, 0x00699400, 0x00ab06c0 }, + { 0x80400040, 0xd48045b1, 0x00699480, 0x00ab06c2 }, + { 0x80400040, 0xd42045b1, 0x00699420, 0x00ab06e0 }, + { 0x80400040, 0xd4a045b1, 0x006994a0, 0x00ab06e2 }, + { 0x80400040, 0xd41045b1, 0x00699410, 0x00ab06d0 }, + { 0x80400040, 0xd49045b1, 0x00699490, 0x00ab06d2 }, + { 0x80400040, 0xd43045b1, 0x00699430, 0x00ab06f0 }, + { 0x80400040, 0xd4b045b1, 0x006994b0, 0x00ab06f2 }, + { 0x00000040, 0x220a3d8c, 0x0000020a, 0x00080008 }, + { 0x00110040, 0x220a3d8c, 0x0200020a, 0x00300030 }, + { 0x01000010, 0x20003d2c, 0x02000020, 0x00060006 }, + { 0x00000040, 0x20203d29, 0x00000020, 0x00020002 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00100001 }, + { 0x00000040, 0x220c358c, 0x0000020c, 0x0000003a }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffc1a }, + { 0x00000001, 0x220001ec, 0x00000000, 0x01400140 }, + { 0x00000001, 0x220201ec, 0x00000000, 0x01c001c0 }, + { 0x00800401, 0x20400236, 0x01ee8000, 0x00000000 }, + { 0x00800801, 0x20500236, 0x01ee8010, 0x00000000 }, + { 0x00800401, 0x20600236, 0x01ee8020, 0x00000000 }, + { 0x00800801, 0x20700236, 0x01ee8030, 0x00000000 }, + { 0x00800401, 0x20800236, 0x01ee8040, 0x00000000 }, + { 0x00800801, 0x20900236, 0x01ee8050, 0x00000000 }, + { 0x00800401, 0x20a00236, 0x01ee8060, 0x00000000 }, + { 0x00800801, 0x20b00236, 0x01ee8070, 0x00000000 }, + { 0x00800401, 0x20c00236, 0x01ee8100, 0x00000000 }, + { 0x00800801, 0x20d00236, 0x01ee8110, 0x00000000 }, + { 0x00800401, 0x20e00236, 0x01ee8120, 0x00000000 }, + { 0x00800801, 0x20f00236, 0x01ee8130, 0x00000000 }, + { 0x00800401, 0x21000236, 0x01ee8140, 0x00000000 }, + { 0x00800801, 0x21100236, 0x01ee8150, 0x00000000 }, + { 0x00800401, 0x21200236, 0x01ee8160, 0x00000000 }, + { 0x00800801, 0x21300236, 0x01ee8170, 0x00000000 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x1218a000 }, + { 0x00200401, 0x204001a5, 0x00450064, 0x00000000 }, + { 0x00000801, 0x20480061, 0x00000000, 0x000f000f }, + { 0x01600031, 0x21400021, 0x508d0040, 0x00000200 }, + { 0x00000001, 0x220001ec, 0x00000000, 0x03400340 }, + { 0x00800401, 0x41600236, 0x00d28000, 0x00000000 }, + { 0x00800801, 0x41610236, 0x00d28080, 0x00000000 }, + { 0x00800401, 0x41800236, 0x00d28020, 0x00000000 }, + { 0x00800801, 0x41810236, 0x00d280a0, 0x00000000 }, + { 0x00800401, 0x41a00236, 0x00d28040, 0x00000000 }, + { 0x00800801, 0x41a10236, 0x00d280c0, 0x00000000 }, + { 0x00800401, 0x41c00236, 0x00d28060, 0x00000000 }, + { 0x00800801, 0x41c10236, 0x00d280e0, 0x00000000 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x0a18a001 }, + { 0x0000040c, 0x20443ca5, 0x00000044, 0x00010001 }, + { 0x00000801, 0x20480061, 0x00000000, 0x0007000f }, + { 0x0a600031, 0x21600021, 0x508d0040, 0x00000200 }, + { 0x00000001, 0x21400021, 0x00000140, 0x00000000 }, + { 0x00000001, 0x21600021, 0x00000160, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x01000005, 0x20602d29, 0x02000042, 0x00070007 }, + { 0x01000005, 0x20422d29, 0x00000042, 0xfff8fff8 }, + { 0x00000001, 0x202001e9, 0x00000000, 0x100c100c }, + { 0x00600001, 0x27e00021, 0x008d0000, 0x00000000 }, + { 0x00000009, 0x27e82d21, 0x00000040, 0x00060006 }, + { 0x00600001, 0x20a00022, 0x008d0000, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000100 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x01800031, 0x20801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20400022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20480c22, 0x000007e8, 0x00000080 }, + { 0x02800031, 0x21001d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20600022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20680c22, 0x000007e8, 0x00000100 }, + { 0x03800031, 0x21801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20800022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20880c22, 0x000007e8, 0x00000180 }, + { 0x04800031, 0x22001d09, 0x408d0000, 0x02488400 }, + { 0x02600005, 0x20000c20, 0x02000090, 0x00002000 }, + { 0x00000006, 0x20880c21, 0x00000088, 0x00200000 }, + { 0x00200009, 0x20845529, 0x0000009c, 0x00450020 }, + { 0x00200001, 0x40840231, 0x00450094, 0x00000000 }, + { 0x00310001, 0x20840229, 0x02450094, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000095, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000094 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0080, 0x00000000 }, + { 0x00802001, 0x20800061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x020000d0, 0x00002000 }, + { 0x00000006, 0x20c80c21, 0x000000c8, 0x00200000 }, + { 0x00200009, 0x20c45529, 0x000000dc, 0x00450020 }, + { 0x00200001, 0x40c40231, 0x004500d4, 0x00000000 }, + { 0x00310001, 0x20c40229, 0x024500d4, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x000000d5, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x000000d4 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d00c0, 0x00000000 }, + { 0x00802001, 0x20c00061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000110, 0x00002000 }, + { 0x00000006, 0x21080c21, 0x00000108, 0x00200000 }, + { 0x00200009, 0x21045529, 0x0000011c, 0x00450020 }, + { 0x00200001, 0x41040231, 0x00450114, 0x00000000 }, + { 0x00310001, 0x21040229, 0x02450114, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000115, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000114 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0100, 0x00000000 }, + { 0x00802001, 0x21000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000150, 0x00002000 }, + { 0x00000006, 0x21480c21, 0x00000148, 0x00200000 }, + { 0x00200009, 0x21445529, 0x0000015c, 0x00450020 }, + { 0x00200001, 0x41440231, 0x00450154, 0x00000000 }, + { 0x00310001, 0x21440229, 0x02450154, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000155, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000154 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0140, 0x00000000 }, + { 0x00802001, 0x21400061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000190, 0x00002000 }, + { 0x00000006, 0x21880c21, 0x00000188, 0x00200000 }, + { 0x00200009, 0x21845529, 0x0000019c, 0x00450020 }, + { 0x00200001, 0x41840231, 0x00450194, 0x00000000 }, + { 0x00310001, 0x21840229, 0x02450194, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000195, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000194 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0180, 0x00000000 }, + { 0x00802001, 0x21800061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x020001d0, 0x00002000 }, + { 0x00000006, 0x21c80c21, 0x000001c8, 0x00200000 }, + { 0x00200009, 0x21c45529, 0x000001dc, 0x00450020 }, + { 0x00200001, 0x41c40231, 0x004501d4, 0x00000000 }, + { 0x00310001, 0x21c40229, 0x024501d4, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x000001d5, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x000001d4 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d01c0, 0x00000000 }, + { 0x00802001, 0x21c00061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000210, 0x00002000 }, + { 0x00000006, 0x22080c21, 0x00000208, 0x00200000 }, + { 0x00200009, 0x22045529, 0x0000021c, 0x00450020 }, + { 0x00200001, 0x42040231, 0x00450214, 0x00000000 }, + { 0x00310001, 0x22040229, 0x02450214, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000215, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000214 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0200, 0x00000000 }, + { 0x00802001, 0x22000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000250, 0x00002000 }, + { 0x00000006, 0x22480c21, 0x00000248, 0x00200000 }, + { 0x00200009, 0x22445529, 0x0000025c, 0x00450020 }, + { 0x00200001, 0x42440231, 0x00450254, 0x00000000 }, + { 0x00310001, 0x22440229, 0x02450254, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000255, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000254 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0240, 0x00000000 }, + { 0x00802001, 0x22400061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x01000040, 0x20423dad, 0x00000042, 0xfff8fff8 }, + { 0x00600001, 0x21600022, 0x008d07e0, 0x00000000 }, + { 0x00802001, 0x21800022, 0x008d0080, 0x00000000 }, + { 0x00802001, 0x21c00022, 0x008d00c0, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20280c22, 0x000007e8, 0x00000080 }, + { 0x00802001, 0x20400022, 0x008d0100, 0x00000000 }, + { 0x00802001, 0x20800022, 0x008d0140, 0x00000000 }, + { 0x01800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x21680c22, 0x000007e8, 0x00000100 }, + { 0x00802001, 0x21800022, 0x008d0180, 0x00000000 }, + { 0x00802001, 0x21c00022, 0x008d01c0, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x20280c22, 0x000007e8, 0x00000180 }, + { 0x00802001, 0x20400022, 0x008d0200, 0x00000000 }, + { 0x00802001, 0x20800022, 0x008d0240, 0x00000000 }, + { 0x01800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x27e80c21, 0x000007e8, 0x00000200 }, + { 0x01000010, 0x20002da0, 0x02000060, 0x00000000 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffff00 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000040 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x01800031, 0x20801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20400022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20480c22, 0x000007e8, 0x00000080 }, + { 0x02800031, 0x21001d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20600022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20680c22, 0x000007e8, 0x00000100 }, + { 0x03800031, 0x21801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20800022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20880c22, 0x000007e8, 0x00000180 }, + { 0x04800031, 0x22001d09, 0x408d0000, 0x02488400 }, + { 0x00000001, 0x220400e0, 0x00000000, 0x00a00080 }, + { 0x00600001, 0x20a00022, 0x008d0000, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02008810, 0x00002000 }, + { 0x01000040, 0x20603dad, 0x00000060, 0xffffffff }, + { 0x00000006, 0xa8080c21, 0x00008808, 0x00200000 }, + { 0x00200009, 0xa8045529, 0x0000881c, 0x00450020 }, + { 0x00200001, 0xc8040231, 0x00458814, 0x00000000 }, + { 0x00310001, 0xa8040229, 0x02458814, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00008815, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00008814 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d8800, 0x00000000 }, + { 0x00802001, 0xa8000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00600001, 0x21600022, 0x008d07e0, 0x00000000 }, + { 0x00802001, 0x21800022, 0x008d8800, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000040, 0x27e80c21, 0x000007e8, 0x00000040 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00400040 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffffda }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x01000005, 0x20602d29, 0x02000042, 0x00070007 }, + { 0x01000005, 0x20422d29, 0x00000042, 0xfff8fff8 }, + { 0x00600001, 0x27e00021, 0x008d0000, 0x00000000 }, + { 0x00000009, 0x27e82d21, 0x00000040, 0x00060006 }, + { 0x00600001, 0x20a00022, 0x008d0000, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000260 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x01800031, 0x20801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20400022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20480c22, 0x000007e8, 0x00000080 }, + { 0x02800031, 0x21001d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20600022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20680c22, 0x000007e8, 0x00000100 }, + { 0x03800031, 0x21801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20800022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20880c22, 0x000007e8, 0x00000180 }, + { 0x04800031, 0x22001d09, 0x408d0000, 0x02488400 }, + { 0x00200001, 0x20640229, 0x00450094, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000090, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000090, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000095, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000095, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000090, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x20881c21, 0x00000088, 0x00200000 }, + { 0x00200001, 0x20840129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0080, 0x00000000 }, + { 0x00802001, 0x20800061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000090, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000009c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000009c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200009c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000090, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200009c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200009c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x20843d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x20862d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x40840231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x004500d4, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x020000d0, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x000000d0, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x000000d5, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x000000d5, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x000000d0, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x20c81c21, 0x000000c8, 0x00200000 }, + { 0x00200001, 0x20c40129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d00c0, 0x00000000 }, + { 0x00802001, 0x20c00061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x000000d0, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x000000dc, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x000000dc, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x020000dc, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x020000d0, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x020000dc, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x020000dc, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x20c43d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x20c62d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x40c40231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450114, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000110, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000110, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000115, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000115, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000110, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x21081c21, 0x00000108, 0x00200000 }, + { 0x00200001, 0x21040129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0100, 0x00000000 }, + { 0x00802001, 0x21000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000110, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000011c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000011c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200011c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000110, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200011c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200011c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x21043d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x21062d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x41040231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450154, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000150, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000150, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000155, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000155, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000150, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x21481c21, 0x00000148, 0x00200000 }, + { 0x00200001, 0x21440129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0140, 0x00000000 }, + { 0x00802001, 0x21400061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000150, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000015c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000015c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200015c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000150, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200015c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200015c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x21443d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x21462d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x41440231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450194, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000190, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000190, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000195, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000195, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000190, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x21881c21, 0x00000188, 0x00200000 }, + { 0x00200001, 0x21840129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0180, 0x00000000 }, + { 0x00802001, 0x21800061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000190, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000019c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000019c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200019c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000190, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200019c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200019c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x21843d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x21862d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x41840231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x004501d4, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x020001d0, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x000001d0, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x000001d5, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x000001d5, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x000001d0, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x21c81c21, 0x000001c8, 0x00200000 }, + { 0x00200001, 0x21c40129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d01c0, 0x00000000 }, + { 0x00802001, 0x21c00061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x000001d0, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x000001dc, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x000001dc, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x020001dc, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x020001d0, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x020001dc, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x020001dc, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x21c43d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x21c62d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x41c40231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450214, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000210, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000210, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000215, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000215, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000210, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x22081c21, 0x00000208, 0x00200000 }, + { 0x00200001, 0x22040129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0200, 0x00000000 }, + { 0x00802001, 0x22000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000210, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000021c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000021c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200021c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000210, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200021c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200021c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x22043d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x22062d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x42040231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450254, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000250, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000250, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000255, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000255, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000250, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x22481c21, 0x00000248, 0x00200000 }, + { 0x00200001, 0x22440129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0240, 0x00000000 }, + { 0x00802001, 0x22400061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000250, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000025c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000025c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200025c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000250, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200025c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200025c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x22443d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x22462d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x42440231, 0x00660064, 0x00000000 }, + { 0x01000040, 0x20423dad, 0x00000042, 0xfff8fff8 }, + { 0x00600001, 0x21600022, 0x008d07e0, 0x00000000 }, + { 0x00802001, 0x21800022, 0x008d0080, 0x00000000 }, + { 0x00802001, 0x21c00022, 0x008d00c0, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20280c22, 0x000007e8, 0x00000080 }, + { 0x00802001, 0x20400022, 0x008d0100, 0x00000000 }, + { 0x00802001, 0x20800022, 0x008d0140, 0x00000000 }, + { 0x01800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x21680c22, 0x000007e8, 0x00000100 }, + { 0x00802001, 0x21800022, 0x008d0180, 0x00000000 }, + { 0x00802001, 0x21c00022, 0x008d01c0, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x20280c22, 0x000007e8, 0x00000180 }, + { 0x00802001, 0x20400022, 0x008d0200, 0x00000000 }, + { 0x00802001, 0x20800022, 0x008d0240, 0x00000000 }, + { 0x01800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x27e80c21, 0x000007e8, 0x00000200 }, + { 0x01000010, 0x20002da0, 0x02000060, 0x00000000 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xfffffda0 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000006e }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x01800031, 0x20801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20400022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20480c22, 0x000007e8, 0x00000080 }, + { 0x02800031, 0x21001d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20600022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20680c22, 0x000007e8, 0x00000100 }, + { 0x03800031, 0x21801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20800022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20880c22, 0x000007e8, 0x00000180 }, + { 0x04800031, 0x22001d09, 0x408d0000, 0x02488400 }, + { 0x00000001, 0x220400e0, 0x00000000, 0x00a00080 }, + { 0x00600001, 0x20a00022, 0x008d0000, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00458814, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02008810, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00008810, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00008815, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00008815, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00008810, 0x00002000 }, + { 0x01000040, 0x20603dad, 0x02000060, 0xffffffff }, + { 0x00000006, 0xa8080c21, 0x00008808, 0x00200000 }, + { 0x00200001, 0xa8040129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d8800, 0x00000000 }, + { 0x00802001, 0xa8000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000022 }, + { 0x02600005, 0x20000d20, 0x02000066, 0x00000001 }, + { 0x02600005, 0x20000c20, 0x00008810, 0x00004000 }, + { 0x00000001, 0x20680129, 0x0000881c, 0x00000000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000881c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200881c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02008810, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200881c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200881c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x00000068, 0x000d0070 }, + { 0x01000040, 0x20603dad, 0x02000060, 0x00000000 }, + { 0x00000009, 0xa8043d09, 0x00000600, 0x000c000c }, + { 0x00000005, 0xa8062d09, 0x00000600, 0xf000f000 }, + { 0x00200001, 0xc8040231, 0x00660064, 0x00000000 }, + { 0x00600001, 0x21600022, 0x008d07e0, 0x00000000 }, + { 0x00802001, 0x21800022, 0x008d8800, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000040, 0x27e80c21, 0x000007e8, 0x00000040 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00400040 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xffffffac }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, diff --git a/shaders/h264/mc/avc_mc.g4b.gen5 b/shaders/h264/mc/avc_mc.g4b.gen5 new file mode 100644 index 0000000..5a91f32 --- /dev/null +++ b/shaders/h264/mc/avc_mc.g4b.gen5 @@ -0,0 +1,2938 @@ + { 0x00600001, 0x27c00021, 0x008d0000, 0x00000000 }, + { 0x00200009, 0x27fa3e29, 0x00450064, 0x00040004 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x000000d4 }, + { 0x00000005, 0x220e3e2c, 0x00000070, 0x000f000f }, + { 0x00000001, 0x26a00221, 0x00009c38, 0x00000000 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a0 }, + { 0x00a02001, 0x24000229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24400229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24800229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24c00229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x25000229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x25400229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x25800229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x25c00229, 0x00b10624, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000086 }, + { 0x00000001, 0x22080060, 0x00000000, 0x06440640 }, + { 0x00a02001, 0x24000229, 0x00009003, 0x00000000 }, + { 0x00a02001, 0x24400229, 0x0000900b, 0x00000000 }, + { 0x00a02001, 0x24800229, 0x00009013, 0x00000000 }, + { 0x00a02001, 0x24c00229, 0x0000901b, 0x00000000 }, + { 0x00a02001, 0x25000229, 0x00009023, 0x00000000 }, + { 0x00a02001, 0x25400229, 0x0000902b, 0x00000000 }, + { 0x00a02001, 0x25800229, 0x00009033, 0x00000000 }, + { 0x00a02001, 0x25c00229, 0x0000903b, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000072 }, + { 0x02600005, 0x20000c20, 0x0000006c, 0x00000002 }, + { 0x00600005, 0x24000c20, 0x0000006c, 0x00000011 }, + { 0x01600007, 0x20000c00, 0x028d0400, 0x00000011 }, + { 0x00780001, 0x26240169, 0x00000000, 0x80808080 }, + { 0x00780001, 0x66430231, 0x028d0624, 0x00000000 }, + { 0x00780001, 0x66630231, 0x028d062c, 0x00000000 }, + { 0x00780001, 0x26240231, 0x00cf0643, 0x00000000 }, + { 0x00780001, 0x262c0231, 0x00cf0663, 0x00000000 }, + { 0x00800040, 0x25e04629, 0x00cf0643, 0x00b10624 }, + { 0x00600040, 0x25e02529, 0x008d05e0, 0x008d05f0 }, + { 0x00400040, 0x25e02529, 0x006905e0, 0x006905e8 }, + { 0x00200040, 0x25e02529, 0x004505e0, 0x004505e4 }, + { 0x00a02040, 0x2400252c, 0x000005e0, 0x000005e2 }, + { 0x00a02040, 0x24003d8c, 0x00b10400, 0x00100010 }, + { 0x00a02008, 0x24003d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x24403d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x24803d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x24c03d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x25003d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x25403d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x25803d89, 0x00b10400, 0x00050005 }, + { 0x00a02008, 0x25c03d89, 0x00b10400, 0x00050005 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000044 }, + { 0x00600041, 0x26806e2d, 0x008d062c, 0x89abcdef }, + { 0x00600041, 0x26906e2d, 0x008d0623, 0xfedcba98 }, + { 0x00600041, 0x26a06e2d, 0x00cf0663, 0x89abcdef }, + { 0x00600041, 0x26b06e2d, 0x00cf0643, 0x0fedcba9 }, + { 0x00000041, 0x26be3e2d, 0x00000623, 0xfff8fff8 }, + { 0x00802040, 0x268035ad, 0x008d4680, 0x008d0690 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0688 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0684 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0682 }, + { 0x00800001, 0x240001ec, 0x00000000, 0x00200020 }, + { 0x00200048, 0x24003dac, 0x00a00680, 0x00050005 }, + { 0x00200008, 0x26e03d8d, 0x00450400, 0x00060006 }, + { 0x00800001, 0x240001ec, 0x00000000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x00000633, 0x00100010 }, + { 0x00800048, 0x26c03e2d, 0x0000067f, 0x00100010 }, + { 0x00800048, 0x272055ad, 0x000006e0, 0x00b10040 }, + { 0x00600041, 0x268055ad, 0x000006e2, 0x00ae0040 }, + { 0x00600041, 0x26a055ad, 0x000006e2, 0x00ae0041 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00680 }, + { 0x80a02008, 0x44003d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00682 }, + { 0x80a02008, 0x44403d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00684 }, + { 0x80a02008, 0x44803d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00686 }, + { 0x80a02008, 0x44c03d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b00688 }, + { 0x80a02008, 0x45003d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b0068a }, + { 0x80a02008, 0x45403d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b0068c }, + { 0x80a02008, 0x45803d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00b0068e }, + { 0x80a02008, 0x45c03d91, 0x00b10400, 0x00050005 }, + { 0x00000001, 0x22040060, 0x00000000, 0x00900080 }, + { 0x01000005, 0x20000c20, 0x02000068, 0x00003c00 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000020 }, + { 0x80802040, 0x440045b1, 0x008d8800, 0x00ae0400 }, + { 0x80802040, 0x444045b1, 0x008d8820, 0x00ae0440 }, + { 0x80802040, 0x448045b1, 0x008d8840, 0x00ae0480 }, + { 0x80802040, 0x44c045b1, 0x008d8860, 0x00ae04c0 }, + { 0x80802040, 0x441045b1, 0x008d8880, 0x00ae0410 }, + { 0x80802040, 0x445045b1, 0x008d88a0, 0x00ae0450 }, + { 0x80802040, 0x449045b1, 0x008d88c0, 0x00ae0490 }, + { 0x80802040, 0x44d045b1, 0x008d88e0, 0x00ae04d0 }, + { 0x80802040, 0x450045b1, 0x008d8900, 0x00ae0500 }, + { 0x80802040, 0x454045b1, 0x008d8920, 0x00ae0540 }, + { 0x80802040, 0x458045b1, 0x008d8940, 0x00ae0580 }, + { 0x80802040, 0x45c045b1, 0x008d8960, 0x00ae05c0 }, + { 0x80802040, 0x451045b1, 0x008d8980, 0x00ae0510 }, + { 0x80802040, 0x455045b1, 0x008d89a0, 0x00ae0550 }, + { 0x80802040, 0x459045b1, 0x008d89c0, 0x00ae0590 }, + { 0x80802040, 0x45d045b1, 0x008d89e0, 0x00ae05d0 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x01800180 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x000000e2 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000018 }, + { 0x00000001, 0x27800021, 0x00000780, 0x00000000 }, + { 0x00000001, 0x27a00021, 0x000007a0, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00200440, 0x27c055a5, 0x004507fa, 0x0045002a }, + { 0x00000801, 0x27c80061, 0x00000000, 0x0000001b }, + { 0x00000040, 0x22000d20, 0x00000062, 0x02186000 }, + { 0x01600031, 0x26200021, 0x408d07c0, 0x00000200 }, + { 0x00000440, 0x27c43ca5, 0x000007c4, 0x00010001 }, + { 0x00000801, 0x27c80061, 0x00000000, 0x000f0003 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x00100000 }, + { 0x02600031, 0x26400021, 0x408d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00000008, 0x27fc3dad, 0x000007fc, 0x00010001 }, + { 0x00200440, 0x27c055a5, 0x004507fa, 0x0045002a }, + { 0x00000801, 0x27c80061, 0x00000000, 0x00000013 }, + { 0x00000040, 0x22000c00, 0x00000200, 0xefffc001 }, + { 0x01600031, 0x26200021, 0x408d07c0, 0x00000200 }, + { 0x00000440, 0x27c43ca5, 0x000007c4, 0x00010001 }, + { 0x00000801, 0x27c80061, 0x00000000, 0x00070003 }, + { 0x05600031, 0x26400021, 0x408d07c0, 0x00000200 }, + { 0x00000008, 0x220e3e2c, 0x0000006c, 0x00060006 }, + { 0x00000001, 0x26a002a5, 0x00009c3c, 0x00000000 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a0 }, + { 0x02600005, 0x20000c20, 0x0000006c, 0x00000002 }, + { 0x00780001, 0x26240169, 0x00000000, 0x80808080 }, + { 0x02400005, 0x20000c20, 0x0200006c, 0x00000001 }, + { 0x00560001, 0x46420129, 0x02690624, 0x00000000 }, + { 0x02400005, 0x20000c20, 0x0200006c, 0x00000010 }, + { 0x00560001, 0x46520129, 0x0269062c, 0x00000000 }, + { 0x00780001, 0x26240129, 0x00ae0642, 0x00000000 }, + { 0x00800040, 0x24004629, 0x00b10624, 0x00650642 }, + { 0x00600040, 0x24002529, 0x00650400, 0x00650404 }, + { 0x00600040, 0x25202529, 0x00050400, 0x00050404 }, + { 0x00600040, 0x25702529, 0x00050408, 0x0005040c }, + { 0x00560001, 0x26240169, 0x00000000, 0x80808080 }, + { 0x00460001, 0x26240129, 0x028a0652, 0x00000000 }, + { 0x00560001, 0x46520129, 0x02690624, 0x00000000 }, + { 0x02400005, 0x20000c20, 0x0200006c, 0x00000001 }, + { 0x00560001, 0x46420169, 0x02000000, 0x80808080 }, + { 0x00460001, 0x46420129, 0x0069062c, 0x00000000 }, + { 0x00560001, 0x262c0129, 0x008a0642, 0x00000000 }, + { 0x00600040, 0x24004629, 0x008d0624, 0x00650652 }, + { 0x00600040, 0x24104629, 0x00650642, 0x008d062c }, + { 0x00600040, 0x24002529, 0x00650400, 0x00650404 }, + { 0x00600040, 0x25302529, 0x00050408, 0x0005040c }, + { 0x00600040, 0x25602529, 0x00050400, 0x00050404 }, + { 0x00a02040, 0x24003d2c, 0x00b10520, 0x00040004 }, + { 0x00a02008, 0x24003d89, 0x00b10400, 0x00030003 }, + { 0x00a02008, 0x24403d89, 0x00b10400, 0x00030003 }, + { 0x00a02040, 0x24003d2c, 0x00b10560, 0x00040004 }, + { 0x00a02008, 0x24803d89, 0x00b10400, 0x00030003 }, + { 0x00a02008, 0x24c03d89, 0x00b10400, 0x00030003 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000048 }, + { 0x00000001, 0x22080060, 0x00000000, 0x06440640 }, + { 0x00a02001, 0x24000229, 0x00059002, 0x00000000 }, + { 0x00a02001, 0x24400229, 0x0005900a, 0x00000000 }, + { 0x00a02001, 0x24800229, 0x00059012, 0x00000000 }, + { 0x00a02001, 0x24c00229, 0x0005901a, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000003c }, + { 0x00a02001, 0x24000229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24400229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24800229, 0x00b10624, 0x00000000 }, + { 0x00a02001, 0x24c00229, 0x00b10624, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000032 }, + { 0x00600041, 0x26806e2d, 0x008d062c, 0x44332211 }, + { 0x00600041, 0x26906e2d, 0x008d0622, 0xffeeddcc }, + { 0x00600041, 0x26a06e2d, 0x00650652, 0x44332211 }, + { 0x00600041, 0x26b06e2d, 0x00650642, 0x00ffeedd }, + { 0x00200041, 0x26bc3e2d, 0x00450622, 0xfffcfffc }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0690 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0688 }, + { 0x00802040, 0x268035ad, 0x008d0680, 0x008d0684 }, + { 0x00800001, 0x240001ec, 0x00000000, 0x00200020 }, + { 0x00400048, 0x24003dac, 0x00a50680, 0x00220022 }, + { 0x00400008, 0x26e03d8d, 0x00690400, 0x00060006 }, + { 0x00800001, 0x240001ec, 0x00000000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x00050632, 0x00100010 }, + { 0x00800048, 0x26c03e2d, 0x0005065e, 0x00100010 }, + { 0x00800048, 0x272055ad, 0x000506e0, 0x00240044 }, + { 0x00600041, 0x268055ad, 0x000506e4, 0x00440044 }, + { 0x00600041, 0x26a055ad, 0x000506e4, 0x00440045 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00050680 }, + { 0x80a02008, 0x44003d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00050684 }, + { 0x80a02008, 0x44403d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x00050688 }, + { 0x80a02008, 0x44803d91, 0x00b10400, 0x00050005 }, + { 0x00a02040, 0x240035ac, 0x00b10720, 0x0005068c }, + { 0x80a02008, 0x44c03d91, 0x00b10400, 0x00050005 }, + { 0x00000040, 0x22063d8c, 0x00000204, 0x00100010 }, + { 0x80802040, 0x640045b1, 0x008d8880, 0x00cf0400 }, + { 0x80802040, 0x644045b1, 0x008d88a0, 0x00cf0440 }, + { 0x80802040, 0x648045b1, 0x008d88c0, 0x00cf0480 }, + { 0x80802040, 0x64c045b1, 0x008d88e0, 0x00cf04c0 }, + { 0x80802040, 0x640245b1, 0x008d8900, 0x00cf0402 }, + { 0x80802040, 0x644245b1, 0x008d8920, 0x00cf0442 }, + { 0x80802040, 0x648245b1, 0x008d8940, 0x00cf0482 }, + { 0x80802040, 0x64c245b1, 0x008d8960, 0x00cf04c2 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x0007000f }, + { 0x00200801, 0x27c001a1, 0x004507fa, 0x00000000 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x08004000 }, + { 0x00800001, 0x20400232, 0x00d20400, 0x00000000 }, + { 0x00800001, 0x20500232, 0x00d20420, 0x00000000 }, + { 0x00800001, 0x20600232, 0x00d20440, 0x00000000 }, + { 0x00800001, 0x20700232, 0x00d20460, 0x00000000 }, + { 0x00800001, 0x20800232, 0x00d20480, 0x00000000 }, + { 0x00800001, 0x20900232, 0x00d204a0, 0x00000000 }, + { 0x00800001, 0x20a00232, 0x00d204c0, 0x00000000 }, + { 0x00800001, 0x20b00232, 0x00d204e0, 0x00000000 }, + { 0x01600031, 0x27a00021, 0x508d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00200801, 0x27c001a1, 0x004507fa, 0x00000000 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x0ff04000 }, + { 0x00000001, 0x22080060, 0x00000000, 0x04400400 }, + { 0x00a02001, 0x20400232, 0x00d29000, 0x00000000 }, + { 0x00a02001, 0x20500232, 0x00d29020, 0x00000000 }, + { 0x00a02001, 0x20800232, 0x00d29080, 0x00000000 }, + { 0x00a02001, 0x20900232, 0x00d290a0, 0x00000000 }, + { 0x00a02001, 0x20c00232, 0x00d29100, 0x00000000 }, + { 0x00a02001, 0x20d00232, 0x00d29120, 0x00000000 }, + { 0x00a02001, 0x21000232, 0x00d29180, 0x00000000 }, + { 0x00a02001, 0x21100232, 0x00d291a0, 0x00000000 }, + { 0x01600031, 0x27800021, 0x508d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00600001, 0x27c00021, 0x008d0000, 0x00000000 }, + { 0x00200009, 0x27fa3e29, 0x00450064, 0x00040004 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0xffffff04 }, + { 0x00000001, 0x220401ec, 0x00000000, 0x00800080 }, + { 0x00000001, 0x22080060, 0x00000000, 0x04100400 }, + { 0x00200008, 0x27006e2d, 0x0000006c, 0x00000040 }, + { 0x00400001, 0x27400061, 0x00000000, 0x00000000 }, + { 0x00800001, 0x272001a9, 0x00b10620, 0x00000000 }, + { 0x02600005, 0x20001c20, 0x0000006c, 0x00000008 }, + { 0x00110001, 0x27230231, 0x00000624, 0x00000000 }, + { 0x00600001, 0x27420231, 0x00cf0643, 0x00000000 }, + { 0x00110001, 0x27410231, 0x00000643, 0x00000000 }, + { 0x00240001, 0x27400231, 0x00000623, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x0200006c, 0x00000002 }, + { 0x00010001, 0x27400231, 0x02000624, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x0200006c, 0x00000001 }, + { 0x00110001, 0x27420231, 0x02000623, 0x00000000 }, + { 0x00000005, 0x26803e2d, 0x00000070, 0x000f000f }, + { 0x00000001, 0x270801ad, 0x00000700, 0x00000000 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000084 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00800001, 0x27200231, 0x00b10628, 0x00000000 }, + { 0x02600005, 0x20001c20, 0x0200006c, 0x00000004 }, + { 0x00680001, 0x272c0231, 0x028d0634, 0x00000000 }, + { 0x00780001, 0x272c0231, 0x02000633, 0x00000000 }, + { 0x00400001, 0x27420231, 0x00a0040e, 0x00000000 }, + { 0x00400001, 0x27460231, 0x00a0044e, 0x00000000 }, + { 0x00200001, 0x27400231, 0x0000062b, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x0200006c, 0x00000002 }, + { 0x00010001, 0x27400231, 0x0200062c, 0x00000000 }, + { 0x00110001, 0x27410231, 0x0200040e, 0x00000000 }, + { 0x00000008, 0x26803e2d, 0x00000070, 0x00040004 }, + { 0x00200040, 0x22083d8c, 0x00450208, 0x00800080 }, + { 0x00000006, 0x27083dad, 0x00000700, 0x00010001 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000064 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00a02001, 0xb3800231, 0x00d20400, 0x00000000 }, + { 0x00a02001, 0xb3a00231, 0x00d20480, 0x00000000 }, + { 0x00a02001, 0xb3c00231, 0x00d20440, 0x00000000 }, + { 0x00a02001, 0xb3e00231, 0x00d204c0, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x00000060, 0x00000002 }, + { 0x01000005, 0x20001c20, 0x0200006c, 0x00000001 }, + { 0x01010005, 0x20001c20, 0x00000060, 0x00004000 }, + { 0x02010005, 0x20001c20, 0x02000060, 0x00000010 }, + { 0x00030001, 0x27230231, 0x0000065f, 0x00000000 }, + { 0x00030220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02600005, 0x20001c20, 0x0000006c, 0x00000001 }, + { 0x00460001, 0x27200231, 0x0069065c, 0x00000000 }, + { 0x00110001, 0x27230231, 0x00000458, 0x00000000 }, + { 0x00600001, 0x27240231, 0x008d0458, 0x00000000 }, + { 0x00600001, 0x272c0231, 0x008d0478, 0x00000000 }, + { 0x00600001, 0x27420231, 0x00cf0663, 0x00000000 }, + { 0x00000001, 0x27400231, 0x00000458, 0x00000000 }, + { 0x00010001, 0x27410231, 0x0000065f, 0x00000000 }, + { 0x00110001, 0x27410231, 0x00000663, 0x00000000 }, + { 0x02000005, 0x20001c20, 0x0200006c, 0x00000010 }, + { 0x00110001, 0x27420231, 0x0200065f, 0x00000000 }, + { 0x00000005, 0x26803e2d, 0x00000071, 0x000f000f }, + { 0x00000006, 0x27083dad, 0x00000702, 0x00020002 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000030 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00400001, 0x27200231, 0x0069045c, 0x00000000 }, + { 0x00600001, 0x27240231, 0x008d0478, 0x00000000 }, + { 0x00800001, 0x272c0231, 0x0000047f, 0x00000000 }, + { 0x00400001, 0x27420231, 0x00a0048e, 0x00000000 }, + { 0x00400001, 0x27460231, 0x00a004ce, 0x00000000 }, + { 0x00000001, 0x27400231, 0x00000478, 0x00000000 }, + { 0x00000001, 0x27410231, 0x0000045f, 0x00000000 }, + { 0x00000008, 0x26803e2d, 0x00000071, 0x00040004 }, + { 0x00200040, 0x22083d8c, 0x00450208, 0x00800080 }, + { 0x00000006, 0x27083dad, 0x00000700, 0x00030003 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00a02001, 0xb3800231, 0x00d20480, 0x00000000 }, + { 0x00a02001, 0xb3a00231, 0x00d20500, 0x00000000 }, + { 0x00a02001, 0xb3c00231, 0x00d204c0, 0x00000000 }, + { 0x00a02001, 0xb3e00231, 0x00d20540, 0x00000000 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000100 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0xfffffe74 }, + { 0x00000001, 0x27800021, 0x00000780, 0x00000000 }, + { 0x00000001, 0x27a00021, 0x000007a0, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000001, 0x27340231, 0x00000733, 0x00000000 }, + { 0x00600001, 0x274a0231, 0x00000749, 0x00000000 }, + { 0x00800040, 0x24003e2c, 0x00b10723, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10724, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10725, 0x00010001 }, + { 0x00800008, 0x26c03d8d, 0x008d0400, 0x00020002 }, + { 0x00800040, 0x24003e2c, 0x00b10740, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10741, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10742, 0x00010001 }, + { 0x00800008, 0x26a03d8d, 0x008d0400, 0x00020002 }, + { 0x00800001, 0x27240231, 0x00d206c0, 0x00000000 }, + { 0x00600001, 0x27400231, 0x00ae06a2, 0x00000000 }, + { 0x00000001, 0x27230231, 0x000006a0, 0x00000000 }, + { 0x00000005, 0x220e3dac, 0x00000680, 0x000f000f }, + { 0x00000001, 0x26a00221, 0x00009c2c, 0x00000000 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a0 }, + { 0x80800040, 0xd00045b1, 0x00b18800, 0x000d0724 }, + { 0x80800040, 0xd02045b1, 0x00b18820, 0x000d0724 }, + { 0x80800040, 0xd04045b1, 0x00b18840, 0x000d0724 }, + { 0x80800040, 0xd06045b1, 0x00b18860, 0x000d0724 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x80800040, 0xd00045b1, 0x00b18800, 0x002c0740 }, + { 0x80800040, 0xd02045b1, 0x00b18820, 0x002c0742 }, + { 0x80800040, 0xd04045b1, 0x00b18840, 0x002c0744 }, + { 0x80800040, 0xd06045b1, 0x00b18860, 0x002c0746 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x02800005, 0x20003da0, 0x00000708, 0x00020002 }, + { 0x02600005, 0x20003da0, 0x02000708, 0x00010001 }, + { 0x009a0001, 0x27200169, 0x00000000, 0x80808080 }, + { 0x00780001, 0x27400231, 0x028d0724, 0x00000000 }, + { 0x00780001, 0x27240231, 0x008d0740, 0x00000000 }, + { 0x00600040, 0x25e04629, 0x008d0724, 0x008d0740 }, + { 0x00400040, 0x25e02529, 0x006905e0, 0x006905e8 }, + { 0x00200040, 0x25e02529, 0x004505e0, 0x004505e4 }, + { 0x00800040, 0x2400252c, 0x000005e0, 0x000005e2 }, + { 0x00800040, 0x24003d8c, 0x008d0400, 0x00080008 }, + { 0x00800008, 0x26803d8d, 0x008d0400, 0x00040004 }, + { 0x80800040, 0xd00035b1, 0x00b18800, 0x00b10680 }, + { 0x80800040, 0xd02035b1, 0x00b18820, 0x00b10680 }, + { 0x80800040, 0xd04035b1, 0x00b18840, 0x00b10680 }, + { 0x80800040, 0xd06035b1, 0x00b18860, 0x00b10680 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00600001, 0x27340231, 0x008d0733, 0x00000000 }, + { 0x00800040, 0x24003e2c, 0x00b10726, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10725, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10724, 0x00010001 }, + { 0x00800008, 0x26803d8d, 0x00b10400, 0x00020002 }, + { 0x80800040, 0xd00035b1, 0x00b18800, 0x002d0680 }, + { 0x80800040, 0xd02035b1, 0x00b18820, 0x002d0684 }, + { 0x80800040, 0xd04035b1, 0x00b18840, 0x002d0688 }, + { 0x80800040, 0xd06035b1, 0x00b18860, 0x002d068c }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00400009, 0x26c05421, 0x00000744, 0x00690050 }, + { 0x00400009, 0x26d05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26a00231, 0x00cf06c3, 0x00000000 }, + { 0x00800001, 0x26a80231, 0x00b10723, 0x00000000 }, + { 0x00800040, 0x24003e2c, 0x00b106a2, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a1, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a0, 0x00010001 }, + { 0x00800008, 0x26a03d8d, 0x00b10400, 0x00020002 }, + { 0x00200040, 0x220c3eac, 0x00450056, 0x06a006a0 }, + { 0x80800040, 0xd06035b1, 0x01ed9800, 0x00b18860 }, + { 0x80800040, 0xd04035b1, 0x01ed9804, 0x00b18840 }, + { 0x80800040, 0xd02035b1, 0x01ed9808, 0x00b18820 }, + { 0x80800040, 0xd00035b1, 0x01ed980c, 0x00b18800 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00400009, 0x26c05421, 0x00000744, 0x00690050 }, + { 0x00400009, 0x26d05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26a00231, 0x00cf06c3, 0x00000000 }, + { 0x00800001, 0x26a80231, 0x00b10723, 0x00000000 }, + { 0x00800042, 0x25c04629, 0x00b106a8, 0x00b106a9 }, + { 0x00800040, 0x24003e2c, 0x00b106a3, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a2, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a1, 0x00010001 }, + { 0x00800008, 0x26c03d8d, 0x008d0400, 0x00020002 }, + { 0x00600001, 0x26a00231, 0x00ae06c0, 0x00000000 }, + { 0x00600001, 0x46a60231, 0x00ae06cc, 0x00000000 }, + { 0x00600001, 0x46a70231, 0x00ae05c0, 0x00000000 }, + { 0x00200040, 0x220c3eac, 0x00450036, 0x06a006a0 }, + { 0x80800040, 0xd0603631, 0x01ee9800, 0x00b18860 }, + { 0x80800040, 0xd0403631, 0x01ee9802, 0x00b18840 }, + { 0x80800040, 0xd0203631, 0x01ee9804, 0x00b18820 }, + { 0x80800040, 0xd0003631, 0x01ee9806, 0x00b18800 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00400009, 0x26c05421, 0x00000744, 0x00690050 }, + { 0x00400009, 0x26d05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26a00231, 0x00ab06c3, 0x00000000 }, + { 0x00800001, 0x26a80231, 0x00b10723, 0x00000000 }, + { 0x00800040, 0x24003e2c, 0x00b106a2, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a1, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b106a0, 0x00010001 }, + { 0x00800008, 0x25c03d89, 0x008d0400, 0x00020002 }, + { 0x00800042, 0x26a0462d, 0x00b106a0, 0x00b106a1 }, + { 0x00600001, 0x46a10231, 0x00ae05c0, 0x00000000 }, + { 0x00600001, 0x26b00231, 0x00ae05d0, 0x00000000 }, + { 0x00200040, 0x220c3eac, 0x00450056, 0x06a006a0 }, + { 0x80800040, 0xd0603631, 0x01ed9800, 0x00b18860 }, + { 0x80800040, 0xd0403631, 0x01ed9804, 0x00b18840 }, + { 0x80800040, 0xd0203631, 0x01ed9808, 0x00b18820 }, + { 0x80800040, 0xd0003631, 0x01ed980c, 0x00b18800 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00800042, 0x25c04629, 0x00b10724, 0x00b10725 }, + { 0x00800040, 0x24003e2c, 0x00b10726, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10725, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10724, 0x00010001 }, + { 0x00800008, 0x25e03d89, 0x00240400, 0x00020002 }, + { 0x80800040, 0xd0003531, 0x00ad05c0, 0x00b18800 }, + { 0x80800040, 0xd0203531, 0x00ad05c2, 0x00b18820 }, + { 0x80800040, 0xd0403531, 0x00ad05c4, 0x00b18840 }, + { 0x80800040, 0xd0603531, 0x00ad05c6, 0x00b18860 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00600001, 0x27480231, 0x00000747, 0x00000000 }, + { 0x00800042, 0x25c04629, 0x00b10740, 0x00b10741 }, + { 0x00800040, 0x24003e2c, 0x00b10742, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10741, 0x00020002 }, + { 0x00800048, 0x24003e2c, 0x00b10740, 0x00010001 }, + { 0x00800008, 0x25e03d89, 0x00240400, 0x00020002 }, + { 0x00800001, 0x45c10231, 0x00d205e0, 0x00000000 }, + { 0x80800040, 0xd0003631, 0x004d05c0, 0x00b18800 }, + { 0x80800040, 0xd0203631, 0x004d05c4, 0x00b18820 }, + { 0x80800040, 0xd0403631, 0x004d05c8, 0x00b18840 }, + { 0x80800040, 0xd0603631, 0x004d05cc, 0x00b18860 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00200801, 0x27c001a1, 0x004507fa, 0x00000000 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x0ff04000 }, + { 0x00800001, 0x20400232, 0x00cd0400, 0x00000000 }, + { 0x00800001, 0x20500232, 0x00cd0408, 0x00000000 }, + { 0x00800001, 0x20600232, 0x00cd0410, 0x00000000 }, + { 0x00800001, 0x20700232, 0x00cd0418, 0x00000000 }, + { 0x00800001, 0x20800232, 0x00cd0440, 0x00000000 }, + { 0x00800001, 0x20900232, 0x00cd0448, 0x00000000 }, + { 0x00800001, 0x20a00232, 0x00cd0450, 0x00000000 }, + { 0x00800001, 0x20b00232, 0x00cd0458, 0x00000000 }, + { 0x00800001, 0x20c00232, 0x00cd0480, 0x00000000 }, + { 0x00800001, 0x20d00232, 0x00cd0488, 0x00000000 }, + { 0x00800001, 0x20e00232, 0x00cd0490, 0x00000000 }, + { 0x00800001, 0x20f00232, 0x00cd0498, 0x00000000 }, + { 0x00800001, 0x21000232, 0x00cd04c0, 0x00000000 }, + { 0x00800001, 0x21100232, 0x00cd04c8, 0x00000000 }, + { 0x00800001, 0x21200232, 0x00cd04d0, 0x00000000 }, + { 0x00800001, 0x21300232, 0x00cd04d8, 0x00000000 }, + { 0x01600031, 0x27800021, 0x508d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00600001, 0x27c00021, 0x008d0000, 0x00000000 }, + { 0x00200009, 0x27fa3e29, 0x00450064, 0x00040004 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0xfffffd34 }, + { 0x00000001, 0x220401ec, 0x00000000, 0x00800080 }, + { 0x00000001, 0x220601ec, 0x00000000, 0x04000400 }, + { 0x00200008, 0x27006e2d, 0x0000006c, 0x00000040 }, + { 0x02600005, 0x20003e20, 0x0000006c, 0x00040004 }, + { 0x00780001, 0x26340231, 0x00000633, 0x00000000 }, + { 0x00800001, 0x27200231, 0x00b10620, 0x00000000 }, + { 0x00600001, 0x27400231, 0x00cf0643, 0x00000000 }, + { 0x00400008, 0x26806e2d, 0x00240070, 0x00004040 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000006a }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00000006, 0x27003dad, 0x00000700, 0x00010001 }, + { 0x00800001, 0x27200231, 0x00b10628, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00808c26, 0x00000000 }, + { 0x00400001, 0x27440231, 0x00808c66, 0x00000000 }, + { 0x00400008, 0x26806e2d, 0x00240072, 0x00004040 }, + { 0x00000040, 0x22063d8c, 0x00000206, 0x00800080 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000058 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00000006, 0x27003dad, 0x00000702, 0x00020002 }, + { 0x00800001, 0xaf800231, 0x00d28f80, 0x00000000 }, + { 0x00800001, 0xaf900231, 0x00d28fa0, 0x00000000 }, + { 0x00800001, 0xafc00231, 0x00d28fc0, 0x00000000 }, + { 0x00800001, 0xafd00231, 0x00d28fe0, 0x00000000 }, + { 0x00800001, 0xafa00231, 0x00d28c00, 0x00000000 }, + { 0x00800001, 0xafb00231, 0x00d28c20, 0x00000000 }, + { 0x00800001, 0xafe00231, 0x00d28c40, 0x00000000 }, + { 0x00800001, 0xaff00231, 0x00d28c60, 0x00000000 }, + { 0x00400001, 0x27200231, 0x0069065c, 0x00000000 }, + { 0x00600001, 0x27240231, 0x00a98fcc, 0x00000000 }, + { 0x00600001, 0x272c0231, 0x00a98fec, 0x00000000 }, + { 0x00600001, 0x27340231, 0x00008fff, 0x00000000 }, + { 0x00800001, 0x2620012d, 0x00b10720, 0x00000000 }, + { 0x00600001, 0x27400231, 0x00cf0663, 0x00000000 }, + { 0x00400008, 0x26806e2d, 0x00240074, 0x00004040 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000032 }, + { 0x00000040, 0x22043d8c, 0x00000204, 0x00800080 }, + { 0x00000006, 0x27003dad, 0x00000700, 0x00010001 }, + { 0x00800001, 0x27200231, 0x00b10628, 0x00000000 }, + { 0x00600001, 0x27300231, 0x008d0638, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00808c26, 0x00000000 }, + { 0x00400001, 0x27440231, 0x00808c66, 0x00000000 }, + { 0x00400008, 0x26806e2d, 0x00240076, 0x00004040 }, + { 0x00000040, 0x22063d8c, 0x00000206, 0x00800080 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x00800001, 0xaf800231, 0x00d28f80, 0x00000000 }, + { 0x00800001, 0xaf900231, 0x00d28fa0, 0x00000000 }, + { 0x00800001, 0xafc00231, 0x00d28fc0, 0x00000000 }, + { 0x00800001, 0xafd00231, 0x00d28fe0, 0x00000000 }, + { 0x00800001, 0xafa00231, 0x00d28c00, 0x00000000 }, + { 0x00800001, 0xafb00231, 0x00d28c20, 0x00000000 }, + { 0x00800001, 0xafe00231, 0x00d28c40, 0x00000000 }, + { 0x00800001, 0xaff00231, 0x00d28c60, 0x00000000 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x000000e0 }, + { 0x00000040, 0x27e01c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001c00, 0x00001400, 0xfffffcce }, + { 0x00000001, 0x27800021, 0x00000780, 0x00000000 }, + { 0x00000001, 0x27a00021, 0x000007a0, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00600001, 0x27800231, 0x008d0740, 0x00000000 }, + { 0x00400005, 0x22083dac, 0x00690680, 0x000f000f }, + { 0x00400040, 0x26a04625, 0x01e09020, 0x00690058 }, + { 0x00000001, 0x26d001ad, 0x00000700, 0x00000000 }, + { 0x00000040, 0x27e41c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a0 }, + { 0x80600040, 0xcc0035b1, 0x00898800, 0x008d0760 }, + { 0x80600040, 0xcc1035b1, 0x00898820, 0x008d0770 }, + { 0x00800001, 0x27200231, 0x008d0724, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00808c06, 0x00000000 }, + { 0x00000006, 0x26d03dad, 0x00000700, 0x00010001 }, + { 0x00000040, 0x27e41c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a4 }, + { 0x80600040, 0xcc2035b1, 0x00898808, 0x008d0760 }, + { 0x80600040, 0xcc3035b1, 0x00898828, 0x008d0770 }, + { 0x00000001, 0x27230231, 0x00000783, 0x00000000 }, + { 0x00400001, 0x27240231, 0x008a8c18, 0x00000000 }, + { 0x00400001, 0x27280231, 0x008a8c38, 0x00000000 }, + { 0x00400001, 0x272c0231, 0x00008c3e, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00690784, 0x00000000 }, + { 0x00000006, 0x26d03dad, 0x00000700, 0x00020002 }, + { 0x00000040, 0x27e41c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006a8 }, + { 0x80600040, 0xcc4035b1, 0x00898840, 0x008d0760 }, + { 0x80600040, 0xcc5035b1, 0x00898860, 0x008d0770 }, + { 0x00800001, 0x27200231, 0x008d0724, 0x00000000 }, + { 0x00600001, 0x27280231, 0x00000727, 0x00000000 }, + { 0x00400001, 0x27400231, 0x00808c46, 0x00000000 }, + { 0x00000006, 0x26d03dad, 0x00000700, 0x00030003 }, + { 0x00000040, 0x27e41c01, 0x00001400, 0x00000020 }, + { 0x00000220, 0x34001400, 0x00001400, 0x000006ac }, + { 0x80600040, 0xcc6035b1, 0x00898848, 0x008d0760 }, + { 0x80600040, 0xcc7035b1, 0x00898868, 0x008d0770 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00800001, 0x2760022d, 0x00090724, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00800001, 0x2760022d, 0x00280740, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x02802005, 0x20003da0, 0x000006d0, 0x00020002 }, + { 0x02802005, 0x20003da0, 0x020006d0, 0x00010001 }, + { 0x009a0001, 0x27200169, 0x00000000, 0x80808080 }, + { 0x00780001, 0x27400231, 0x028d0724, 0x00000000 }, + { 0x00780001, 0x27240231, 0x008d0740, 0x00000000 }, + { 0x00400040, 0x25e04629, 0x00690724, 0x00690740 }, + { 0x00200040, 0x25e02529, 0x004505e0, 0x004505e4 }, + { 0x00800040, 0x2400252c, 0x000005e0, 0x000005e2 }, + { 0x00800040, 0x24003d8c, 0x008d0400, 0x00040004 }, + { 0x00800008, 0x27603d8d, 0x008d0400, 0x00030003 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00600001, 0x26c00231, 0x008d0724, 0x00000000 }, + { 0x00400001, 0x26c80231, 0x0069072b, 0x00000000 }, + { 0x00600040, 0x24003e2c, 0x008d06c2, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c1, 0x00020002 }, + { 0x00600048, 0x25e03e29, 0x008d06c0, 0x00010001 }, + { 0x00800008, 0x27603d2d, 0x002905e0, 0x00020002 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00400009, 0x26e05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26c40231, 0x008d0723, 0x00000000 }, + { 0x00400001, 0x26c00231, 0x00ab06e3, 0x00000000 }, + { 0x00600040, 0x24003e2c, 0x008d06c2, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c1, 0x00020002 }, + { 0x00600048, 0x26c03e2d, 0x008d06c0, 0x00010001 }, + { 0x00400040, 0x22083eac, 0x00690054, 0x06c006c0 }, + { 0x00800008, 0x27603dad, 0x01e99000, 0x00020002 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00400009, 0x26e05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26c40231, 0x008d0723, 0x00000000 }, + { 0x00400001, 0x26c00231, 0x00ab06e3, 0x00000000 }, + { 0x00600042, 0x25c04629, 0x008d06c4, 0x008d06c5 }, + { 0x00600040, 0x24003e2c, 0x008d06c3, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c2, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c1, 0x00010001 }, + { 0x00600008, 0x26c03d8d, 0x008d0400, 0x00020002 }, + { 0x00400001, 0x46c401ad, 0x006906c4, 0x00000000 }, + { 0x00400001, 0x46c6012d, 0x006905c0, 0x00000000 }, + { 0x00400040, 0x22083eac, 0x00690054, 0x06c006c0 }, + { 0x00800001, 0x276001ad, 0x01ea9000, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00400009, 0x26e05421, 0x00000740, 0x00690050 }, + { 0x00600001, 0x26c40231, 0x008d0723, 0x00000000 }, + { 0x00400001, 0x26c00231, 0x00ab06e3, 0x00000000 }, + { 0x00600042, 0x25c04629, 0x008d06c0, 0x008d06c1 }, + { 0x00600040, 0x24003e2c, 0x008d06c2, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d06c1, 0x00020002 }, + { 0x00600048, 0x26e03e2d, 0x008d06c0, 0x00010001 }, + { 0x00400008, 0x46c23dad, 0x006906e0, 0x00020002 }, + { 0x00200008, 0x26d03dad, 0x004506e8, 0x00020002 }, + { 0x00400001, 0x46c0012d, 0x006905c0, 0x00000000 }, + { 0x00400009, 0x22083eac, 0x00690054, 0x00010001 }, + { 0x00400040, 0x22083d8c, 0x00690208, 0x06c006c0 }, + { 0x00800001, 0x276001ad, 0x01e99000, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00600042, 0x45c04629, 0x008d0724, 0x008d0725 }, + { 0x00600040, 0x24003e2c, 0x008d0726, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d0725, 0x00020002 }, + { 0x00600048, 0x25e03e29, 0x008d0724, 0x00010001 }, + { 0x00600008, 0x45c23d29, 0x008d05e0, 0x00020002 }, + { 0x00800001, 0x2760012d, 0x002a05c0, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00600001, 0x27440231, 0x00000743, 0x00000000 }, + { 0x00600042, 0x45c04629, 0x008d0740, 0x008d0741 }, + { 0x00600040, 0x24003e2c, 0x008d0742, 0x00020002 }, + { 0x00600048, 0x24003e2c, 0x008d0741, 0x00020002 }, + { 0x00600048, 0x25e03e29, 0x008d0740, 0x00010001 }, + { 0x00600008, 0x45c23d29, 0x008d05e0, 0x00020002 }, + { 0x00800001, 0x2760012d, 0x004905c0, 0x00000000 }, + { 0x00000001, 0x34000020, 0x000007e4, 0x00000000 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00200801, 0x27c001a1, 0x004507fa, 0x00000000 }, + { 0x00000040, 0x22000c00, 0x00000200, 0x0ff04000 }, + { 0x00800001, 0x20400232, 0x00a90400, 0x00000000 }, + { 0x00800001, 0x20500232, 0x00a90404, 0x00000000 }, + { 0x00800001, 0x20600232, 0x00a90408, 0x00000000 }, + { 0x00800001, 0x20700232, 0x00a9040c, 0x00000000 }, + { 0x00800001, 0x20800232, 0x00a90440, 0x00000000 }, + { 0x00800001, 0x20900232, 0x00a90444, 0x00000000 }, + { 0x00800001, 0x20a00232, 0x00a90448, 0x00000000 }, + { 0x00800001, 0x20b00232, 0x00a9044c, 0x00000000 }, + { 0x00800001, 0x20c00232, 0x00a90480, 0x00000000 }, + { 0x00800001, 0x20d00232, 0x00a90484, 0x00000000 }, + { 0x00800001, 0x20e00232, 0x00a90488, 0x00000000 }, + { 0x00800001, 0x20f00232, 0x00a9048c, 0x00000000 }, + { 0x00800001, 0x21000232, 0x00a904c0, 0x00000000 }, + { 0x00800001, 0x21100232, 0x00a904c4, 0x00000000 }, + { 0x00800001, 0x21200232, 0x00a904c8, 0x00000000 }, + { 0x00800001, 0x21300232, 0x00a904cc, 0x00000000 }, + { 0x01600031, 0x27800021, 0x508d07c0, 0x00000200 }, + { 0x00000001, 0x34000020, 0x000007e0, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00600001, 0x27c00021, 0x008d0000, 0x00000000 }, + { 0x00200009, 0x27fa3e29, 0x00450064, 0x00040004 }, + { 0x00000401, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00200809, 0x27c03e21, 0x00450064, 0x00040004 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x1218a000 }, + { 0x00a02401, 0x20400232, 0x00b10080, 0x00000000 }, + { 0x00a02801, 0x20500232, 0x00b10090, 0x00000000 }, + { 0x00a02401, 0x20800232, 0x00b100c0, 0x00000000 }, + { 0x00a02801, 0x20900232, 0x00b100d0, 0x00000000 }, + { 0x00a02401, 0x20c00232, 0x00b10100, 0x00000000 }, + { 0x00a02801, 0x20d00232, 0x00b10110, 0x00000000 }, + { 0x00a02401, 0x21000232, 0x00b10140, 0x00000000 }, + { 0x00a02801, 0x21100232, 0x00b10150, 0x00000000 }, + { 0x01600031, 0x27800021, 0x508d07c0, 0x00000200 }, + { 0x00000401, 0x20280062, 0x00000000, 0x0007000f }, + { 0x0000080c, 0x20243c22, 0x000007c4, 0x00010001 }, + { 0x00000040, 0x22001c00, 0x00000200, 0xf8000001 }, + { 0x00800001, 0x40400232, 0x00b10180, 0x00000000 }, + { 0x00800001, 0x40410232, 0x00b101c0, 0x00000000 }, + { 0x00800001, 0x40600232, 0x00b10190, 0x00000000 }, + { 0x00800001, 0x40610232, 0x00b101d0, 0x00000000 }, + { 0x00800001, 0x40800232, 0x00b101a0, 0x00000000 }, + { 0x00800001, 0x40810232, 0x00b101e0, 0x00000000 }, + { 0x00800001, 0x40a00232, 0x00b101b0, 0x00000000 }, + { 0x00800001, 0x40a10232, 0x00b101f0, 0x00000000 }, + { 0x01600031, 0x27a00001, 0x508d0000, 0x00000200 }, + { 0x00000001, 0x27800021, 0x00000780, 0x00000000 }, + { 0x00000001, 0x27a00021, 0x000007a0, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00600001, 0x20400021, 0x008d0000, 0x00000000 }, + { 0x00000005, 0x20203e2d, 0x00000061, 0x001f001f }, + { 0x00200009, 0x20643e2d, 0x00450064, 0x00040004 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00000000 }, + { 0x01000010, 0x20003dac, 0x00000020, 0x00160016 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x01200010, 0x20003e2c, 0x0200006c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000024 }, + { 0x00800009, 0x25403dad, 0x00050064, 0x00020002 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00180018 }, + { 0x00200040, 0x45483dad, 0x00660548, 0x00100010 }, + { 0x00200040, 0x45523dad, 0x00660552, 0x00100010 }, + { 0x00400040, 0x25583dad, 0x00690558, 0x00100010 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0x00200020 }, + { 0x00800040, 0x20a035ad, 0x00b100a0, 0x00b10540 }, + { 0x00600040, 0x45423dad, 0x00ae0542, 0x00200020 }, + { 0x00800040, 0x20e035ad, 0x00b100e0, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0xffe0ffe0 }, + { 0x00800040, 0x20c035ad, 0x00b100c0, 0x00b10540 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000014 }, + { 0x06600010, 0x20003dac, 0x02000020, 0x00030003 }, + { 0x00000001, 0x206c0171, 0x00000000, 0x00000000 }, + { 0x00610001, 0x208000a5, 0x02050080, 0x00000000 }, + { 0x00610001, 0x212000a5, 0x02090100, 0x00000000 }, + { 0x00410001, 0x211000a5, 0x02690100, 0x00000000 }, + { 0x00200009, 0x25403dad, 0x00450064, 0x00020002 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00050540 }, + { 0x00200040, 0x40883dad, 0x00660088, 0x00200020 }, + { 0x00200040, 0x40923dad, 0x00660092, 0x00200020 }, + { 0x00400040, 0x20983dad, 0x00690098, 0x00200020 }, + { 0x00200401, 0x22080060, 0x00000000, 0x03400140 }, + { 0x00000c01, 0x220c0060, 0x00000000, 0x04400080 }, + { 0x00000801, 0x22040060, 0x00000000, 0x01000070 }, + { 0x00000001, 0x20200169, 0x00000000, 0x00000000 }, + { 0x0000000c, 0x2458262d, 0x0000006d, 0x00000020 }, + { 0x00000001, 0x210e0169, 0x00000000, 0x00010001 }, + { 0x00000005, 0x24583dad, 0x00000458, 0x00030003 }, + { 0x01000010, 0x200035ac, 0x0000010e, 0x00000458 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x000002fe }, + { 0x01000010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00010401, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00010805, 0x245a3e2d, 0x02008800, 0x007f007f }, + { 0x00110401, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00110805, 0x245a3e2d, 0x02008804, 0x007f007f }, + { 0x0000000c, 0x211e362d, 0x0000006c, 0x00000020 }, + { 0x00000040, 0x20780d21, 0x0000045a, 0x0208a002 }, + { 0x02000005, 0x20003dac, 0x0200011e, 0x00030003 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000001be }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000012 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00700000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c000c }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x0007000c }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00110040, 0x20441da5, 0x02000442, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c0007 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00070007 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x01600006, 0x20003dac, 0x0000044a, 0x00000000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00010040, 0x22000c20, 0x00000078, 0x00200010 }, + { 0x00110040, 0x22000c20, 0x00000078, 0x00300010 }, + { 0x00010001, 0x20480061, 0x00000000, 0x00030009 }, + { 0x00110001, 0x20480061, 0x00000000, 0x00040009 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001a }, + { 0x00600001, 0x26220231, 0x008d05c0, 0x00000000 }, + { 0x00600401, 0x26120231, 0x008d05b8, 0x00000000 }, + { 0x00600801, 0x26020231, 0x008d05b0, 0x00000000 }, + { 0x00600401, 0x25f20231, 0x008d05a8, 0x00000000 }, + { 0x00600801, 0x25e20231, 0x008d05a0, 0x00000000 }, + { 0x00600401, 0x25d20231, 0x008d0598, 0x00000000 }, + { 0x00600801, 0x25c20231, 0x008d0590, 0x00000000 }, + { 0x00600401, 0x25b20231, 0x008d0588, 0x00000000 }, + { 0x00600801, 0x25a20231, 0x008d0580, 0x00000000 }, + { 0x00600401, 0x25920231, 0x008d0578, 0x00000000 }, + { 0x00600801, 0x25820231, 0x008d0570, 0x00000000 }, + { 0x00600001, 0x25720231, 0x008d0568, 0x00000000 }, + { 0x00600001, 0x25620231, 0x008d0560, 0x00000000 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00810001, 0xbc000229, 0x02ad8000, 0x00000000 }, + { 0x00810001, 0xbc200229, 0x02ad8020, 0x00000000 }, + { 0x00810001, 0xbc400229, 0x02ad8040, 0x00000000 }, + { 0x00810001, 0xbc600229, 0x02ad8060, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000012a }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000066 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000064 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04a00480 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00a02040, 0x2400462c, 0x00ad8000, 0x00ad8005 }, + { 0x00a02048, 0x24003e2c, 0x00ad8001, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8002, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8003, 0x00140014 }, + { 0x00a02048, 0xb8003e2d, 0x00ad8004, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8040, 0x00ad8045 }, + { 0x00a02048, 0x24003e2c, 0x00ad8041, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8042, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8043, 0x00140014 }, + { 0x00a02048, 0xb8403e2d, 0x00ad8044, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8080, 0x00ad8085 }, + { 0x00a02048, 0x24003e2c, 0x00ad8081, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8082, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8083, 0x00140014 }, + { 0x00a02048, 0xb8803e2d, 0x00ad8084, 0xfffbfffb }, + { 0x00600040, 0x2400462c, 0x008d80c0, 0x008d80c5 }, + { 0x00600048, 0x24003e2c, 0x008d80c1, 0xfffbfffb }, + { 0x00600048, 0x24003e2c, 0x008d80c2, 0x00140014 }, + { 0x00600048, 0x24003e2c, 0x008d80c3, 0x00140014 }, + { 0x00600048, 0xb8c03e2d, 0x008d80c4, 0xfffbfffb }, + { 0x00000401, 0x22000060, 0x00000000, 0x04a00480 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04a00490 }, + { 0x00010802, 0x220c2d28, 0x00000454, 0x04800480 }, + { 0x00a02040, 0x24003dac, 0x00b18000, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8800, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8820, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18020, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8820, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8840, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18040, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x000a000a }, + { 0x00a02040, 0x24003dac, 0x00b18040, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18060, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8860, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8880, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18080, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8880, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed88a0, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x000000b4 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000004c }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000004a }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010002, 0x220c2d28, 0x02000454, 0x04800480 }, + { 0x00800040, 0x24003e2c, 0x01ed8000, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8001, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8021, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8002, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8022, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8003, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8023, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8004, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8024, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8005, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8025, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00800040, 0x24003e2c, 0x01ed8040, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8060, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8041, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8061, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8042, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8062, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8043, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8063, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8044, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8064, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8045, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8065, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000005c }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000062 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000042 }, + { 0x00200401, 0x22000128, 0x0066044c, 0x00000000 }, + { 0x00200c01, 0x22040128, 0x0045044e, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x03400010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220c0168, 0x02000000, 0x04800480 }, + { 0x00a02040, 0x24003e2c, 0x00ad83e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8be0, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8800, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8800, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8820, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00a02040, 0x24003e2c, 0x00ad8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8040, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8840, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8860, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8060, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8860, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8880, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000001c }, + { 0x00200401, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x01200010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01200010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00210040, 0x22002d08, 0x00450200, 0x00010001 }, + { 0x00210040, 0x22002d08, 0x02450200, 0x00100010 }, + { 0x00800001, 0xb8000229, 0x01ed8000, 0x00000000 }, + { 0x00800001, 0xb8200229, 0x01ed8020, 0x00000000 }, + { 0x00800001, 0xb8400229, 0x01ed8040, 0x00000000 }, + { 0x00800001, 0xb8600229, 0x01ed8060, 0x00000000 }, + { 0x80800042, 0xd8004631, 0x00d29800, 0x00d20480 }, + { 0x80800042, 0xd8204631, 0x00d29820, 0x00d204a0 }, + { 0x80800042, 0xd8404631, 0x00d29840, 0x00d204c0 }, + { 0x80800042, 0xd8604631, 0x00d29860, 0x00d204e0 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000c01, 0x220201e8, 0x00000000, 0x07100710 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00200040, 0x244c3d09, 0x00450200, 0x00100010 }, + { 0x00800041, 0x24002628, 0x00ad8000, 0x00000540 }, + { 0x00800041, 0x24202628, 0x00ad8020, 0x00000540 }, + { 0x00800048, 0x24002628, 0x00ad8002, 0x00000542 }, + { 0x00800048, 0x24202628, 0x00ad8022, 0x00000542 }, + { 0x00200001, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00800048, 0x24002628, 0x01ed8000, 0x00000544 }, + { 0x00800048, 0x24202628, 0x01ed8020, 0x00000544 }, + { 0x00800048, 0xbc002629, 0x01ed8002, 0x00000546 }, + { 0x00801048, 0xbc202629, 0x01ed8022, 0x00000546 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000126 }, + { 0x00000001, 0x20220169, 0x00000000, 0x00040004 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00500000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x00080008 }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00030008 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0x22000c20, 0x00000078, 0x00100010 }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00000001, 0x20480061, 0x00000000, 0x00020005 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00410001, 0xbc000229, 0x02698000, 0x00000000 }, + { 0x00410001, 0xbc100229, 0x02698010, 0x00000000 }, + { 0x00410001, 0xbc200229, 0x02698020, 0x00000000 }, + { 0x00410001, 0xbc300229, 0x02698030, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000000be }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000044 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000042 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04d004c0 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00802040, 0x2400462c, 0x00a98000, 0x00a98005 }, + { 0x00802048, 0x24003e2c, 0x00a98001, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98002, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98003, 0x00140014 }, + { 0x00802048, 0xb8003e2d, 0x00a98004, 0xfffbfffb }, + { 0x00802040, 0x2400462c, 0x00a98040, 0x00a98045 }, + { 0x00802048, 0x24003e2c, 0x00a98041, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98042, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98043, 0x00140014 }, + { 0x00802048, 0xb8203e2d, 0x00a98044, 0xfffbfffb }, + { 0x00400040, 0x2400462c, 0x00698080, 0x00698085 }, + { 0x00400048, 0x24003e2c, 0x00698081, 0xfffbfffb }, + { 0x00400048, 0x24003e2c, 0x00698082, 0x00140014 }, + { 0x00400048, 0x24003e2c, 0x00698083, 0x00140014 }, + { 0x00400048, 0xb8403e2d, 0x00698084, 0xfffbfffb }, + { 0x00000401, 0x220c0060, 0x00000000, 0x04e004d0 }, + { 0x00000c01, 0x22000060, 0x00000000, 0x04d004c8 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04e004d8 }, + { 0x00000801, 0x22080060, 0x00000000, 0x04f004e8 }, + { 0x00800040, 0x24003dac, 0x00b104c0, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x00b104e0, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed9800, 0x00140014 }, + { 0x00010001, 0x220c0168, 0x00000000, 0x05400540 }, + { 0x00110001, 0x220c0168, 0x00000000, 0x04c004c0 }, + { 0x00800048, 0x24003dac, 0x01e98000, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01e98020, 0x00010001 }, + { 0x00800048, 0x24003dac, 0x01e98800, 0x00140014 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000060 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000022 }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010001, 0x220e0168, 0x02000000, 0x05400540 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x00800040, 0x24003e2c, 0x01e98000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98001, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98002, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98003, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98004, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98005, 0x00010001 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x03600010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x00400401, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x00400840, 0x22083d2c, 0x0069044c, 0x00100010 }, + { 0x00610040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00800040, 0x24003e2c, 0x01e983e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e993e0, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e99000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e99020, 0x00010001 }, + { 0x00000001, 0x220e0168, 0x00000000, 0x05400540 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x00400001, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x01400010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01400010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00010001 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00100010 }, + { 0x00800001, 0x25400229, 0x01e98000, 0x00000000 }, + { 0x80800042, 0x45404631, 0x00d20540, 0x00d204c0 }, + { 0x00000001, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x00400001, 0xd8000231, 0x008a0540, 0x00000000 }, + { 0x00400001, 0xd8100231, 0x008a0548, 0x00000000 }, + { 0x00400001, 0xd8200231, 0x008a0550, 0x00000000 }, + { 0x00400001, 0xd8300231, 0x008a0558, 0x00000000 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00600041, 0x24002628, 0x00898000, 0x00000540 }, + { 0x00600048, 0x24002628, 0x00898002, 0x00000542 }, + { 0x00600048, 0x24002628, 0x00898008, 0x00000544 }, + { 0x00600048, 0x2540262d, 0x0089800a, 0x00000546 }, + { 0x00400401, 0xbc0001a9, 0x00690540, 0x00000000 }, + { 0x00400801, 0xbc1001a9, 0x00690548, 0x00000000 }, + { 0x01000010, 0x20003d2c, 0x00000022, 0x00030003 }, + { 0x01000040, 0x20223d29, 0x02000022, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00080008 }, + { 0x00110040, 0x24540c21, 0x00000454, 0x00080008 }, + { 0x00010040, 0x24540c21, 0x00000454, 0x00180038 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffee6 }, + { 0x01800010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0xffe0ffe0 }, + { 0x00010001, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00110001, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00000001, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00800040, 0x24003dac, 0x00b19c00, 0x00200020 }, + { 0x00800040, 0x24203dac, 0x00b19c20, 0x00200020 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00060006 }, + { 0x8080000c, 0xdc203d91, 0x00b10420, 0x00060006 }, + { 0x02000040, 0x210e3d29, 0x0200010e, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00040004 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffcf8 }, + { 0x01000005, 0x25643e2d, 0x00000060, 0x00c000c0 }, + { 0x01000010, 0x20003dac, 0x02000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000086 }, + { 0x00a02001, 0x46400231, 0x00d20780, 0x00000000 }, + { 0x00a02001, 0x46800231, 0x00d207c0, 0x00000000 }, + { 0x00a02001, 0x46c00231, 0x00d20480, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000007e }, + { 0x80a02042, 0x46404631, 0x00d20640, 0x00d20780 }, + { 0x80a02042, 0x46804631, 0x00d20680, 0x00d207c0 }, + { 0x80a02042, 0x46c04631, 0x00d206c0, 0x00d20480 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000076 }, + { 0x01000010, 0x20003dac, 0x02000564, 0x00800080 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffffe4 }, + { 0x00200401, 0x256801ed, 0x00000000, 0x00200020 }, + { 0x00200801, 0x256c01ed, 0x00000000, 0x00060006 }, + { 0x00400001, 0x258001ed, 0x00000000, 0x00000000 }, + { 0x00600001, 0x45a001ad, 0x00058c00, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000028 }, + { 0x02600005, 0x20006e28, 0x02008c0c, 0x88848421 }, + { 0x0220000c, 0x25663dad, 0x00000458, 0x00010001 }, + { 0x0000000c, 0x25643dad, 0x00000564, 0x00060006 }, + { 0x00310001, 0x25c401ad, 0x00000458, 0x00000000 }, + { 0x00210001, 0x25c00061, 0x00000000, 0x00010001 }, + { 0x00310040, 0x25c03dad, 0x004545c4, 0x00010001 }, + { 0x00610001, 0x25a00061, 0x02000000, 0x00000080 }, + { 0x00710001, 0x45a002ad, 0x02ae8c00, 0x00000000 }, + { 0x00710001, 0x45a202ad, 0x02ae8c01, 0x00000000 }, + { 0x00800041, 0x25a035ad, 0x00b105a0, 0x000905c0 }, + { 0x01200010, 0x20003e2c, 0x0245006e, 0x00000000 }, + { 0x00310009, 0x244c45ad, 0x02000564, 0x0045006e }, + { 0x00210001, 0x244c01ed, 0x02000000, 0x00000000 }, + { 0x0031000c, 0x244c3dad, 0x0245044c, 0x00010001 }, + { 0x00200009, 0x256835ad, 0x0045044c, 0x00000566 }, + { 0x00210040, 0x256835ad, 0x02450568, 0x00000566 }, + { 0x00200040, 0x256c362d, 0x0045006e, 0x00000566 }, + { 0x00400040, 0x240035ac, 0x00ab05a2, 0x00ab05a6 }, + { 0x00400040, 0x2400358c, 0x00690400, 0x00000566 }, + { 0x0040000c, 0x2580358d, 0x00690400, 0x00000566 }, + { 0x00800041, 0x2400362c, 0x00d20640, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d20660, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d20780, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207a0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x464035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x466035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d20680, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d206a0, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d207c0, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207e0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x468035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x46a035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d206c0, 0x000705a8 }, + { 0x00800041, 0x2420362c, 0x00d206e0, 0x000705a8 }, + { 0x00800048, 0x2400362c, 0x00d20480, 0x000705ac }, + { 0x00800048, 0x2420362c, 0x00d204a0, 0x000705ac }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x0000056a }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x0000056a }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056e }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056e }, + { 0x80800040, 0x46c035b1, 0x00b104c0, 0x00050582 }, + { 0x80800040, 0x46e035b1, 0x00b104e0, 0x00050582 }, + { 0x01800005, 0x20003d2c, 0x02000020, 0x00020002 }, + { 0x80800040, 0xd00045b1, 0x00b19000, 0x00d20640 }, + { 0x80800040, 0xd02045b1, 0x00b19020, 0x00d20660 }, + { 0x80800040, 0xd04045b1, 0x00b19040, 0x00d20680 }, + { 0x80800040, 0xd06045b1, 0x00b19060, 0x00d206a0 }, + { 0x00000040, 0x22083d8c, 0x00000208, 0x00800080 }, + { 0x80400040, 0xd40045b1, 0x00699400, 0x00ab06c0 }, + { 0x80400040, 0xd48045b1, 0x00699480, 0x00ab06c2 }, + { 0x80400040, 0xd42045b1, 0x00699420, 0x00ab06e0 }, + { 0x80400040, 0xd4a045b1, 0x006994a0, 0x00ab06e2 }, + { 0x80400040, 0xd41045b1, 0x00699410, 0x00ab06d0 }, + { 0x80400040, 0xd49045b1, 0x00699490, 0x00ab06d2 }, + { 0x80400040, 0xd43045b1, 0x00699430, 0x00ab06f0 }, + { 0x80400040, 0xd4b045b1, 0x006994b0, 0x00ab06f2 }, + { 0x00000040, 0x220a3d8c, 0x0000020a, 0x00080008 }, + { 0x00110040, 0x220a3d8c, 0x0200020a, 0x00300030 }, + { 0x01000010, 0x20003d2c, 0x02000020, 0x00060006 }, + { 0x00000040, 0x20203d29, 0x00000020, 0x00020002 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00100001 }, + { 0x00000040, 0x220c358c, 0x0000020c, 0x0000003a }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffc36 }, + { 0x00000001, 0x220001ec, 0x00000000, 0x01400140 }, + { 0x00000001, 0x220201ec, 0x00000000, 0x01c001c0 }, + { 0x00800401, 0x20400236, 0x01ee8000, 0x00000000 }, + { 0x00800801, 0x20500236, 0x01ee8010, 0x00000000 }, + { 0x00800401, 0x20600236, 0x01ee8020, 0x00000000 }, + { 0x00800801, 0x20700236, 0x01ee8030, 0x00000000 }, + { 0x00800401, 0x20800236, 0x01ee8040, 0x00000000 }, + { 0x00800801, 0x20900236, 0x01ee8050, 0x00000000 }, + { 0x00800401, 0x20a00236, 0x01ee8060, 0x00000000 }, + { 0x00800801, 0x20b00236, 0x01ee8070, 0x00000000 }, + { 0x00800401, 0x20c00236, 0x01ee8100, 0x00000000 }, + { 0x00800801, 0x20d00236, 0x01ee8110, 0x00000000 }, + { 0x00800401, 0x20e00236, 0x01ee8120, 0x00000000 }, + { 0x00800801, 0x20f00236, 0x01ee8130, 0x00000000 }, + { 0x00800401, 0x21000236, 0x01ee8140, 0x00000000 }, + { 0x00800801, 0x21100236, 0x01ee8150, 0x00000000 }, + { 0x00800401, 0x21200236, 0x01ee8160, 0x00000000 }, + { 0x00800801, 0x21300236, 0x01ee8170, 0x00000000 }, + { 0x00200401, 0x204001a5, 0x00450064, 0x00000000 }, + { 0x00000801, 0x20480061, 0x00000000, 0x000f000f }, + { 0x01600031, 0x21401c21, 0x508d0040, 0x1218a000 }, + { 0x00000001, 0x220001ec, 0x00000000, 0x03400340 }, + { 0x00800401, 0x41600236, 0x00d28000, 0x00000000 }, + { 0x00800801, 0x41610236, 0x00d28080, 0x00000000 }, + { 0x00800401, 0x41800236, 0x00d28020, 0x00000000 }, + { 0x00800801, 0x41810236, 0x00d280a0, 0x00000000 }, + { 0x00800401, 0x41a00236, 0x00d28040, 0x00000000 }, + { 0x00800801, 0x41a10236, 0x00d280c0, 0x00000000 }, + { 0x00800401, 0x41c00236, 0x00d28060, 0x00000000 }, + { 0x00800801, 0x41c10236, 0x00d280e0, 0x00000000 }, + { 0x0000040c, 0x20443ca5, 0x00000044, 0x00010001 }, + { 0x00000801, 0x20480061, 0x00000000, 0x0007000f }, + { 0x0a600031, 0x21601c21, 0x508d0040, 0x0a18a001 }, + { 0x00000001, 0x21400021, 0x00000140, 0x00000000 }, + { 0x00000001, 0x21600021, 0x00000160, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00600001, 0x20400021, 0x008d0000, 0x00000000 }, + { 0x00000005, 0x20203e2d, 0x00000061, 0x001f001f }, + { 0x00200009, 0x20643e2d, 0x00450064, 0x00040004 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00000000 }, + { 0x01000010, 0x20003dac, 0x00000020, 0x00160016 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x01200010, 0x20003e2c, 0x0200006c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000024 }, + { 0x00800009, 0x25403dad, 0x00050064, 0x00020002 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00180018 }, + { 0x00200040, 0x45483dad, 0x00660548, 0x00100010 }, + { 0x00200040, 0x45523dad, 0x00660552, 0x00100010 }, + { 0x00400040, 0x25583dad, 0x00690558, 0x00100010 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0x00200020 }, + { 0x00800040, 0x20a035ad, 0x00b100a0, 0x00b10540 }, + { 0x00600040, 0x45423dad, 0x00ae0542, 0x00200020 }, + { 0x00800040, 0x20e035ad, 0x00b100e0, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0xffe0ffe0 }, + { 0x00800040, 0x20c035ad, 0x00b100c0, 0x00b10540 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000014 }, + { 0x06600010, 0x20003dac, 0x02000020, 0x00030003 }, + { 0x00000001, 0x206c0171, 0x00000000, 0x00000000 }, + { 0x00610001, 0x208000a5, 0x02050080, 0x00000000 }, + { 0x00610001, 0x212000a5, 0x02090100, 0x00000000 }, + { 0x00410001, 0x211000a5, 0x02690100, 0x00000000 }, + { 0x00200009, 0x25403dad, 0x00450064, 0x00020002 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00050540 }, + { 0x00200040, 0x40883dad, 0x00660088, 0x00200020 }, + { 0x00200040, 0x40923dad, 0x00660092, 0x00200020 }, + { 0x00400040, 0x20983dad, 0x00690098, 0x00200020 }, + { 0x00200401, 0x22080060, 0x00000000, 0x03400140 }, + { 0x00000c01, 0x220c0060, 0x00000000, 0x04400080 }, + { 0x00000801, 0x22040060, 0x00000000, 0x01000070 }, + { 0x00000005, 0x203e2e29, 0x00000063, 0x00010001 }, + { 0x00000001, 0x20200169, 0x00000000, 0x00000000 }, + { 0x0000000c, 0x2458262d, 0x0000006d, 0x00000020 }, + { 0x00000001, 0x210e0169, 0x00000000, 0x00010001 }, + { 0x00000005, 0x24583dad, 0x00000458, 0x00030003 }, + { 0x01000010, 0x200035ac, 0x0000010e, 0x00000458 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000316 }, + { 0x01000010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00010401, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00010805, 0x245a3e2d, 0x02008800, 0x007f007f }, + { 0x00110401, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00110805, 0x245a3e2d, 0x02008804, 0x007f007f }, + { 0x00010005, 0x245c3e2d, 0x02008800, 0x00800080 }, + { 0x00110005, 0x245c3e2d, 0x02008804, 0x00800080 }, + { 0x00000009, 0x245c3dad, 0x0000045c, 0x00010001 }, + { 0x0000000c, 0x211e362d, 0x0000006c, 0x00000020 }, + { 0x00000040, 0x240035ac, 0x0000003e, 0x0000045c }, + { 0x01000010, 0x20003d8c, 0x00210400, 0x00010001 }, + { 0x01000010, 0x20003d8c, 0x02210400, 0x01000100 }, + { 0x00000001, 0x203c01ed, 0x00000000, 0x00000000 }, + { 0x00010001, 0x203c01ed, 0x00000000, 0x00020002 }, + { 0x00010001, 0x203c01ed, 0x02000000, 0xfffefffe }, + { 0x00000040, 0x20780d21, 0x0000045a, 0x0208e602 }, + { 0x00000040, 0x20782421, 0x00000078, 0x0000045c }, + { 0x02000005, 0x20003dac, 0x0200011e, 0x00030003 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000001c0 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000012 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00700000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c000c }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x0007000c }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00110040, 0x20441da5, 0x02000442, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c0007 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00070007 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0xb80235ad, 0x00009802, 0x0000003c }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x01600006, 0x20003dac, 0x0000044a, 0x00000000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00010040, 0x22000c20, 0x00000078, 0x00200010 }, + { 0x00110040, 0x22000c20, 0x00000078, 0x00300010 }, + { 0x00010001, 0x20480061, 0x00000000, 0x00030009 }, + { 0x00110001, 0x20480061, 0x00000000, 0x00040009 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001a }, + { 0x00600001, 0x26220231, 0x008d05c0, 0x00000000 }, + { 0x00600401, 0x26120231, 0x008d05b8, 0x00000000 }, + { 0x00600801, 0x26020231, 0x008d05b0, 0x00000000 }, + { 0x00600401, 0x25f20231, 0x008d05a8, 0x00000000 }, + { 0x00600801, 0x25e20231, 0x008d05a0, 0x00000000 }, + { 0x00600401, 0x25d20231, 0x008d0598, 0x00000000 }, + { 0x00600801, 0x25c20231, 0x008d0590, 0x00000000 }, + { 0x00600401, 0x25b20231, 0x008d0588, 0x00000000 }, + { 0x00600801, 0x25a20231, 0x008d0580, 0x00000000 }, + { 0x00600401, 0x25920231, 0x008d0578, 0x00000000 }, + { 0x00600801, 0x25820231, 0x008d0570, 0x00000000 }, + { 0x00600001, 0x25720231, 0x008d0568, 0x00000000 }, + { 0x00600001, 0x25620231, 0x008d0560, 0x00000000 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00810001, 0xbc000229, 0x02ad8000, 0x00000000 }, + { 0x00810001, 0xbc200229, 0x02ad8020, 0x00000000 }, + { 0x00810001, 0xbc400229, 0x02ad8040, 0x00000000 }, + { 0x00810001, 0xbc600229, 0x02ad8060, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000012a }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000066 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000064 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04a00480 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00a02040, 0x2400462c, 0x00ad8000, 0x00ad8005 }, + { 0x00a02048, 0x24003e2c, 0x00ad8001, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8002, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8003, 0x00140014 }, + { 0x00a02048, 0xb8003e2d, 0x00ad8004, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8040, 0x00ad8045 }, + { 0x00a02048, 0x24003e2c, 0x00ad8041, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8042, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8043, 0x00140014 }, + { 0x00a02048, 0xb8403e2d, 0x00ad8044, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8080, 0x00ad8085 }, + { 0x00a02048, 0x24003e2c, 0x00ad8081, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8082, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8083, 0x00140014 }, + { 0x00a02048, 0xb8803e2d, 0x00ad8084, 0xfffbfffb }, + { 0x00600040, 0x2400462c, 0x008d80c0, 0x008d80c5 }, + { 0x00600048, 0x24003e2c, 0x008d80c1, 0xfffbfffb }, + { 0x00600048, 0x24003e2c, 0x008d80c2, 0x00140014 }, + { 0x00600048, 0x24003e2c, 0x008d80c3, 0x00140014 }, + { 0x00600048, 0xb8c03e2d, 0x008d80c4, 0xfffbfffb }, + { 0x00000401, 0x22000060, 0x00000000, 0x04a00480 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04a00490 }, + { 0x00010802, 0x220c2d28, 0x00000454, 0x04800480 }, + { 0x00a02040, 0x24003dac, 0x00b18000, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8800, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8820, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18020, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8820, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8840, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18040, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x000a000a }, + { 0x00a02040, 0x24003dac, 0x00b18040, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18060, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8860, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8880, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18080, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8880, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed88a0, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x000000b4 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000004c }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000004a }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010002, 0x220c2d28, 0x02000454, 0x04800480 }, + { 0x00800040, 0x24003e2c, 0x01ed8000, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8001, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8021, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8002, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8022, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8003, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8023, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8004, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8024, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8005, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8025, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00800040, 0x24003e2c, 0x01ed8040, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8060, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8041, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8061, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8042, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8062, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8043, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8063, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8044, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8064, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8045, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8065, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000005c }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000062 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000042 }, + { 0x00200401, 0x22000128, 0x0066044c, 0x00000000 }, + { 0x00200c01, 0x22040128, 0x0045044e, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x03400010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220c0168, 0x02000000, 0x04800480 }, + { 0x00a02040, 0x24003e2c, 0x00ad83e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8be0, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8800, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8800, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8820, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00a02040, 0x24003e2c, 0x00ad8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8040, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8840, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8860, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8060, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8860, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8880, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000001c }, + { 0x00200401, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x01200010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01200010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00210040, 0x22002d08, 0x00450200, 0x00010001 }, + { 0x00210040, 0x22002d08, 0x02450200, 0x00100010 }, + { 0x00800001, 0xb8000229, 0x01ed8000, 0x00000000 }, + { 0x00800001, 0xb8200229, 0x01ed8020, 0x00000000 }, + { 0x00800001, 0xb8400229, 0x01ed8040, 0x00000000 }, + { 0x00800001, 0xb8600229, 0x01ed8060, 0x00000000 }, + { 0x80800042, 0xd8004631, 0x00d29800, 0x00d20480 }, + { 0x80800042, 0xd8204631, 0x00d29820, 0x00d204a0 }, + { 0x80800042, 0xd8404631, 0x00d29840, 0x00d204c0 }, + { 0x80800042, 0xd8604631, 0x00d29860, 0x00d204e0 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000c01, 0x220201e8, 0x00000000, 0x07100710 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00200040, 0x244c3d09, 0x00450200, 0x00100010 }, + { 0x00800041, 0x24002628, 0x00ad8000, 0x00000540 }, + { 0x00800041, 0x24202628, 0x00ad8020, 0x00000540 }, + { 0x00800048, 0x24002628, 0x00ad8002, 0x00000542 }, + { 0x00800048, 0x24202628, 0x00ad8022, 0x00000542 }, + { 0x00200001, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00800048, 0x24002628, 0x01ed8000, 0x00000544 }, + { 0x00800048, 0x24202628, 0x01ed8020, 0x00000544 }, + { 0x00800048, 0xbc002629, 0x01ed8002, 0x00000546 }, + { 0x00801048, 0xbc202629, 0x01ed8022, 0x00000546 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000128 }, + { 0x00000001, 0x20220169, 0x00000000, 0x00040004 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00500000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x00080008 }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00030008 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0xb80235ad, 0x00009802, 0x0000003c }, + { 0x00000040, 0x22000c20, 0x00000078, 0x00100010 }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00000001, 0x20480061, 0x00000000, 0x00020005 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00410001, 0xbc000229, 0x02698000, 0x00000000 }, + { 0x00410001, 0xbc100229, 0x02698010, 0x00000000 }, + { 0x00410001, 0xbc200229, 0x02698020, 0x00000000 }, + { 0x00410001, 0xbc300229, 0x02698030, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000000be }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000044 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000042 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04d004c0 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00802040, 0x2400462c, 0x00a98000, 0x00a98005 }, + { 0x00802048, 0x24003e2c, 0x00a98001, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98002, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98003, 0x00140014 }, + { 0x00802048, 0xb8003e2d, 0x00a98004, 0xfffbfffb }, + { 0x00802040, 0x2400462c, 0x00a98040, 0x00a98045 }, + { 0x00802048, 0x24003e2c, 0x00a98041, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98042, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98043, 0x00140014 }, + { 0x00802048, 0xb8203e2d, 0x00a98044, 0xfffbfffb }, + { 0x00400040, 0x2400462c, 0x00698080, 0x00698085 }, + { 0x00400048, 0x24003e2c, 0x00698081, 0xfffbfffb }, + { 0x00400048, 0x24003e2c, 0x00698082, 0x00140014 }, + { 0x00400048, 0x24003e2c, 0x00698083, 0x00140014 }, + { 0x00400048, 0xb8403e2d, 0x00698084, 0xfffbfffb }, + { 0x00000401, 0x220c0060, 0x00000000, 0x04e004d0 }, + { 0x00000c01, 0x22000060, 0x00000000, 0x04d004c8 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04e004d8 }, + { 0x00000801, 0x22080060, 0x00000000, 0x04f004e8 }, + { 0x00800040, 0x24003dac, 0x00b104c0, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x00b104e0, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed9800, 0x00140014 }, + { 0x00010001, 0x220c0168, 0x00000000, 0x05400540 }, + { 0x00110001, 0x220c0168, 0x00000000, 0x04c004c0 }, + { 0x00800048, 0x24003dac, 0x01e98000, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01e98020, 0x00010001 }, + { 0x00800048, 0x24003dac, 0x01e98800, 0x00140014 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000060 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000022 }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010001, 0x220e0168, 0x02000000, 0x05400540 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x00800040, 0x24003e2c, 0x01e98000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98001, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98002, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98003, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98004, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98005, 0x00010001 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x03600010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x00400401, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x00400840, 0x22083d2c, 0x0069044c, 0x00100010 }, + { 0x00610040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00800040, 0x24003e2c, 0x01e983e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e993e0, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e99000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e99020, 0x00010001 }, + { 0x00000001, 0x220e0168, 0x00000000, 0x05400540 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x00400001, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x01400010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01400010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00010001 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00100010 }, + { 0x00800001, 0x25400229, 0x01e98000, 0x00000000 }, + { 0x80800042, 0x45404631, 0x00d20540, 0x00d204c0 }, + { 0x00000001, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x00400001, 0xd8000231, 0x008a0540, 0x00000000 }, + { 0x00400001, 0xd8100231, 0x008a0548, 0x00000000 }, + { 0x00400001, 0xd8200231, 0x008a0550, 0x00000000 }, + { 0x00400001, 0xd8300231, 0x008a0558, 0x00000000 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00600041, 0x24002628, 0x00898000, 0x00000540 }, + { 0x00600048, 0x24002628, 0x00898002, 0x00000542 }, + { 0x00600048, 0x24002628, 0x00898008, 0x00000544 }, + { 0x00600048, 0x2540262d, 0x0089800a, 0x00000546 }, + { 0x00400401, 0xbc0001a9, 0x00690540, 0x00000000 }, + { 0x00400801, 0xbc1001a9, 0x00690548, 0x00000000 }, + { 0x01000010, 0x20003d2c, 0x00000022, 0x00030003 }, + { 0x01000040, 0x20223d29, 0x02000022, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00080008 }, + { 0x00110040, 0x24540c21, 0x00000454, 0x00080008 }, + { 0x00010040, 0x24540c21, 0x00000454, 0x00180038 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffee4 }, + { 0x01800010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0xffe0ffe0 }, + { 0x00010001, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00110001, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00000001, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00800040, 0x24003dac, 0x00b19c00, 0x00200020 }, + { 0x00800040, 0x24203dac, 0x00b19c20, 0x00200020 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00060006 }, + { 0x8080000c, 0xdc203d91, 0x00b10420, 0x00060006 }, + { 0x02000040, 0x210e3d29, 0x0200010e, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00040004 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffce0 }, + { 0x01000005, 0x25643e2d, 0x00000060, 0x00c000c0 }, + { 0x01000010, 0x20003dac, 0x02000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000086 }, + { 0x00a02001, 0x46400231, 0x00d20780, 0x00000000 }, + { 0x00a02001, 0x46800231, 0x00d207c0, 0x00000000 }, + { 0x00a02001, 0x46c00231, 0x00d20480, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000007e }, + { 0x80a02042, 0x46404631, 0x00d20640, 0x00d20780 }, + { 0x80a02042, 0x46804631, 0x00d20680, 0x00d207c0 }, + { 0x80a02042, 0x46c04631, 0x00d206c0, 0x00d20480 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000076 }, + { 0x01000010, 0x20003dac, 0x02000564, 0x00800080 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffffe4 }, + { 0x00200401, 0x256801ed, 0x00000000, 0x00200020 }, + { 0x00200801, 0x256c01ed, 0x00000000, 0x00060006 }, + { 0x00400001, 0x258001ed, 0x00000000, 0x00000000 }, + { 0x00600001, 0x45a001ad, 0x00058c00, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000028 }, + { 0x02600005, 0x20006e28, 0x02008c0c, 0x88848421 }, + { 0x0220000c, 0x25663dad, 0x00000458, 0x00010001 }, + { 0x0000000c, 0x25643dad, 0x00000564, 0x00060006 }, + { 0x00310001, 0x25c401ad, 0x00000458, 0x00000000 }, + { 0x00210001, 0x25c00061, 0x00000000, 0x00010001 }, + { 0x00310040, 0x25c03dad, 0x004545c4, 0x00010001 }, + { 0x00610001, 0x25a00061, 0x02000000, 0x00000080 }, + { 0x00710001, 0x45a002ad, 0x02ae8c00, 0x00000000 }, + { 0x00710001, 0x45a202ad, 0x02ae8c01, 0x00000000 }, + { 0x00800041, 0x25a035ad, 0x00b105a0, 0x000905c0 }, + { 0x01200010, 0x20003e2c, 0x0245006e, 0x00000000 }, + { 0x00310009, 0x244c45ad, 0x02000564, 0x0045006e }, + { 0x00210001, 0x244c01ed, 0x02000000, 0x00000000 }, + { 0x0031000c, 0x244c3dad, 0x0245044c, 0x00010001 }, + { 0x00200009, 0x256835ad, 0x0045044c, 0x00000566 }, + { 0x00210040, 0x256835ad, 0x02450568, 0x00000566 }, + { 0x00200040, 0x256c362d, 0x0045006e, 0x00000566 }, + { 0x00400040, 0x240035ac, 0x00ab05a2, 0x00ab05a6 }, + { 0x00400040, 0x2400358c, 0x00690400, 0x00000566 }, + { 0x0040000c, 0x2580358d, 0x00690400, 0x00000566 }, + { 0x00800041, 0x2400362c, 0x00d20640, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d20660, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d20780, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207a0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x464035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x466035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d20680, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d206a0, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d207c0, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207e0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x468035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x46a035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d206c0, 0x000705a8 }, + { 0x00800041, 0x2420362c, 0x00d206e0, 0x000705a8 }, + { 0x00800048, 0x2400362c, 0x00d20480, 0x000705ac }, + { 0x00800048, 0x2420362c, 0x00d204a0, 0x000705ac }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x0000056a }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x0000056a }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056e }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056e }, + { 0x80800040, 0x46c035b1, 0x00b104c0, 0x00050582 }, + { 0x80800040, 0x46e035b1, 0x00b104e0, 0x00050582 }, + { 0x01800005, 0x20003d2c, 0x02000020, 0x00020002 }, + { 0x80800040, 0xd00045b1, 0x00b19000, 0x00d20640 }, + { 0x80800040, 0xd02045b1, 0x00b19020, 0x00d20660 }, + { 0x80800040, 0xd04045b1, 0x00b19040, 0x00d20680 }, + { 0x80800040, 0xd06045b1, 0x00b19060, 0x00d206a0 }, + { 0x00000040, 0x22083d8c, 0x00000208, 0x00800080 }, + { 0x80400040, 0xd40045b1, 0x00699400, 0x00ab06c0 }, + { 0x80400040, 0xd48045b1, 0x00699480, 0x00ab06c2 }, + { 0x80400040, 0xd42045b1, 0x00699420, 0x00ab06e0 }, + { 0x80400040, 0xd4a045b1, 0x006994a0, 0x00ab06e2 }, + { 0x80400040, 0xd41045b1, 0x00699410, 0x00ab06d0 }, + { 0x80400040, 0xd49045b1, 0x00699490, 0x00ab06d2 }, + { 0x80400040, 0xd43045b1, 0x00699430, 0x00ab06f0 }, + { 0x80400040, 0xd4b045b1, 0x006994b0, 0x00ab06f2 }, + { 0x00000040, 0x220a3d8c, 0x0000020a, 0x00080008 }, + { 0x00110040, 0x220a3d8c, 0x0200020a, 0x00300030 }, + { 0x01000010, 0x20003d2c, 0x02000020, 0x00060006 }, + { 0x00000040, 0x20203d29, 0x00000020, 0x00020002 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00100001 }, + { 0x00000040, 0x220c358c, 0x0000020c, 0x0000003a }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffc1e }, + { 0x00000001, 0x220001ec, 0x00000000, 0x01400140 }, + { 0x00000001, 0x220201ec, 0x00000000, 0x01c001c0 }, + { 0x00800401, 0x20400236, 0x01ee8000, 0x00000000 }, + { 0x00800801, 0x20500236, 0x01ee8010, 0x00000000 }, + { 0x00800401, 0x20600236, 0x01ee8020, 0x00000000 }, + { 0x00800801, 0x20700236, 0x01ee8030, 0x00000000 }, + { 0x00800401, 0x20800236, 0x01ee8040, 0x00000000 }, + { 0x00800801, 0x20900236, 0x01ee8050, 0x00000000 }, + { 0x00800401, 0x20a00236, 0x01ee8060, 0x00000000 }, + { 0x00800801, 0x20b00236, 0x01ee8070, 0x00000000 }, + { 0x00800401, 0x20c00236, 0x01ee8100, 0x00000000 }, + { 0x00800801, 0x20d00236, 0x01ee8110, 0x00000000 }, + { 0x00800401, 0x20e00236, 0x01ee8120, 0x00000000 }, + { 0x00800801, 0x20f00236, 0x01ee8130, 0x00000000 }, + { 0x00800401, 0x21000236, 0x01ee8140, 0x00000000 }, + { 0x00800801, 0x21100236, 0x01ee8150, 0x00000000 }, + { 0x00800401, 0x21200236, 0x01ee8160, 0x00000000 }, + { 0x00800801, 0x21300236, 0x01ee8170, 0x00000000 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x1218a600 }, + { 0x00200401, 0x204001a5, 0x00450064, 0x00000000 }, + { 0x00000801, 0x20480061, 0x00000000, 0x000f000f }, + { 0x01600031, 0x21400021, 0x508d0040, 0x00000200 }, + { 0x00000001, 0x220001ec, 0x00000000, 0x03400340 }, + { 0x00800401, 0x41600236, 0x00d28000, 0x00000000 }, + { 0x00800801, 0x41610236, 0x00d28080, 0x00000000 }, + { 0x00800401, 0x41800236, 0x00d28020, 0x00000000 }, + { 0x00800801, 0x41810236, 0x00d280a0, 0x00000000 }, + { 0x00800401, 0x41a00236, 0x00d28040, 0x00000000 }, + { 0x00800801, 0x41a10236, 0x00d280c0, 0x00000000 }, + { 0x00800401, 0x41c00236, 0x00d28060, 0x00000000 }, + { 0x00800801, 0x41c10236, 0x00d280e0, 0x00000000 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x0a18a601 }, + { 0x0000040c, 0x20443ca5, 0x00000044, 0x00010001 }, + { 0x00000801, 0x20480061, 0x00000000, 0x0007000f }, + { 0x0a600031, 0x21600021, 0x508d0040, 0x00000200 }, + { 0x00000001, 0x21400021, 0x00000140, 0x00000000 }, + { 0x00000001, 0x21600021, 0x00000160, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00600001, 0x20400021, 0x008d0000, 0x00000000 }, + { 0x00000005, 0x20203e2d, 0x00000061, 0x001f001f }, + { 0x00200009, 0x20643e2d, 0x00450064, 0x00040004 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00000000 }, + { 0x01000010, 0x20003dac, 0x00000020, 0x00160016 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x01200010, 0x20003e2c, 0x0200006c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000024 }, + { 0x00800009, 0x25403dad, 0x00050064, 0x00020002 }, + { 0x00000001, 0x203a01ed, 0x00000000, 0x00180018 }, + { 0x00200040, 0x45483dad, 0x00660548, 0x00100010 }, + { 0x00200040, 0x45523dad, 0x00660552, 0x00100010 }, + { 0x00400040, 0x25583dad, 0x00690558, 0x00100010 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0x00200020 }, + { 0x00800040, 0x20a035ad, 0x00b100a0, 0x00b10540 }, + { 0x00600040, 0x45423dad, 0x00ae0542, 0x00200020 }, + { 0x00800040, 0x20e035ad, 0x00b100e0, 0x00b10540 }, + { 0x00600040, 0x45403dad, 0x00ae0540, 0xffe0ffe0 }, + { 0x00800040, 0x20c035ad, 0x00b100c0, 0x00b10540 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000014 }, + { 0x06600010, 0x20003dac, 0x02000020, 0x00030003 }, + { 0x00000001, 0x206c0171, 0x00000000, 0x00000000 }, + { 0x00610001, 0x208000a5, 0x02050080, 0x00000000 }, + { 0x00610001, 0x212000a5, 0x02090100, 0x00000000 }, + { 0x00410001, 0x211000a5, 0x02690100, 0x00000000 }, + { 0x00200009, 0x25403dad, 0x00450064, 0x00020002 }, + { 0x00800040, 0x208035ad, 0x00b10080, 0x00050540 }, + { 0x00200040, 0x40883dad, 0x00660088, 0x00200020 }, + { 0x00200040, 0x40923dad, 0x00660092, 0x00200020 }, + { 0x00400040, 0x20983dad, 0x00690098, 0x00200020 }, + { 0x00200401, 0x22080060, 0x00000000, 0x03400140 }, + { 0x00000c01, 0x220c0060, 0x00000000, 0x04400080 }, + { 0x00000801, 0x22040060, 0x00000000, 0x01000070 }, + { 0x02000005, 0x20002e28, 0x00000061, 0x00400040 }, + { 0x00010005, 0x203e2e29, 0x00000063, 0x00010001 }, + { 0x00110001, 0x203e0169, 0x00000000, 0x00030003 }, + { 0x00000001, 0x20200169, 0x00000000, 0x00000000 }, + { 0x0000000c, 0x2458262d, 0x0000006d, 0x00000020 }, + { 0x00000001, 0x210e0169, 0x00000000, 0x00010001 }, + { 0x00000005, 0x24583dad, 0x00000458, 0x00030003 }, + { 0x01000010, 0x200035ac, 0x0000010e, 0x00000458 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000031a }, + { 0x01000010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00010401, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00010805, 0x245a3e2d, 0x02008800, 0x007f007f }, + { 0x00110401, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00110805, 0x245a3e2d, 0x02008804, 0x007f007f }, + { 0x00010005, 0x245c3e2d, 0x02008800, 0x00800080 }, + { 0x00110005, 0x245c3e2d, 0x02008804, 0x00800080 }, + { 0x00000009, 0x245c3dad, 0x0000045c, 0x00010001 }, + { 0x0000000c, 0x211e362d, 0x0000006c, 0x00000020 }, + { 0x00000040, 0x240035ac, 0x0000003e, 0x0000045c }, + { 0x01000010, 0x20003d8c, 0x00210400, 0x00010001 }, + { 0x01000010, 0x20003d8c, 0x02210400, 0x01000100 }, + { 0x00000001, 0x203c01ed, 0x00000000, 0x00000000 }, + { 0x00010001, 0x203c01ed, 0x00000000, 0x00020002 }, + { 0x00010001, 0x203c01ed, 0x02000000, 0xfffefffe }, + { 0x02000005, 0x20002e28, 0x00000061, 0x00400040 }, + { 0x00010040, 0x244c0d21, 0x0000045a, 0x0208e602 }, + { 0x00110040, 0x20780d21, 0x0000045a, 0x0208a002 }, + { 0x00010040, 0x20782421, 0x0000044c, 0x0000045c }, + { 0x02000005, 0x20003dac, 0x0200011e, 0x00030003 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000001c0 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000012 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00700000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c000c }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x0007000c }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00400000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00110040, 0x20441da5, 0x02000442, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x000c0007 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00070007 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0xb80235ad, 0x00009802, 0x0000003c }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x01600006, 0x20003dac, 0x0000044a, 0x00000000 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00010040, 0x22000c20, 0x00000078, 0x00200010 }, + { 0x00110040, 0x22000c20, 0x00000078, 0x00300010 }, + { 0x00010001, 0x20480061, 0x00000000, 0x00030009 }, + { 0x00110001, 0x20480061, 0x00000000, 0x00040009 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x01000005, 0x20003dac, 0x00009800, 0x00030003 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000001a }, + { 0x00600001, 0x26220231, 0x008d05c0, 0x00000000 }, + { 0x00600401, 0x26120231, 0x008d05b8, 0x00000000 }, + { 0x00600801, 0x26020231, 0x008d05b0, 0x00000000 }, + { 0x00600401, 0x25f20231, 0x008d05a8, 0x00000000 }, + { 0x00600801, 0x25e20231, 0x008d05a0, 0x00000000 }, + { 0x00600401, 0x25d20231, 0x008d0598, 0x00000000 }, + { 0x00600801, 0x25c20231, 0x008d0590, 0x00000000 }, + { 0x00600401, 0x25b20231, 0x008d0588, 0x00000000 }, + { 0x00600801, 0x25a20231, 0x008d0580, 0x00000000 }, + { 0x00600401, 0x25920231, 0x008d0578, 0x00000000 }, + { 0x00600801, 0x25820231, 0x008d0570, 0x00000000 }, + { 0x00600001, 0x25720231, 0x008d0568, 0x00000000 }, + { 0x00600001, 0x25620231, 0x008d0560, 0x00000000 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00810001, 0xbc000229, 0x02ad8000, 0x00000000 }, + { 0x00810001, 0xbc200229, 0x02ad8020, 0x00000000 }, + { 0x00810001, 0xbc400229, 0x02ad8040, 0x00000000 }, + { 0x00810001, 0xbc600229, 0x02ad8060, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000012a }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000066 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000064 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04a00480 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00a02040, 0x2400462c, 0x00ad8000, 0x00ad8005 }, + { 0x00a02048, 0x24003e2c, 0x00ad8001, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8002, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8003, 0x00140014 }, + { 0x00a02048, 0xb8003e2d, 0x00ad8004, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8040, 0x00ad8045 }, + { 0x00a02048, 0x24003e2c, 0x00ad8041, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8042, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8043, 0x00140014 }, + { 0x00a02048, 0xb8403e2d, 0x00ad8044, 0xfffbfffb }, + { 0x00a02040, 0x2400462c, 0x00ad8080, 0x00ad8085 }, + { 0x00a02048, 0x24003e2c, 0x00ad8081, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8082, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8083, 0x00140014 }, + { 0x00a02048, 0xb8803e2d, 0x00ad8084, 0xfffbfffb }, + { 0x00600040, 0x2400462c, 0x008d80c0, 0x008d80c5 }, + { 0x00600048, 0x24003e2c, 0x008d80c1, 0xfffbfffb }, + { 0x00600048, 0x24003e2c, 0x008d80c2, 0x00140014 }, + { 0x00600048, 0x24003e2c, 0x008d80c3, 0x00140014 }, + { 0x00600048, 0xb8c03e2d, 0x008d80c4, 0xfffbfffb }, + { 0x00000401, 0x22000060, 0x00000000, 0x04a00480 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04a00490 }, + { 0x00010802, 0x220c2d28, 0x00000454, 0x04800480 }, + { 0x00a02040, 0x24003dac, 0x00b18000, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8800, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8820, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18020, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8820, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8840, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18040, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x000a000a }, + { 0x00a02040, 0x24003dac, 0x00b18040, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x01ed8840, 0xfffbfffb }, + { 0x00800048, 0x24203dac, 0x01ed8860, 0xfffbfffb }, + { 0x00a02048, 0x24003dac, 0x00b18060, 0x00140014 }, + { 0x00800048, 0x24003dac, 0x01ed8860, 0x00140014 }, + { 0x00800048, 0x24203dac, 0x01ed8880, 0x00140014 }, + { 0x00a02048, 0x24003dac, 0x00b18080, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed8880, 0x00010001 }, + { 0x00800048, 0x24203dac, 0x01ed88a0, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x000a000a }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x000000b4 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000004c }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000004a }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010002, 0x220c2d28, 0x02000454, 0x04800480 }, + { 0x00800040, 0x24003e2c, 0x01ed8000, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8001, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8021, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8002, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8022, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8003, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8023, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8004, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8024, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8005, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8025, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00800040, 0x24003e2c, 0x01ed8040, 0x00100010 }, + { 0x00800040, 0x24203e2c, 0x01ed8060, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8041, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8061, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8042, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8062, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8043, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8063, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8044, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8064, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8045, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8065, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000005c }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000062 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000042 }, + { 0x00200401, 0x22000128, 0x0066044c, 0x00000000 }, + { 0x00200c01, 0x22040128, 0x0045044e, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x03400010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220c0168, 0x02000000, 0x04800480 }, + { 0x00a02040, 0x24003e2c, 0x00ad83e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8be0, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8800, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8800, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8820, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0x00010001 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8203d91, 0x00b10420, 0x00050005 }, + { 0x00a02040, 0x24003e2c, 0x00ad8020, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01ed8820, 0xfffbfffb }, + { 0x00800048, 0x24203e2c, 0x01ed8840, 0xfffbfffb }, + { 0x00a02048, 0x24003e2c, 0x00ad8040, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01ed8840, 0x00140014 }, + { 0x00800048, 0x24203e2c, 0x01ed8860, 0x00140014 }, + { 0x00a02048, 0x24003e2c, 0x00ad8060, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01ed8860, 0x00010001 }, + { 0x00800048, 0x24203e2c, 0x01ed8880, 0x00010001 }, + { 0x8080000c, 0xd8403d91, 0x00b10400, 0x00050005 }, + { 0x8080100c, 0xd8603d91, 0x00b10420, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000001c }, + { 0x00200401, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00000801, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x01200010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01200010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00210040, 0x22002d08, 0x00450200, 0x00010001 }, + { 0x00210040, 0x22002d08, 0x02450200, 0x00100010 }, + { 0x00800001, 0xb8000229, 0x01ed8000, 0x00000000 }, + { 0x00800001, 0xb8200229, 0x01ed8020, 0x00000000 }, + { 0x00800001, 0xb8400229, 0x01ed8040, 0x00000000 }, + { 0x00800001, 0xb8600229, 0x01ed8060, 0x00000000 }, + { 0x80800042, 0xd8004631, 0x00d29800, 0x00d20480 }, + { 0x80800042, 0xd8204631, 0x00d29820, 0x00d204a0 }, + { 0x80800042, 0xd8404631, 0x00d29840, 0x00d204c0 }, + { 0x80800042, 0xd8604631, 0x00d29860, 0x00d204e0 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000c01, 0x220201e8, 0x00000000, 0x07100710 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00200040, 0x244c3d09, 0x00450200, 0x00100010 }, + { 0x00800041, 0x24002628, 0x00ad8000, 0x00000540 }, + { 0x00800041, 0x24202628, 0x00ad8020, 0x00000540 }, + { 0x00800048, 0x24002628, 0x00ad8002, 0x00000542 }, + { 0x00800048, 0x24202628, 0x00ad8022, 0x00000542 }, + { 0x00200001, 0x22000128, 0x0045044c, 0x00000000 }, + { 0x00800048, 0x24002628, 0x01ed8000, 0x00000544 }, + { 0x00800048, 0x24202628, 0x01ed8020, 0x00000544 }, + { 0x00800048, 0xbc002629, 0x01ed8002, 0x00000546 }, + { 0x00801048, 0xbc202629, 0x01ed8022, 0x00000546 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000128 }, + { 0x00000001, 0x20220169, 0x00000000, 0x00040004 }, + { 0x00200005, 0x24443dad, 0x00459800, 0x00030003 }, + { 0x0020000c, 0x24403dad, 0x00459800, 0x00020002 }, + { 0x01600006, 0x20003dac, 0x02000446, 0x00000000 }, + { 0x00010040, 0x22000c20, 0x02000078, 0x00200000 }, + { 0x00110040, 0x22000c20, 0x02000078, 0x00500000 }, + { 0x00310040, 0x20401da5, 0x02450440, 0xfffffffe }, + { 0x00110001, 0x20480061, 0x02000000, 0x00080008 }, + { 0x00010040, 0x20401da5, 0x02000440, 0xfffffffe }, + { 0x00010001, 0x204401a5, 0x02000442, 0x00000000 }, + { 0x00010001, 0x20480061, 0x02000000, 0x00030008 }, + { 0x01600031, 0x25600021, 0x408d0040, 0x00000200 }, + { 0x00000040, 0xb80235ad, 0x00009802, 0x0000003c }, + { 0x00000040, 0x22000c20, 0x00000078, 0x00100010 }, + { 0x0020040c, 0x24403dad, 0x00459800, 0x00030003 }, + { 0x00200805, 0x24483dad, 0x00459800, 0x00070007 }, + { 0x00200001, 0x204001a5, 0x00450440, 0x00000000 }, + { 0x00000009, 0x20403ca5, 0x00000040, 0x00010001 }, + { 0x00000001, 0x20480061, 0x00000000, 0x00020005 }, + { 0x02600031, 0x27000021, 0x408d0040, 0x00000200 }, + { 0x00110001, 0x220001ec, 0x02000000, 0x05820582 }, + { 0x00010001, 0x220001ec, 0x02000000, 0x05620562 }, + { 0x00000001, 0x220e0128, 0x00000454, 0x00000000 }, + { 0x01800006, 0x200035ac, 0x02000446, 0x00000444 }, + { 0x00410001, 0xbc000229, 0x02698000, 0x00000000 }, + { 0x00410001, 0xbc100229, 0x02698010, 0x00000000 }, + { 0x00410001, 0xbc200229, 0x02698020, 0x00000000 }, + { 0x00410001, 0xbc300229, 0x02698030, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x000000be }, + { 0x00600001, 0x2024018d, 0x008d0200, 0x00000000 }, + { 0x01000041, 0x245e35ad, 0x00000446, 0x00000444 }, + { 0x02000005, 0x20003db0, 0x0200045e, 0x00010001 }, + { 0x00000040, 0x22022d08, 0x00000200, 0x00100010 }, + { 0x00200040, 0x22042d08, 0x00450200, 0x00200020 }, + { 0x00400001, 0x244c0109, 0x00690200, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000044 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000042 }, + { 0x00000040, 0x22003d08, 0x00000200, 0xffdeffde }, + { 0x00000440, 0x22023d08, 0x00000202, 0xffeeffee }, + { 0x00000801, 0x220c0060, 0x00000000, 0x04d004c0 }, + { 0x01000010, 0x20003db0, 0x0000045e, 0x00040004 }, + { 0x00802040, 0x2400462c, 0x00a98000, 0x00a98005 }, + { 0x00802048, 0x24003e2c, 0x00a98001, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98002, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98003, 0x00140014 }, + { 0x00802048, 0xb8003e2d, 0x00a98004, 0xfffbfffb }, + { 0x00802040, 0x2400462c, 0x00a98040, 0x00a98045 }, + { 0x00802048, 0x24003e2c, 0x00a98041, 0xfffbfffb }, + { 0x00802048, 0x24003e2c, 0x00a98042, 0x00140014 }, + { 0x00802048, 0x24003e2c, 0x00a98043, 0x00140014 }, + { 0x00802048, 0xb8203e2d, 0x00a98044, 0xfffbfffb }, + { 0x00400040, 0x2400462c, 0x00698080, 0x00698085 }, + { 0x00400048, 0x24003e2c, 0x00698081, 0xfffbfffb }, + { 0x00400048, 0x24003e2c, 0x00698082, 0x00140014 }, + { 0x00400048, 0x24003e2c, 0x00698083, 0x00140014 }, + { 0x00400048, 0xb8403e2d, 0x00698084, 0xfffbfffb }, + { 0x00000401, 0x220c0060, 0x00000000, 0x04e004d0 }, + { 0x00000c01, 0x22000060, 0x00000000, 0x04d004c8 }, + { 0x00000c01, 0x22040060, 0x00000000, 0x04e004d8 }, + { 0x00000801, 0x22080060, 0x00000000, 0x04f004e8 }, + { 0x00800040, 0x24003dac, 0x00b104c0, 0x02000200 }, + { 0x00800048, 0x24003dac, 0x00b104e0, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01ed9800, 0x00140014 }, + { 0x00010001, 0x220c0168, 0x00000000, 0x05400540 }, + { 0x00110001, 0x220c0168, 0x00000000, 0x04c004c0 }, + { 0x00800048, 0x24003dac, 0x01e98000, 0xfffbfffb }, + { 0x00800048, 0x24003dac, 0x01e98020, 0x00010001 }, + { 0x00800048, 0x24003dac, 0x01e98800, 0x00140014 }, + { 0x8080000c, 0xd8003d91, 0x00b10400, 0x000a000a }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000060 }, + { 0x01000010, 0x20003db0, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000022 }, + { 0x00400040, 0x22003d28, 0x0069044c, 0xfffefffe }, + { 0x03400010, 0x20003dac, 0x00000446, 0x00020002 }, + { 0x01000010, 0x20003db0, 0x02000444, 0x00020002 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00100010 }, + { 0x01000010, 0x20003dac, 0x00000446, 0x00000000 }, + { 0x00010001, 0x220e0168, 0x02000000, 0x05400540 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x00800040, 0x24003e2c, 0x01e98000, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98001, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98002, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98003, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98004, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e98005, 0x00010001 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x00000004 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000036 }, + { 0x01000010, 0x20003db0, 0x00000446, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000024 }, + { 0x03600010, 0x20003dac, 0x02000444, 0x00020002 }, + { 0x00400401, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x00400840, 0x22083d2c, 0x0069044c, 0x00100010 }, + { 0x00610040, 0x22002d08, 0x02690200, 0x00010001 }, + { 0x01000010, 0x20003dac, 0x00000444, 0x00000000 }, + { 0x01000010, 0x20003db0, 0x02000446, 0x00020002 }, + { 0x00800040, 0x24003e2c, 0x01e983e0, 0x00100010 }, + { 0x00800048, 0x24003e2c, 0x01e98000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e98020, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e993e0, 0xfffbfffb }, + { 0x00800048, 0x24003e2c, 0x01e99000, 0x00140014 }, + { 0x00800048, 0x24003e2c, 0x01e99020, 0x00010001 }, + { 0x00000001, 0x220e0168, 0x00000000, 0x05400540 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000002 }, + { 0x00110001, 0x220e0168, 0x02000000, 0x04c004c0 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00050005 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x00400001, 0x22000128, 0x0069044c, 0x00000000 }, + { 0x01400010, 0x20003dac, 0x00000444, 0x00030003 }, + { 0x01400010, 0x20003dac, 0x02000446, 0x00030003 }, + { 0x00410040, 0x22002d08, 0x00690200, 0x00010001 }, + { 0x00410040, 0x22002d08, 0x02690200, 0x00100010 }, + { 0x00800001, 0x25400229, 0x01e98000, 0x00000000 }, + { 0x80800042, 0x45404631, 0x00d20540, 0x00d204c0 }, + { 0x00000001, 0x220c0128, 0x00000454, 0x00000000 }, + { 0x00400001, 0xd8000231, 0x008a0540, 0x00000000 }, + { 0x00400001, 0xd8100231, 0x008a0548, 0x00000000 }, + { 0x00400001, 0xd8200231, 0x008a0550, 0x00000000 }, + { 0x00400001, 0xd8300231, 0x008a0558, 0x00000000 }, + { 0x00600001, 0x220001ac, 0x008d0024, 0x00000000 }, + { 0x00200040, 0x244c3dad, 0x00450448, 0xfff8fff8 }, + { 0x00000401, 0x220001ec, 0x00000000, 0x07000700 }, + { 0x00000801, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00000441, 0x254635ad, 0x00000448, 0x0000044a }, + { 0x00000c41, 0x254025ad, 0x0000444c, 0x0000444e }, + { 0x00000c41, 0x254225ad, 0x00000448, 0x0000444e }, + { 0x00000841, 0x254435ad, 0x0000444c, 0x0000044a }, + { 0x00600041, 0x24002628, 0x00898000, 0x00000540 }, + { 0x00600048, 0x24002628, 0x00898002, 0x00000542 }, + { 0x00600048, 0x24002628, 0x00898008, 0x00000544 }, + { 0x00600048, 0x2540262d, 0x0089800a, 0x00000546 }, + { 0x00400401, 0xbc0001a9, 0x00690540, 0x00000000 }, + { 0x00400801, 0xbc1001a9, 0x00690548, 0x00000000 }, + { 0x01000010, 0x20003d2c, 0x00000022, 0x00030003 }, + { 0x01000040, 0x20223d29, 0x02000022, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00080008 }, + { 0x00110040, 0x24540c21, 0x00000454, 0x00080008 }, + { 0x00010040, 0x24540c21, 0x00000454, 0x00180038 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffee4 }, + { 0x01800010, 0x20003dac, 0x0200010e, 0x00010001 }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0xffe0ffe0 }, + { 0x00010001, 0x24540061, 0x02000000, 0x06c00640 }, + { 0x00110001, 0x24540061, 0x02000000, 0x04800780 }, + { 0x00000001, 0x220e0128, 0x00000456, 0x00000000 }, + { 0x00800040, 0x24003dac, 0x00b19c00, 0x00200020 }, + { 0x00800040, 0x24203dac, 0x00b19c20, 0x00200020 }, + { 0x8080000c, 0xdc003d91, 0x00b10400, 0x00060006 }, + { 0x8080000c, 0xdc203d91, 0x00b10420, 0x00060006 }, + { 0x02000040, 0x210e3d29, 0x0200010e, 0xffffffff }, + { 0x00000040, 0x220c3d8c, 0x0000020c, 0x00040004 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffcdc }, + { 0x01000005, 0x25643e2d, 0x00000060, 0x00c000c0 }, + { 0x01000010, 0x20003dac, 0x02000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0x00000016 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000086 }, + { 0x00a02001, 0x46400231, 0x00d20780, 0x00000000 }, + { 0x00a02001, 0x46800231, 0x00d207c0, 0x00000000 }, + { 0x00a02001, 0x46c00231, 0x00d20480, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000007e }, + { 0x80a02042, 0x46404631, 0x00d20640, 0x00d20780 }, + { 0x80a02042, 0x46804631, 0x00d20680, 0x00d207c0 }, + { 0x80a02042, 0x46c04631, 0x00d206c0, 0x00d20480 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000076 }, + { 0x01000010, 0x20003dac, 0x02000564, 0x00800080 }, + { 0x00110220, 0x34001c00, 0x02001400, 0x0000000e }, + { 0x01000010, 0x20003dac, 0x00000458, 0x00020002 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffffe4 }, + { 0x00200401, 0x256801ed, 0x00000000, 0x00200020 }, + { 0x00200801, 0x256c01ed, 0x00000000, 0x00060006 }, + { 0x00400001, 0x258001ed, 0x00000000, 0x00000000 }, + { 0x00600001, 0x45a001ad, 0x00058c00, 0x00000000 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000028 }, + { 0x02600005, 0x20006e28, 0x02008c0c, 0x88848421 }, + { 0x0220000c, 0x25663dad, 0x00000458, 0x00010001 }, + { 0x0000000c, 0x25643dad, 0x00000564, 0x00060006 }, + { 0x00310001, 0x25c401ad, 0x00000458, 0x00000000 }, + { 0x00210001, 0x25c00061, 0x00000000, 0x00010001 }, + { 0x00310040, 0x25c03dad, 0x004545c4, 0x00010001 }, + { 0x00610001, 0x25a00061, 0x02000000, 0x00000080 }, + { 0x00710001, 0x45a002ad, 0x02ae8c00, 0x00000000 }, + { 0x00710001, 0x45a202ad, 0x02ae8c01, 0x00000000 }, + { 0x00800041, 0x25a035ad, 0x00b105a0, 0x000905c0 }, + { 0x01200010, 0x20003e2c, 0x0245006e, 0x00000000 }, + { 0x00310009, 0x244c45ad, 0x02000564, 0x0045006e }, + { 0x00210001, 0x244c01ed, 0x02000000, 0x00000000 }, + { 0x0031000c, 0x244c3dad, 0x0245044c, 0x00010001 }, + { 0x00200009, 0x256835ad, 0x0045044c, 0x00000566 }, + { 0x00210040, 0x256835ad, 0x02450568, 0x00000566 }, + { 0x00200040, 0x256c362d, 0x0045006e, 0x00000566 }, + { 0x00400040, 0x240035ac, 0x00ab05a2, 0x00ab05a6 }, + { 0x00400040, 0x2400358c, 0x00690400, 0x00000566 }, + { 0x0040000c, 0x2580358d, 0x00690400, 0x00000566 }, + { 0x00800041, 0x2400362c, 0x00d20640, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d20660, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d20780, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207a0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x464035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x466035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d20680, 0x000005a0 }, + { 0x00800041, 0x2420362c, 0x00d206a0, 0x000005a0 }, + { 0x00800048, 0x2400362c, 0x00d207c0, 0x000005a4 }, + { 0x00800048, 0x2420362c, 0x00d207e0, 0x000005a4 }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x00000568 }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x00000568 }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056c }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056c }, + { 0x80800040, 0x468035b1, 0x00b104c0, 0x00000580 }, + { 0x80800040, 0x46a035b1, 0x00b104e0, 0x00000580 }, + { 0x00800041, 0x2400362c, 0x00d206c0, 0x000705a8 }, + { 0x00800041, 0x2420362c, 0x00d206e0, 0x000705a8 }, + { 0x00800048, 0x2400362c, 0x00d20480, 0x000705ac }, + { 0x00800048, 0x2420362c, 0x00d204a0, 0x000705ac }, + { 0x00800040, 0x2400358c, 0x00b10400, 0x0000056a }, + { 0x00800040, 0x2420358c, 0x00b10420, 0x0000056a }, + { 0x0080000c, 0x24c0358d, 0x00b10400, 0x0000056e }, + { 0x0080000c, 0x24e0358d, 0x00b10420, 0x0000056e }, + { 0x80800040, 0x46c035b1, 0x00b104c0, 0x00050582 }, + { 0x80800040, 0x46e035b1, 0x00b104e0, 0x00050582 }, + { 0x01800005, 0x20003d2c, 0x02000020, 0x00020002 }, + { 0x80800040, 0xd00045b1, 0x00b19000, 0x00d20640 }, + { 0x80800040, 0xd02045b1, 0x00b19020, 0x00d20660 }, + { 0x80800040, 0xd04045b1, 0x00b19040, 0x00d20680 }, + { 0x80800040, 0xd06045b1, 0x00b19060, 0x00d206a0 }, + { 0x00000040, 0x22083d8c, 0x00000208, 0x00800080 }, + { 0x80400040, 0xd40045b1, 0x00699400, 0x00ab06c0 }, + { 0x80400040, 0xd48045b1, 0x00699480, 0x00ab06c2 }, + { 0x80400040, 0xd42045b1, 0x00699420, 0x00ab06e0 }, + { 0x80400040, 0xd4a045b1, 0x006994a0, 0x00ab06e2 }, + { 0x80400040, 0xd41045b1, 0x00699410, 0x00ab06d0 }, + { 0x80400040, 0xd49045b1, 0x00699490, 0x00ab06d2 }, + { 0x80400040, 0xd43045b1, 0x00699430, 0x00ab06f0 }, + { 0x80400040, 0xd4b045b1, 0x006994b0, 0x00ab06f2 }, + { 0x00000040, 0x220a3d8c, 0x0000020a, 0x00080008 }, + { 0x00110040, 0x220a3d8c, 0x0200020a, 0x00300030 }, + { 0x01000010, 0x20003d2c, 0x02000020, 0x00060006 }, + { 0x00000040, 0x20203d29, 0x00000020, 0x00020002 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00100001 }, + { 0x00000040, 0x220c358c, 0x0000020c, 0x0000003a }, + { 0x00110220, 0x34001c00, 0x02001400, 0xfffffc1a }, + { 0x00000001, 0x220001ec, 0x00000000, 0x01400140 }, + { 0x00000001, 0x220201ec, 0x00000000, 0x01c001c0 }, + { 0x00800401, 0x20400236, 0x01ee8000, 0x00000000 }, + { 0x00800801, 0x20500236, 0x01ee8010, 0x00000000 }, + { 0x00800401, 0x20600236, 0x01ee8020, 0x00000000 }, + { 0x00800801, 0x20700236, 0x01ee8030, 0x00000000 }, + { 0x00800401, 0x20800236, 0x01ee8040, 0x00000000 }, + { 0x00800801, 0x20900236, 0x01ee8050, 0x00000000 }, + { 0x00800401, 0x20a00236, 0x01ee8060, 0x00000000 }, + { 0x00800801, 0x20b00236, 0x01ee8070, 0x00000000 }, + { 0x00800401, 0x20c00236, 0x01ee8100, 0x00000000 }, + { 0x00800801, 0x20d00236, 0x01ee8110, 0x00000000 }, + { 0x00800401, 0x20e00236, 0x01ee8120, 0x00000000 }, + { 0x00800801, 0x20f00236, 0x01ee8130, 0x00000000 }, + { 0x00800401, 0x21000236, 0x01ee8140, 0x00000000 }, + { 0x00800801, 0x21100236, 0x01ee8150, 0x00000000 }, + { 0x00800401, 0x21200236, 0x01ee8160, 0x00000000 }, + { 0x00800801, 0x21300236, 0x01ee8170, 0x00000000 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x1218a000 }, + { 0x00200401, 0x204001a5, 0x00450064, 0x00000000 }, + { 0x00000801, 0x20480061, 0x00000000, 0x000f000f }, + { 0x01600031, 0x21400021, 0x508d0040, 0x00000200 }, + { 0x00000001, 0x220001ec, 0x00000000, 0x03400340 }, + { 0x00800401, 0x41600236, 0x00d28000, 0x00000000 }, + { 0x00800801, 0x41610236, 0x00d28080, 0x00000000 }, + { 0x00800401, 0x41800236, 0x00d28020, 0x00000000 }, + { 0x00800801, 0x41810236, 0x00d280a0, 0x00000000 }, + { 0x00800401, 0x41a00236, 0x00d28040, 0x00000000 }, + { 0x00800801, 0x41a10236, 0x00d280c0, 0x00000000 }, + { 0x00800401, 0x41c00236, 0x00d28060, 0x00000000 }, + { 0x00800801, 0x41c10236, 0x00d280e0, 0x00000000 }, + { 0x00000040, 0x22000d20, 0x00000062, 0x0a18a001 }, + { 0x0000040c, 0x20443ca5, 0x00000044, 0x00010001 }, + { 0x00000801, 0x20480061, 0x00000000, 0x0007000f }, + { 0x0a600031, 0x21600021, 0x508d0040, 0x00000200 }, + { 0x00000001, 0x21400021, 0x00000140, 0x00000000 }, + { 0x00000001, 0x21600021, 0x00000160, 0x00000000 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x01000005, 0x20602d29, 0x02000042, 0x00070007 }, + { 0x01000005, 0x20422d29, 0x00000042, 0xfff8fff8 }, + { 0x00000001, 0x202001e9, 0x00000000, 0x100c100c }, + { 0x00600001, 0x27e00021, 0x008d0000, 0x00000000 }, + { 0x00000009, 0x27e82d21, 0x00000040, 0x00060006 }, + { 0x00600001, 0x20a00022, 0x008d0000, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000100 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x01800031, 0x20801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20400022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20480c22, 0x000007e8, 0x00000080 }, + { 0x02800031, 0x21001d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20600022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20680c22, 0x000007e8, 0x00000100 }, + { 0x03800031, 0x21801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20800022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20880c22, 0x000007e8, 0x00000180 }, + { 0x04800031, 0x22001d09, 0x408d0000, 0x02488400 }, + { 0x02600005, 0x20000c20, 0x02000090, 0x00002000 }, + { 0x00000006, 0x20880c21, 0x00000088, 0x00200000 }, + { 0x00200009, 0x20845529, 0x0000009c, 0x00450020 }, + { 0x00200001, 0x40840231, 0x00450094, 0x00000000 }, + { 0x00310001, 0x20840229, 0x02450094, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000095, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000094 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0080, 0x00000000 }, + { 0x00802001, 0x20800061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x020000d0, 0x00002000 }, + { 0x00000006, 0x20c80c21, 0x000000c8, 0x00200000 }, + { 0x00200009, 0x20c45529, 0x000000dc, 0x00450020 }, + { 0x00200001, 0x40c40231, 0x004500d4, 0x00000000 }, + { 0x00310001, 0x20c40229, 0x024500d4, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x000000d5, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x000000d4 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d00c0, 0x00000000 }, + { 0x00802001, 0x20c00061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000110, 0x00002000 }, + { 0x00000006, 0x21080c21, 0x00000108, 0x00200000 }, + { 0x00200009, 0x21045529, 0x0000011c, 0x00450020 }, + { 0x00200001, 0x41040231, 0x00450114, 0x00000000 }, + { 0x00310001, 0x21040229, 0x02450114, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000115, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000114 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0100, 0x00000000 }, + { 0x00802001, 0x21000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000150, 0x00002000 }, + { 0x00000006, 0x21480c21, 0x00000148, 0x00200000 }, + { 0x00200009, 0x21445529, 0x0000015c, 0x00450020 }, + { 0x00200001, 0x41440231, 0x00450154, 0x00000000 }, + { 0x00310001, 0x21440229, 0x02450154, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000155, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000154 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0140, 0x00000000 }, + { 0x00802001, 0x21400061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000190, 0x00002000 }, + { 0x00000006, 0x21880c21, 0x00000188, 0x00200000 }, + { 0x00200009, 0x21845529, 0x0000019c, 0x00450020 }, + { 0x00200001, 0x41840231, 0x00450194, 0x00000000 }, + { 0x00310001, 0x21840229, 0x02450194, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000195, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000194 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0180, 0x00000000 }, + { 0x00802001, 0x21800061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x020001d0, 0x00002000 }, + { 0x00000006, 0x21c80c21, 0x000001c8, 0x00200000 }, + { 0x00200009, 0x21c45529, 0x000001dc, 0x00450020 }, + { 0x00200001, 0x41c40231, 0x004501d4, 0x00000000 }, + { 0x00310001, 0x21c40229, 0x024501d4, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x000001d5, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x000001d4 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d01c0, 0x00000000 }, + { 0x00802001, 0x21c00061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000210, 0x00002000 }, + { 0x00000006, 0x22080c21, 0x00000208, 0x00200000 }, + { 0x00200009, 0x22045529, 0x0000021c, 0x00450020 }, + { 0x00200001, 0x42040231, 0x00450214, 0x00000000 }, + { 0x00310001, 0x22040229, 0x02450214, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000215, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000214 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0200, 0x00000000 }, + { 0x00802001, 0x22000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x02600005, 0x20000c20, 0x02000250, 0x00002000 }, + { 0x00000006, 0x22480c21, 0x00000248, 0x00200000 }, + { 0x00200009, 0x22445529, 0x0000025c, 0x00450020 }, + { 0x00200001, 0x42440231, 0x00450254, 0x00000000 }, + { 0x00310001, 0x22440229, 0x02450254, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00000255, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00000254 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0240, 0x00000000 }, + { 0x00802001, 0x22400061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x01000040, 0x20423dad, 0x00000042, 0xfff8fff8 }, + { 0x00600001, 0x21600022, 0x008d07e0, 0x00000000 }, + { 0x00802001, 0x21800022, 0x008d0080, 0x00000000 }, + { 0x00802001, 0x21c00022, 0x008d00c0, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20280c22, 0x000007e8, 0x00000080 }, + { 0x00802001, 0x20400022, 0x008d0100, 0x00000000 }, + { 0x00802001, 0x20800022, 0x008d0140, 0x00000000 }, + { 0x01800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x21680c22, 0x000007e8, 0x00000100 }, + { 0x00802001, 0x21800022, 0x008d0180, 0x00000000 }, + { 0x00802001, 0x21c00022, 0x008d01c0, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x20280c22, 0x000007e8, 0x00000180 }, + { 0x00802001, 0x20400022, 0x008d0200, 0x00000000 }, + { 0x00802001, 0x20800022, 0x008d0240, 0x00000000 }, + { 0x01800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x27e80c21, 0x000007e8, 0x00000200 }, + { 0x01000010, 0x20002da0, 0x02000060, 0x00000000 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffff00 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x00000040 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x01800031, 0x20801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20400022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20480c22, 0x000007e8, 0x00000080 }, + { 0x02800031, 0x21001d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20600022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20680c22, 0x000007e8, 0x00000100 }, + { 0x03800031, 0x21801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20800022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20880c22, 0x000007e8, 0x00000180 }, + { 0x04800031, 0x22001d09, 0x408d0000, 0x02488400 }, + { 0x00000001, 0x220400e0, 0x00000000, 0x00a00080 }, + { 0x00600001, 0x20a00022, 0x008d0000, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02008810, 0x00002000 }, + { 0x01000040, 0x20603dad, 0x00000060, 0xffffffff }, + { 0x00000006, 0xa8080c21, 0x00008808, 0x00200000 }, + { 0x00200009, 0xa8045529, 0x0000881c, 0x00450020 }, + { 0x00200001, 0xc8040231, 0x00458814, 0x00000000 }, + { 0x00310001, 0xa8040229, 0x02458814, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000000c }, + { 0x00800041, 0x24002628, 0x00008815, 0x00000044 }, + { 0x00800040, 0x24004508, 0x008d0400, 0x00008814 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d8800, 0x00000000 }, + { 0x00802001, 0xa8000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00600001, 0x21600022, 0x008d07e0, 0x00000000 }, + { 0x00802001, 0x21800022, 0x008d8800, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000040, 0x27e80c21, 0x000007e8, 0x00000040 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00400040 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xffffffda }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x01000005, 0x20602d29, 0x02000042, 0x00070007 }, + { 0x01000005, 0x20422d29, 0x00000042, 0xfff8fff8 }, + { 0x00600001, 0x27e00021, 0x008d0000, 0x00000000 }, + { 0x00000009, 0x27e82d21, 0x00000040, 0x00060006 }, + { 0x00600001, 0x20a00022, 0x008d0000, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000260 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x01800031, 0x20801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20400022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20480c22, 0x000007e8, 0x00000080 }, + { 0x02800031, 0x21001d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20600022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20680c22, 0x000007e8, 0x00000100 }, + { 0x03800031, 0x21801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20800022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20880c22, 0x000007e8, 0x00000180 }, + { 0x04800031, 0x22001d09, 0x408d0000, 0x02488400 }, + { 0x00200001, 0x20640229, 0x00450094, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000090, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000090, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000095, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000095, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000090, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x20881c21, 0x00000088, 0x00200000 }, + { 0x00200001, 0x20840129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0080, 0x00000000 }, + { 0x00802001, 0x20800061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000090, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000009c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000009c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200009c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000090, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200009c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200009c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x20843d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x20862d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x40840231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x004500d4, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x020000d0, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x000000d0, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x000000d5, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x000000d5, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x000000d0, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x20c81c21, 0x000000c8, 0x00200000 }, + { 0x00200001, 0x20c40129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d00c0, 0x00000000 }, + { 0x00802001, 0x20c00061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x000000d0, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x000000dc, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x000000dc, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x020000dc, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x020000d0, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x020000dc, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x020000dc, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x20c43d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x20c62d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x40c40231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450114, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000110, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000110, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000115, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000115, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000110, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x21081c21, 0x00000108, 0x00200000 }, + { 0x00200001, 0x21040129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0100, 0x00000000 }, + { 0x00802001, 0x21000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000110, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000011c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000011c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200011c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000110, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200011c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200011c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x21043d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x21062d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x41040231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450154, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000150, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000150, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000155, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000155, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000150, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x21481c21, 0x00000148, 0x00200000 }, + { 0x00200001, 0x21440129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0140, 0x00000000 }, + { 0x00802001, 0x21400061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000150, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000015c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000015c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200015c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000150, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200015c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200015c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x21443d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x21462d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x41440231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450194, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000190, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000190, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000195, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000195, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000190, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x21881c21, 0x00000188, 0x00200000 }, + { 0x00200001, 0x21840129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0180, 0x00000000 }, + { 0x00802001, 0x21800061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000190, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000019c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000019c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200019c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000190, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200019c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200019c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x21843d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x21862d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x41840231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x004501d4, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x020001d0, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x000001d0, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x000001d5, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x000001d5, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x000001d0, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x21c81c21, 0x000001c8, 0x00200000 }, + { 0x00200001, 0x21c40129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d01c0, 0x00000000 }, + { 0x00802001, 0x21c00061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x000001d0, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x000001dc, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x000001dc, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x020001dc, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x020001d0, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x020001dc, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x020001dc, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x21c43d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x21c62d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x41c40231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450214, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000210, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000210, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000215, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000215, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000210, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x22081c21, 0x00000208, 0x00200000 }, + { 0x00200001, 0x22040129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0200, 0x00000000 }, + { 0x00802001, 0x22000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000210, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000021c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000021c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200021c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000210, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200021c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200021c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x22043d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x22062d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x42040231, 0x00660064, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00450254, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02000250, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00000250, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00000255, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00000255, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00000250, 0x00002000 }, + { 0x02600005, 0x20001d20, 0x02000066, 0x00000001 }, + { 0x00000006, 0x22481c21, 0x00000248, 0x00200000 }, + { 0x00200001, 0x22440129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d0240, 0x00000000 }, + { 0x00802001, 0x22400061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x0000001e }, + { 0x02600005, 0x20000c20, 0x00000250, 0x00004000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00000001, 0x20680129, 0x0000025c, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000025c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200025c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02000250, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200025c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200025c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x02000068, 0x000d0070 }, + { 0x00000009, 0x22443d09, 0x00000602, 0x000c000c }, + { 0x00000005, 0x22462d09, 0x00000602, 0xf000f000 }, + { 0x00200001, 0x42440231, 0x00660064, 0x00000000 }, + { 0x01000040, 0x20423dad, 0x00000042, 0xfff8fff8 }, + { 0x00600001, 0x21600022, 0x008d07e0, 0x00000000 }, + { 0x00802001, 0x21800022, 0x008d0080, 0x00000000 }, + { 0x00802001, 0x21c00022, 0x008d00c0, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20280c22, 0x000007e8, 0x00000080 }, + { 0x00802001, 0x20400022, 0x008d0100, 0x00000000 }, + { 0x00802001, 0x20800022, 0x008d0140, 0x00000000 }, + { 0x01800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x21680c22, 0x000007e8, 0x00000100 }, + { 0x00802001, 0x21800022, 0x008d0180, 0x00000000 }, + { 0x00802001, 0x21c00022, 0x008d01c0, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x20280c22, 0x000007e8, 0x00000180 }, + { 0x00802001, 0x20400022, 0x008d0200, 0x00000000 }, + { 0x00802001, 0x20800022, 0x008d0240, 0x00000000 }, + { 0x01800031, 0x20001d0c, 0x508d0000, 0x0a080400 }, + { 0x00000040, 0x27e80c21, 0x000007e8, 0x00000200 }, + { 0x01000010, 0x20002da0, 0x02000060, 0x00000000 }, + { 0x00110220, 0x34001c00, 0x00001400, 0xfffffda0 }, + { 0x00010220, 0x34001c00, 0x02001400, 0x0000006e }, + { 0x00600001, 0x20200022, 0x008d07e0, 0x00000000 }, + { 0x01800031, 0x20801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20400022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20480c22, 0x000007e8, 0x00000080 }, + { 0x02800031, 0x21001d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20600022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20680c22, 0x000007e8, 0x00000100 }, + { 0x03800031, 0x21801d09, 0x408d0000, 0x02488400 }, + { 0x00600001, 0x20800022, 0x008d07e0, 0x00000000 }, + { 0x00000040, 0x20880c22, 0x000007e8, 0x00000180 }, + { 0x04800031, 0x22001d09, 0x408d0000, 0x02488400 }, + { 0x00000001, 0x220400e0, 0x00000000, 0x00a00080 }, + { 0x00600001, 0x20a00022, 0x008d0000, 0x00000000 }, + { 0x00200001, 0x20640229, 0x00458814, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x02008810, 0x01000000 }, + { 0x02600005, 0x20000c20, 0x00008810, 0x00004000 }, + { 0x00600041, 0x24003e2c, 0x00008815, 0x00020002 }, + { 0x00110001, 0x2066018d, 0x02000400, 0x00000000 }, + { 0x00010040, 0x20663d8d, 0x02000400, 0x00010001 }, + { 0x00110001, 0x2066022d, 0x00008815, 0x00000000 }, + { 0x02600005, 0x20000c20, 0x00008810, 0x00002000 }, + { 0x01000040, 0x20603dad, 0x02000060, 0xffffffff }, + { 0x00000006, 0xa8080c21, 0x00008808, 0x00200000 }, + { 0x00200001, 0xa8040129, 0x00450064, 0x00000000 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x00800041, 0x24002528, 0x00000066, 0x00000044 }, + { 0x00800040, 0x24002508, 0x008d0400, 0x00000064 }, + { 0x00000009, 0x20a82d02, 0x00000404, 0x00060006 }, + { 0x00802001, 0x20c00022, 0x008d8800, 0x00000000 }, + { 0x00802001, 0xa8000061, 0x00000000, 0x00000000 }, + { 0x05800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000220, 0x34001c00, 0x00001400, 0x00000022 }, + { 0x02600005, 0x20000d20, 0x02000066, 0x00000001 }, + { 0x02600005, 0x20000c20, 0x00008810, 0x00004000 }, + { 0x00000001, 0x20680129, 0x0000881c, 0x00000000 }, + { 0x00210002, 0x20700421, 0x02450038, 0x00450030 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x0000000e }, + { 0x02600005, 0x20001d20, 0x0000881c, 0x00000010 }, + { 0x00210002, 0x20700421, 0x02450028, 0x00450020 }, + { 0x01600005, 0x20001d20, 0x0200881c, 0x00000001 }, + { 0x00010220, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x02610005, 0x20001c20, 0x02008810, 0x00008000 }, + { 0x02610005, 0x20001d20, 0x0200881c, 0x00000010 }, + { 0x00010006, 0x20681d29, 0x0200881c, 0x00000010 }, + { 0x02800005, 0x2000452c, 0x00000068, 0x000d0070 }, + { 0x01000040, 0x20603dad, 0x02000060, 0x00000000 }, + { 0x00000009, 0xa8043d09, 0x00000600, 0x000c000c }, + { 0x00000005, 0xa8062d09, 0x00000600, 0xf000f000 }, + { 0x00200001, 0xc8040231, 0x00660064, 0x00000000 }, + { 0x00600001, 0x21600022, 0x008d07e0, 0x00000000 }, + { 0x00802001, 0x21800022, 0x008d8800, 0x00000000 }, + { 0x0b800031, 0x20001d0c, 0x508d0000, 0x06080300 }, + { 0x00000040, 0x27e80c21, 0x000007e8, 0x00000040 }, + { 0x00000040, 0x22040c00, 0x00000204, 0x00400040 }, + { 0x00110220, 0x34001c00, 0x02001400, 0xffffffac }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x01600031, 0x20001c20, 0x708d0000, 0x82000010 }, + { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 }, diff --git a/shaders/h264/mc/chromaMVAdjust.asm b/shaders/h264/mc/chromaMVAdjust.asm new file mode 100644 index 0000000..063f554 --- /dev/null +++ b/shaders/h264/mc/chromaMVAdjust.asm @@ -0,0 +1,27 @@ +/*
+ * Adjust chrom MV
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: ChromaMVAdjust.asm
+//
+//
+
+
+//#if !defined(__ChromaMVAdjust__) // Make sure this is only included once
+//#define __ChromaMVAdjust__
+
+
+ // Chroma MV adjustment
+ add (1) acc0:w gPARITY:w gREFPARITY:w
+ cmp.e.f0.0 (1) null:w acc0:w 0x1:w
+ cmp.e.f0.1 (1) null:w acc0:w 0x100:w
+ mov (1) gCHRMVADJ:w 0:w
+ (f0.0) mov (1) gCHRMVADJ:w 2:w
+ (f0.1) mov (1) gCHRMVADJ:w -2:w
+
+//#endif // !defined(__ChromaMVAdjust__)
diff --git a/shaders/h264/mc/export.inc b/shaders/h264/mc/export.inc new file mode 100644 index 0000000..1113841 --- /dev/null +++ b/shaders/h264/mc/export.inc @@ -0,0 +1,209 @@ +#define INTRA_16x16_IP 0 +#define INTRA_16x16_VERTICAL_IP 112 +#define INTRA_16x16_HORIZONTAL_IP 256 +#define INTRA_16x16_DC_IP 416 +#define INTRA_16x16_PLANE_IP 784 +#define End_intra_Pred_16x16_Y_IP 1328 +#define End_add_Error_16x16_Y_IP 1632 +#define load_Intra_Ref_Y_IP 1760 +#define decode_Chroma_Intra_IP 1904 +#define INTRA_CHROMA_DC_IP 2080 +#define INTRA_CHROMA_HORIZONTAL_IP 2560 +#define INTRA_CHROMA_VERTICAL_IP 2656 +#define INTRA_Chroma_PLANE_IP 2736 +#define End_of_intra_Pred_Chroma_IP 3136 +#define save_16x16_Y_IP 3488 +#define INTRA_8x8_IP 3712 +#define INTRA_8x8_BLK2_IP 4544 +#define intra_Pred_8x8_Y_IP 5120 +#define INTRA_8X8_VERTICAL_IP 5376 +#define INTRA_8X8_HORIZONTAL_IP 5456 +#define INTRA_8X8_DC_IP 5536 +#define INTRA_8X8_DIAG_DOWN_LEFT_IP 5792 +#define INTRA_8X8_DIAG_DOWN_RIGHT_IP 5952 +#define INTRA_8X8_VERT_RIGHT_IP 6176 +#define INTRA_8X8_HOR_DOWN_IP 6464 +#define INTRA_8X8_VERT_LEFT_IP 6736 +#define INTRA_8X8_HOR_UP_IP 6896 +#define save_8x8_Y_IP 7088 +#define INTRA_4x4_IP 7424 +#define intra_Pred_4x4_Y_4_IP 8496 +#define ADD_ERROR_SB0_IP 8592 +#define ADD_ERROR_SB1_IP 8704 +#define ADD_ERROR_SB2_IP 8864 +#define ADD_ERROR_SB3_IP 8992 +#define intra_Pred_4x4_Y_IP 9040 +#define INTRA_4X4_VERTICAL_IP 9040 +#define INTRA_4X4_HORIZONTAL_IP 9072 +#define INTRA_4X4_DC_IP 9104 +#define INTRA_4X4_DIAG_DOWN_LEFT_IP 9280 +#define INTRA_4X4_DIAG_DOWN_RIGHT_IP 9392 +#define INTRA_4X4_VERT_RIGHT_IP 9536 +#define INTRA_4X4_HOR_DOWN_IP 9744 +#define INTRA_4X4_VERT_LEFT_IP 9968 +#define INTRA_4X4_HOR_UP_IP 10080 +#define save_4x4_Y_IP 10208 +#define INTRA_PCM_IP 10560 +#define FRAME_MB_IP 11072 +#define INIT_MBPARA_FRM_IP 11120 +#define NOT_8x8_MODE_FRM_IP 11408 +#define CONVERT_MVS_FRM_IP 11488 +#define INIT_ADDRESS_REGS_FRM_IP 11568 +#define LOOP_SUBMB_FRM_IP 11632 +#define LOOP_DIR_FRM_IP 11680 +#define LOADREF_MVXZERO_FRM_IP 12080 +#define EXIT_LOADREF_Y_16x13_FRM_IP 12192 +#define Interpolate_Y_8x8_Func_FRM_IP 12352 +#define Interpolate_Y_8x8_Func2_FRM_IP 12592 +#define Interpolate_Y_H_8x8_FRM_IP 13664 +#define Interpolate_Y_V_8x8_FRM_IP 14320 +#define VFILTER_8x8_FRM_IP 14496 +#define Interpolate_Y_I_8x8_FRM_IP 14880 +#define Average_8x8_FRM_IP 15040 +#define Return_Interpolate_Y_8x8_FRM_IP 15104 +#define Exit_Interpolate_Y_8x8_FRM_IP 15120 +#define Interpolate_C_4x4_Func_FRM_IP 15120 +#define PROCESS4x4_FRM_IP 15424 +#define LOOP_SUBMBPT_FRM_IP 15440 +#define Interpolate_Y_H_4x4_FRM_IP 16528 +#define Interpolate_Y_V_4x4_FRM_IP 16864 +#define VFILTER_4x4_FRM_IP 17136 +#define Interpolate_Y_I_4x4_FRM_IP 17184 +#define Average_4x4_FRM_IP 17280 +#define Return_Interpolate_Y_4x4_FRM_IP 17296 +#define Exit_Interpolate_Y_4x4_FRM_IP 17392 +#define ROUND_SHIFT_C_FRM_IP 17776 +#define LOOP_DIR_CONTINUE_FRM_IP 17840 +#define Weighted_Prediction_FRM_IP 17888 +#define DefaultWeightedPred_UniPred_FRM_IP 17952 +#define DefaultWeightedPred_BiPred_FRM_IP 18048 +#define WeightedPred_FRM_IP 18112 +#define WeightedPred_Explicit_FRM_IP 18256 +#define WeightedPred_LOOP_FRM_IP 18576 +#define Return_WeightedPred_FRM_IP 19056 +#define EXIT_LOOP_FRM_IP 19392 +#define FIELD_MB_IP 19968 +#define INIT_MBPARA_FLD_IP 20016 +#define NOT_8x8_MODE_FLD_IP 20304 +#define CONVERT_MVS_FLD_IP 20384 +#define INIT_ADDRESS_REGS_FLD_IP 20464 +#define LOOP_SUBMB_FLD_IP 20544 +#define LOOP_DIR_FLD_IP 20592 +#define LOADREF_MVXZERO_FLD_IP 21152 +#define EXIT_LOADREF_Y_16x13_FLD_IP 21264 +#define Interpolate_Y_8x8_Func_FLD_IP 21440 +#define Interpolate_Y_8x8_Func2_FLD_IP 21680 +#define Interpolate_Y_H_8x8_FLD_IP 22752 +#define Interpolate_Y_V_8x8_FLD_IP 23408 +#define VFILTER_8x8_FLD_IP 23584 +#define Interpolate_Y_I_8x8_FLD_IP 23968 +#define Average_8x8_FLD_IP 24128 +#define Return_Interpolate_Y_8x8_FLD_IP 24192 +#define Exit_Interpolate_Y_8x8_FLD_IP 24208 +#define Interpolate_C_4x4_Func_FLD_IP 24208 +#define PROCESS4x4_FLD_IP 24512 +#define LOOP_SUBMBPT_FLD_IP 24528 +#define Interpolate_Y_H_4x4_FLD_IP 25632 +#define Interpolate_Y_V_4x4_FLD_IP 25968 +#define VFILTER_4x4_FLD_IP 26240 +#define Interpolate_Y_I_4x4_FLD_IP 26288 +#define Average_4x4_FLD_IP 26384 +#define Return_Interpolate_Y_4x4_FLD_IP 26400 +#define Exit_Interpolate_Y_4x4_FLD_IP 26496 +#define ROUND_SHIFT_C_FLD_IP 26880 +#define LOOP_DIR_CONTINUE_FLD_IP 26944 +#define Weighted_Prediction_FLD_IP 26992 +#define DefaultWeightedPred_UniPred_FLD_IP 27056 +#define DefaultWeightedPred_BiPred_FLD_IP 27152 +#define WeightedPred_FLD_IP 27216 +#define WeightedPred_Explicit_FLD_IP 27360 +#define WeightedPred_LOOP_FLD_IP 27680 +#define Return_WeightedPred_FLD_IP 28160 +#define EXIT_LOOP_FLD_IP 28496 +#define MBAFF_MB_IP 29120 +#define INIT_MBPARA_MBF_IP 29168 +#define NOT_8x8_MODE_MBF_IP 29456 +#define CONVERT_MVS_MBF_IP 29536 +#define INIT_ADDRESS_REGS_MBF_IP 29616 +#define LOOP_SUBMB_MBF_IP 29728 +#define LOOP_DIR_MBF_IP 29776 +#define LOADREF_MVXZERO_MBF_IP 30368 +#define EXIT_LOADREF_Y_16x13_MBF_IP 30480 +#define Interpolate_Y_8x8_Func_MBF_IP 30656 +#define Interpolate_Y_8x8_Func2_MBF_IP 30896 +#define Interpolate_Y_H_8x8_MBF_IP 31968 +#define Interpolate_Y_V_8x8_MBF_IP 32624 +#define VFILTER_8x8_MBF_IP 32800 +#define Interpolate_Y_I_8x8_MBF_IP 33184 +#define Average_8x8_MBF_IP 33344 +#define Return_Interpolate_Y_8x8_MBF_IP 33408 +#define Exit_Interpolate_Y_8x8_MBF_IP 33424 +#define Interpolate_C_4x4_Func_MBF_IP 33424 +#define PROCESS4x4_MBF_IP 33728 +#define LOOP_SUBMBPT_MBF_IP 33744 +#define Interpolate_Y_H_4x4_MBF_IP 34848 +#define Interpolate_Y_V_4x4_MBF_IP 35184 +#define VFILTER_4x4_MBF_IP 35456 +#define Interpolate_Y_I_4x4_MBF_IP 35504 +#define Average_4x4_MBF_IP 35600 +#define Return_Interpolate_Y_4x4_MBF_IP 35616 +#define Exit_Interpolate_Y_4x4_MBF_IP 35712 +#define ROUND_SHIFT_C_MBF_IP 36096 +#define LOOP_DIR_CONTINUE_MBF_IP 36160 +#define Weighted_Prediction_MBF_IP 36208 +#define DefaultWeightedPred_UniPred_MBF_IP 36272 +#define DefaultWeightedPred_BiPred_MBF_IP 36368 +#define WeightedPred_MBF_IP 36432 +#define WeightedPred_Explicit_MBF_IP 36576 +#define WeightedPred_LOOP_MBF_IP 36896 +#define Return_WeightedPred_MBF_IP 37376 +#define EXIT_LOOP_MBF_IP 37712 +#define SETHWSCOREBOARD_IP 38336 +#define SetHWScoreboard_Loop_IP 38448 +#define Parse_8_Loop_0_IP 38816 +#define Parse_8_Loop_2_IP 39008 +#define Parse_8_Loop_4_IP 39200 +#define Parse_8_Loop_6_IP 39392 +#define Parse_8_Loop_8_IP 39584 +#define Parse_8_Loop_10_IP 39776 +#define Parse_8_Loop_12_IP 39968 +#define Parse_8_Loop_14_IP 40160 +#define SetHWScoreboard_Remainder_IP 40496 +#define SetHWScoreboard_Remainder_Loop_IP 40720 +#define Output_Remainder_Intra_IP 40928 +#define SetHWScoreboard_Done_IP 41024 +#define SETHWSCOREBOARD_MBAFF_IP 41088 +#define SetHWScoreboard_MBAFF_Loop_IP 41184 +#define SET_SB_MBAFF_INTRA_0_IP 41664 +#define SET_SB_MBAFF_0_IP 41840 +#define NEXT_MB_MBAFF_0_IP 41904 +#define SET_SB_MBAFF_INTRA_2_IP 42208 +#define SET_SB_MBAFF_2_IP 42384 +#define NEXT_MB_MBAFF_2_IP 42448 +#define SET_SB_MBAFF_INTRA_4_IP 42752 +#define SET_SB_MBAFF_4_IP 42928 +#define NEXT_MB_MBAFF_4_IP 42992 +#define SET_SB_MBAFF_INTRA_6_IP 43296 +#define SET_SB_MBAFF_6_IP 43472 +#define NEXT_MB_MBAFF_6_IP 43536 +#define SET_SB_MBAFF_INTRA_8_IP 43840 +#define SET_SB_MBAFF_8_IP 44016 +#define NEXT_MB_MBAFF_8_IP 44080 +#define SET_SB_MBAFF_INTRA_10_IP 44384 +#define SET_SB_MBAFF_10_IP 44560 +#define NEXT_MB_MBAFF_10_IP 44624 +#define SET_SB_MBAFF_INTRA_12_IP 44928 +#define SET_SB_MBAFF_12_IP 45104 +#define NEXT_MB_MBAFF_12_IP 45168 +#define SET_SB_MBAFF_INTRA_14_IP 45472 +#define SET_SB_MBAFF_14_IP 45648 +#define NEXT_MB_MBAFF_14_IP 45712 +#define SetHWScoreboard_MBAFF_Remainder_IP 46048 +#define SetHWScoreboard_MBAFF_Remainder_Loop_IP 46272 +#define SET_SB_MBAFF_REM_INTRA_IP 46576 +#define SET_SB_MBAFF_REM_IP 46768 +#define Output_MBAFF_Remainder_Intra_IP 46848 +#define SetHWScoreboard_MBAFF_Done_IP 46944 +#define BSDRESET_IP 46960 +#define DCRESETDUMMY_IP 46976 +#define AllAVC_END_IP 46992 diff --git a/shaders/h264/mc/export.inc.gen5 b/shaders/h264/mc/export.inc.gen5 new file mode 100644 index 0000000..63accd9 --- /dev/null +++ b/shaders/h264/mc/export.inc.gen5 @@ -0,0 +1,209 @@ +#define INTRA_16x16_IP_GEN5 0 +#define INTRA_16x16_VERTICAL_IP_GEN5 112 +#define INTRA_16x16_HORIZONTAL_IP_GEN5 256 +#define INTRA_16x16_DC_IP_GEN5 416 +#define INTRA_16x16_PLANE_IP_GEN5 784 +#define End_intra_Pred_16x16_Y_IP_GEN5 1328 +#define End_add_Error_16x16_Y_IP_GEN5 1632 +#define load_Intra_Ref_Y_IP_GEN5 1760 +#define decode_Chroma_Intra_IP_GEN5 1904 +#define INTRA_CHROMA_DC_IP_GEN5 2080 +#define INTRA_CHROMA_HORIZONTAL_IP_GEN5 2560 +#define INTRA_CHROMA_VERTICAL_IP_GEN5 2656 +#define INTRA_Chroma_PLANE_IP_GEN5 2736 +#define End_of_intra_Pred_Chroma_IP_GEN5 3136 +#define save_16x16_Y_IP_GEN5 3488 +#define INTRA_8x8_IP_GEN5 3712 +#define INTRA_8x8_BLK2_IP_GEN5 4544 +#define intra_Pred_8x8_Y_IP_GEN5 5120 +#define INTRA_8X8_VERTICAL_IP_GEN5 5376 +#define INTRA_8X8_HORIZONTAL_IP_GEN5 5456 +#define INTRA_8X8_DC_IP_GEN5 5536 +#define INTRA_8X8_DIAG_DOWN_LEFT_IP_GEN5 5792 +#define INTRA_8X8_DIAG_DOWN_RIGHT_IP_GEN5 5952 +#define INTRA_8X8_VERT_RIGHT_IP_GEN5 6176 +#define INTRA_8X8_HOR_DOWN_IP_GEN5 6464 +#define INTRA_8X8_VERT_LEFT_IP_GEN5 6736 +#define INTRA_8X8_HOR_UP_IP_GEN5 6896 +#define save_8x8_Y_IP_GEN5 7088 +#define INTRA_4x4_IP_GEN5 7424 +#define intra_Pred_4x4_Y_4_IP_GEN5 8496 +#define ADD_ERROR_SB0_IP_GEN5 8592 +#define ADD_ERROR_SB1_IP_GEN5 8704 +#define ADD_ERROR_SB2_IP_GEN5 8864 +#define ADD_ERROR_SB3_IP_GEN5 8992 +#define intra_Pred_4x4_Y_IP_GEN5 9040 +#define INTRA_4X4_VERTICAL_IP_GEN5 9040 +#define INTRA_4X4_HORIZONTAL_IP_GEN5 9072 +#define INTRA_4X4_DC_IP_GEN5 9104 +#define INTRA_4X4_DIAG_DOWN_LEFT_IP_GEN5 9280 +#define INTRA_4X4_DIAG_DOWN_RIGHT_IP_GEN5 9392 +#define INTRA_4X4_VERT_RIGHT_IP_GEN5 9536 +#define INTRA_4X4_HOR_DOWN_IP_GEN5 9744 +#define INTRA_4X4_VERT_LEFT_IP_GEN5 9968 +#define INTRA_4X4_HOR_UP_IP_GEN5 10080 +#define save_4x4_Y_IP_GEN5 10208 +#define INTRA_PCM_IP_GEN5 10560 +#define FRAME_MB_IP_GEN5 11072 +#define INIT_MBPARA_FRM_IP_GEN5 11120 +#define NOT_8x8_MODE_FRM_IP_GEN5 11408 +#define CONVERT_MVS_FRM_IP_GEN5 11488 +#define INIT_ADDRESS_REGS_FRM_IP_GEN5 11568 +#define LOOP_SUBMB_FRM_IP_GEN5 11632 +#define LOOP_DIR_FRM_IP_GEN5 11680 +#define LOADREF_MVXZERO_FRM_IP_GEN5 12080 +#define EXIT_LOADREF_Y_16x13_FRM_IP_GEN5 12192 +#define Interpolate_Y_8x8_Func_FRM_IP_GEN5 12352 +#define Interpolate_Y_8x8_Func2_FRM_IP_GEN5 12592 +#define Interpolate_Y_H_8x8_FRM_IP_GEN5 13664 +#define Interpolate_Y_V_8x8_FRM_IP_GEN5 14320 +#define VFILTER_8x8_FRM_IP_GEN5 14496 +#define Interpolate_Y_I_8x8_FRM_IP_GEN5 14880 +#define Average_8x8_FRM_IP_GEN5 15040 +#define Return_Interpolate_Y_8x8_FRM_IP_GEN5 15104 +#define Exit_Interpolate_Y_8x8_FRM_IP_GEN5 15120 +#define Interpolate_C_4x4_Func_FRM_IP_GEN5 15120 +#define PROCESS4x4_FRM_IP_GEN5 15424 +#define LOOP_SUBMBPT_FRM_IP_GEN5 15440 +#define Interpolate_Y_H_4x4_FRM_IP_GEN5 16528 +#define Interpolate_Y_V_4x4_FRM_IP_GEN5 16864 +#define VFILTER_4x4_FRM_IP_GEN5 17136 +#define Interpolate_Y_I_4x4_FRM_IP_GEN5 17184 +#define Average_4x4_FRM_IP_GEN5 17280 +#define Return_Interpolate_Y_4x4_FRM_IP_GEN5 17296 +#define Exit_Interpolate_Y_4x4_FRM_IP_GEN5 17392 +#define ROUND_SHIFT_C_FRM_IP_GEN5 17776 +#define LOOP_DIR_CONTINUE_FRM_IP_GEN5 17840 +#define Weighted_Prediction_FRM_IP_GEN5 17888 +#define DefaultWeightedPred_UniPred_FRM_IP_GEN5 17952 +#define DefaultWeightedPred_BiPred_FRM_IP_GEN5 18048 +#define WeightedPred_FRM_IP_GEN5 18112 +#define WeightedPred_Explicit_FRM_IP_GEN5 18256 +#define WeightedPred_LOOP_FRM_IP_GEN5 18576 +#define Return_WeightedPred_FRM_IP_GEN5 19056 +#define EXIT_LOOP_FRM_IP_GEN5 19392 +#define FIELD_MB_IP_GEN5 19968 +#define INIT_MBPARA_FLD_IP_GEN5 20016 +#define NOT_8x8_MODE_FLD_IP_GEN5 20304 +#define CONVERT_MVS_FLD_IP_GEN5 20384 +#define INIT_ADDRESS_REGS_FLD_IP_GEN5 20464 +#define LOOP_SUBMB_FLD_IP_GEN5 20544 +#define LOOP_DIR_FLD_IP_GEN5 20592 +#define LOADREF_MVXZERO_FLD_IP_GEN5 21152 +#define EXIT_LOADREF_Y_16x13_FLD_IP_GEN5 21264 +#define Interpolate_Y_8x8_Func_FLD_IP_GEN5 21440 +#define Interpolate_Y_8x8_Func2_FLD_IP_GEN5 21680 +#define Interpolate_Y_H_8x8_FLD_IP_GEN5 22752 +#define Interpolate_Y_V_8x8_FLD_IP_GEN5 23408 +#define VFILTER_8x8_FLD_IP_GEN5 23584 +#define Interpolate_Y_I_8x8_FLD_IP_GEN5 23968 +#define Average_8x8_FLD_IP_GEN5 24128 +#define Return_Interpolate_Y_8x8_FLD_IP_GEN5 24192 +#define Exit_Interpolate_Y_8x8_FLD_IP_GEN5 24208 +#define Interpolate_C_4x4_Func_FLD_IP_GEN5 24208 +#define PROCESS4x4_FLD_IP_GEN5 24512 +#define LOOP_SUBMBPT_FLD_IP_GEN5 24528 +#define Interpolate_Y_H_4x4_FLD_IP_GEN5 25632 +#define Interpolate_Y_V_4x4_FLD_IP_GEN5 25968 +#define VFILTER_4x4_FLD_IP_GEN5 26240 +#define Interpolate_Y_I_4x4_FLD_IP_GEN5 26288 +#define Average_4x4_FLD_IP_GEN5 26384 +#define Return_Interpolate_Y_4x4_FLD_IP_GEN5 26400 +#define Exit_Interpolate_Y_4x4_FLD_IP_GEN5 26496 +#define ROUND_SHIFT_C_FLD_IP_GEN5 26880 +#define LOOP_DIR_CONTINUE_FLD_IP_GEN5 26944 +#define Weighted_Prediction_FLD_IP_GEN5 26992 +#define DefaultWeightedPred_UniPred_FLD_IP_GEN5 27056 +#define DefaultWeightedPred_BiPred_FLD_IP_GEN5 27152 +#define WeightedPred_FLD_IP_GEN5 27216 +#define WeightedPred_Explicit_FLD_IP_GEN5 27360 +#define WeightedPred_LOOP_FLD_IP_GEN5 27680 +#define Return_WeightedPred_FLD_IP_GEN5 28160 +#define EXIT_LOOP_FLD_IP_GEN5 28496 +#define MBAFF_MB_IP_GEN5 29120 +#define INIT_MBPARA_MBF_IP_GEN5 29168 +#define NOT_8x8_MODE_MBF_IP_GEN5 29456 +#define CONVERT_MVS_MBF_IP_GEN5 29536 +#define INIT_ADDRESS_REGS_MBF_IP_GEN5 29616 +#define LOOP_SUBMB_MBF_IP_GEN5 29728 +#define LOOP_DIR_MBF_IP_GEN5 29776 +#define LOADREF_MVXZERO_MBF_IP_GEN5 30368 +#define EXIT_LOADREF_Y_16x13_MBF_IP_GEN5 30480 +#define Interpolate_Y_8x8_Func_MBF_IP_GEN5 30656 +#define Interpolate_Y_8x8_Func2_MBF_IP_GEN5 30896 +#define Interpolate_Y_H_8x8_MBF_IP_GEN5 31968 +#define Interpolate_Y_V_8x8_MBF_IP_GEN5 32624 +#define VFILTER_8x8_MBF_IP_GEN5 32800 +#define Interpolate_Y_I_8x8_MBF_IP_GEN5 33184 +#define Average_8x8_MBF_IP_GEN5 33344 +#define Return_Interpolate_Y_8x8_MBF_IP_GEN5 33408 +#define Exit_Interpolate_Y_8x8_MBF_IP_GEN5 33424 +#define Interpolate_C_4x4_Func_MBF_IP_GEN5 33424 +#define PROCESS4x4_MBF_IP_GEN5 33728 +#define LOOP_SUBMBPT_MBF_IP_GEN5 33744 +#define Interpolate_Y_H_4x4_MBF_IP_GEN5 34848 +#define Interpolate_Y_V_4x4_MBF_IP_GEN5 35184 +#define VFILTER_4x4_MBF_IP_GEN5 35456 +#define Interpolate_Y_I_4x4_MBF_IP_GEN5 35504 +#define Average_4x4_MBF_IP_GEN5 35600 +#define Return_Interpolate_Y_4x4_MBF_IP_GEN5 35616 +#define Exit_Interpolate_Y_4x4_MBF_IP_GEN5 35712 +#define ROUND_SHIFT_C_MBF_IP_GEN5 36096 +#define LOOP_DIR_CONTINUE_MBF_IP_GEN5 36160 +#define Weighted_Prediction_MBF_IP_GEN5 36208 +#define DefaultWeightedPred_UniPred_MBF_IP_GEN5 36272 +#define DefaultWeightedPred_BiPred_MBF_IP_GEN5 36368 +#define WeightedPred_MBF_IP_GEN5 36432 +#define WeightedPred_Explicit_MBF_IP_GEN5 36576 +#define WeightedPred_LOOP_MBF_IP_GEN5 36896 +#define Return_WeightedPred_MBF_IP_GEN5 37376 +#define EXIT_LOOP_MBF_IP_GEN5 37712 +#define SETHWSCOREBOARD_IP_GEN5 38336 +#define SetHWScoreboard_Loop_IP_GEN5 38448 +#define Parse_8_Loop_0_IP_GEN5 38816 +#define Parse_8_Loop_2_IP_GEN5 39008 +#define Parse_8_Loop_4_IP_GEN5 39200 +#define Parse_8_Loop_6_IP_GEN5 39392 +#define Parse_8_Loop_8_IP_GEN5 39584 +#define Parse_8_Loop_10_IP_GEN5 39776 +#define Parse_8_Loop_12_IP_GEN5 39968 +#define Parse_8_Loop_14_IP_GEN5 40160 +#define SetHWScoreboard_Remainder_IP_GEN5 40496 +#define SetHWScoreboard_Remainder_Loop_IP_GEN5 40720 +#define Output_Remainder_Intra_IP_GEN5 40928 +#define SetHWScoreboard_Done_IP_GEN5 41024 +#define SETHWSCOREBOARD_MBAFF_IP_GEN5 41088 +#define SetHWScoreboard_MBAFF_Loop_IP_GEN5 41184 +#define SET_SB_MBAFF_INTRA_0_IP_GEN5 41664 +#define SET_SB_MBAFF_0_IP_GEN5 41840 +#define NEXT_MB_MBAFF_0_IP_GEN5 41904 +#define SET_SB_MBAFF_INTRA_2_IP_GEN5 42208 +#define SET_SB_MBAFF_2_IP_GEN5 42384 +#define NEXT_MB_MBAFF_2_IP_GEN5 42448 +#define SET_SB_MBAFF_INTRA_4_IP_GEN5 42752 +#define SET_SB_MBAFF_4_IP_GEN5 42928 +#define NEXT_MB_MBAFF_4_IP_GEN5 42992 +#define SET_SB_MBAFF_INTRA_6_IP_GEN5 43296 +#define SET_SB_MBAFF_6_IP_GEN5 43472 +#define NEXT_MB_MBAFF_6_IP_GEN5 43536 +#define SET_SB_MBAFF_INTRA_8_IP_GEN5 43840 +#define SET_SB_MBAFF_8_IP_GEN5 44016 +#define NEXT_MB_MBAFF_8_IP_GEN5 44080 +#define SET_SB_MBAFF_INTRA_10_IP_GEN5 44384 +#define SET_SB_MBAFF_10_IP_GEN5 44560 +#define NEXT_MB_MBAFF_10_IP_GEN5 44624 +#define SET_SB_MBAFF_INTRA_12_IP_GEN5 44928 +#define SET_SB_MBAFF_12_IP_GEN5 45104 +#define NEXT_MB_MBAFF_12_IP_GEN5 45168 +#define SET_SB_MBAFF_INTRA_14_IP_GEN5 45472 +#define SET_SB_MBAFF_14_IP_GEN5 45648 +#define NEXT_MB_MBAFF_14_IP_GEN5 45712 +#define SetHWScoreboard_MBAFF_Remainder_IP_GEN5 46048 +#define SetHWScoreboard_MBAFF_Remainder_Loop_IP_GEN5 46272 +#define SET_SB_MBAFF_REM_INTRA_IP_GEN5 46576 +#define SET_SB_MBAFF_REM_IP_GEN5 46768 +#define Output_MBAFF_Remainder_Intra_IP_GEN5 46848 +#define SetHWScoreboard_MBAFF_Done_IP_GEN5 46944 +#define BSDRESET_IP_GEN5 46960 +#define DCRESETDUMMY_IP_GEN5 46976 +#define AllAVC_END_IP_GEN5 46992 diff --git a/shaders/h264/mc/header.inc b/shaders/h264/mc/header.inc new file mode 100644 index 0000000..4a0eecf --- /dev/null +++ b/shaders/h264/mc/header.inc @@ -0,0 +1,303 @@ +/*
+ * Common header file for all AVC MC kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__HEADER__) // Make sure this file is only included once
+#define __HEADER__
+
+// Module name: header.inc
+//
+// Common header file for all AVC MC kernels
+//
+
+#ifndef COMBINED_KERNEL
+#ifdef DEV_CTG
+ #define SW_SCOREBOARD // SW Scoreboard should be enabled for CTG and earlier
+ #undef HW_SCOREBOARD // HW Scoreboard should be disabled for CTG and earlier
+#else
+ #define HW_SCOREBOARD // HW Scoreboard should be enabled for ILK and beyond
+ #undef SW_SCOREBOARD // SW Scoreboard should be disabled for ILK and beyond
+#endif // DEV_CTG
+#endif // COMBINED_KERNEL
+
+//#define MONO // Build Monochrome kernels
+
+// Surface state definition
+//
+#define DESTY 0
+#define DESTUV 1
+#define REFYFM0 2
+#define REFYFM1 3
+#define REFYFM2 4
+#define REFYFM3 5
+#define REFYFM4 6
+#define REFYFM5 7
+#define REFYFM6 8
+#define REFYFM7 9
+#define REFYFM8 10
+#define REFYFM9 11
+#define REFYFM10 12
+#define REFYFM11 13
+#define REFYFM12 14
+#define REFYFM13 15
+#define REFYFM14 16
+#define REFYFM15 17
+#define REFUVFM0 18
+#define REFUVFM1 19
+#define REFUVFM2 20
+#define REFUVFM3 21
+#define REFUVFM4 22
+#define REFUVFM5 23
+#define REFUVFM6 24
+#define REFUVFM7 25
+#define REFUVFM8 26
+#define REFUVFM9 27
+#define REFUVFM10 28
+#define REFUVFM11 29
+#define REFUVFM12 30
+#define REFUVFM13 31
+#define REFUVFM14 32
+#define REFUVFM15 33
+
+.default_execution_size (16)
+.default_register_type :ub
+
+// ----------- Common constant definitions ------------
+//
+// Bit position constants
+//
+#define BIT0 0x01
+#define BIT1 0x02
+#define BIT2 0x04
+#define BIT3 0x08
+#define BIT4 0x10
+#define BIT5 0x20
+#define BIT6 0x40
+#define BIT7 0x80
+#define BIT8 0x0100
+#define BIT9 0x0200
+#define BIT10 0x0400
+#define BIT11 0x0800
+#define BIT12 0x1000
+#define BIT13 0x2000
+#define BIT14 0x4000
+#define BIT15 0x8000
+#define BIT16 0x00010000
+#define BIT17 0x00020000
+#define BIT18 0x00040000
+#define BIT19 0x00080000
+#define BIT20 0x00100000
+#define BIT21 0x00200000
+#define BIT22 0x00400000
+#define BIT23 0x00800000
+#define BIT24 0x01000000
+#define BIT25 0x02000000
+#define BIT26 0x04000000
+#define BIT27 0x08000000
+#define BIT28 0x10000000
+#define BIT29 0x20000000
+#define BIT30 0x40000000
+#define BIT31 0x80000000
+
+#define GRFWIB 32 // GRF register width in byte
+#define GRFWIW 16 // GRF register width in word
+#define GRFWID 8 // GRF register width in dword
+
+#define INST_SIZE 16 // Instruction size = 128b = 16 Bytes
+
+#define REGION(Width,HStride) <Width*HStride;Width,HStride>
+
+#define NULLREG null<1>:ud
+#define NULLREGW null<1>:w
+
+#define TOP_FIELD 0
+#define BOTTOM_FIELD 1
+
+// M2 - M9 for date writing message payload
+.declare MSGPAYLOAD Base=m2 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare MSGPAYLOADB Base=m2 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare MSGPAYLOADW Base=m2 ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+.declare MSGPAYLOADD Base=m2 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+
+// ----------- Common Message Descriptor ------------
+//
+#ifdef DEV_ILK
+#define MSG_GW 0x03 // Message Gateway Extended Message Descriptor,
+#define DAPREAD 0x04 // Data Port Read Extended Message Descriptor,
+#define DAPWRITE 0x05 // Data Port Write Extended Message Descriptor,
+#define TS 0x07 // Thread Spawner Extended Message Descriptor
+#define TS_EOT 0x27 // End of Thread Extended Message Descriptor
+
+#define EOTMSGDSC 0x02000010 // End of Thread Message Descriptor, don't deference URB handle
+
+// Data Port Message Descriptor
+#define DWBRMSGDSC_RC 0x02086000 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_RC_TF 0x02086600 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_RC_BF 0x02086700 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_SC 0x0208A000 // DWORD Block Read Message Descriptor, reading from sampler cache = A.
+#define DWBRMSGDSC_SC_TF 0x0208E600 // DWORD Block Read Message Descriptor, reading top field from field mode sampler cache.
+#define DWBRMSGDSC_SC_BF 0x0208E700 // DWORD Block Read Message Descriptor, reading bottom field from field mode sampler cache.
+
+#define DWBWMSGDSC 0x02082000 // DWORD Block Write Message Descriptor
+#define DWBWMSGDSC_WC 0x0218A000 // DWORD Block Write Message Descriptor + write commit
+
+// Enable Write Commit writeback mesage
+#define ENWRCOM 0x00108000 // Enable "write commit" and set response length = 1
+
+// Thread Spawner Message Descriptor
+#define TSMSGDSC 0x02000011
+
+// Message Gateway Message Descriptors
+#define OGWMSGDSC 0x02000000 // OpenGateway Message Descriptor
+#define CGWMSGDSC 0x02000001 // CloseGateway Message Descriptor
+#define FWDMSGDSC 0x02000002 // ForwardMsg Message Descriptor
+
+#define NOTIFYMSG 0x00008000 // Send notification with ForwardMsg message
+
+#define RESP_LEN(len) 0x100000*len
+#define MSG_LEN(len) 0x2000000*len
+
+#else // Pre DEV_ILK
+
+#define MSG_GW
+#define DAPREAD
+#define DAPWRITE
+#define TS
+#define TS_EOT
+
+#define EOTMSGDSC 0x87100010 // End of Thread Message Descriptor, don't deference URB handle
+
+// Data Port Message Descriptor
+#define DWBRMSGDSC_RC 0x04106000 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_RC_TF 0x04106600 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_RC_BF 0x04106700 // DWORD Block Read Message Descriptor, reading from render cache = 6.
+#define DWBRMSGDSC_SC 0x0410A000 // DWORD Block Read Message Descriptor, reading from sampler cache = A.
+#define DWBRMSGDSC_SC_TF 0x0410A600 // DWORD Block Read Message Descriptor, reading top field from field mode sampler cache.
+#define DWBRMSGDSC_SC_BF 0x0410A700 // DWORD Block Read Message Descriptor, reading bottom field from field mode sampler cache.
+
+#define DWBWMSGDSC 0x05102000 // DWORD Block Write Message Descriptor
+#define DWBWMSGDSC_WC 0x0511A000 // DWORD Block Write Message Descriptor + write commit
+
+// Enable Write Commit writeback mesage
+#define ENWRCOM 0x00018000 // Enable "write commit" and set response length = 1
+
+// Thread Spawner Message Descriptor
+#define TSMSGDSC 0x07100011
+
+// Message Gateway Message Descriptors
+#define OGWMSGDSC 0x03100000 // OpenGateway Message Descriptor
+#define CGWMSGDSC 0x03100001 // CloseGateway Message Descriptor
+#define FWDMSGDSC 0x03100002 // ForwardMsg Message Descriptor
+
+#define NOTIFYMSG 0x00008000 // Send notification with ForwardMsg message
+#define ACKREQMSG 0x00014000 // Acknowledgement required so response length should be 1
+
+#define RESP_LEN(len) 0x10000*len
+#define MSG_LEN(len) 0x100000*len
+
+#endif // DEV_ILK
+
+// Enable frame/field selection in message descriptor
+#define ENMSGDSCFM 0x400 // Enable MSGDSC to select frame surface
+#define ENMSGDSCTF 0x600 // Enable MSGDSC to select top field surface
+#define ENMSGDSCBF 0x700 // Enable MSGDSC to select bottom field surface
+
+// ----------- Message related register ------------
+//
+#define MSGHDR m1 // Message Payload Header
+#define MSGHDRY m1 // Message Payload Header register for Y data
+#define MSGHDRY0 m1 // Message Payload Header register for Y data
+#define MSGHDRY1 m2 // Message Payload Header register for Y data
+#define MSGHDRY2 m3 // Message Payload Header register for Y data
+#define MSGHDRY3 m4 // Message Payload Header register for Y data
+#define MSGHDRUV m5 // Message Payload Header register for U/V data
+#define MSGSRC r62 // Message source register, should never be used for other purposes
+#define MSGDSC a0.0:ud // Message Descriptor register (type DWORD)
+
+#define MH_ORI MSGSRC.0 // DWORD block R/W message header block offset
+#define MH_ORIX MSGSRC.0 // DWORD block R/W message header X offset
+#define MH_ORIY MSGSRC.1 // DWORD block R/W message header Y offset
+#define MH_SIZE MSGSRC.2 // DWORD block R/W message header block width & height
+
+// Data necessary for kernel operations
+//
+// Address registers used as pointers
+//
+// Note: Please keep the register order as is since they are used in compressed instructions
+//
+#define PPREDBUF_Y a0.4 // Pointer to predicted Y picture
+#define PPREDBUF_Y1 a0.5 // Pointer to predicted Y picture for extended instruction
+
+#define PPREDBUF_UV a0.4 // Pointer to predicted U/V picture
+#define PPREDBUF_UV1 a0.5 // Pointer to predicted U/V picture for extended instruction
+
+#define PDECBUF a0.4 // Pointer to decoded picture data
+#define PDECBUF_UD a0.2 // Pointer to decoded picture data in DWORD unit
+
+// ----------- R63 is reserved for global variables ------------
+//
+// Note: Don't program it with values other than what are defined here.
+
+#define G_REG r63
+
+#define RETURN_REG G_REG.0 // Return pointer for all sub-routine calls (type DWORD)
+#define RETURN_REG1 G_REG.1 // Return pointer for second-level calls
+
+#define I_ORIX G_REG.13 // :uw, H. origin of the macroblock in pixel unit, don't overwrite in-line data
+#define I_ORIY G_REG.14 // :uw, V. origin of the macroblock in pixel unit, don't overwrite in-line data
+
+// Macros
+//
+// Note: For macros that require multiple line expansion, insert "\n" at the end of each line.
+//
+#define GRF(reg) r##reg
+#ifdef DEV_ILK
+#define END_THREAD send (8) NULLREG MSGHDR r0:ud TS_EOT EOTMSGDSC
+#else
+#define END_THREAD send (8) NULLREG MSGHDR r0:ud EOTMSGDSC
+#endif // DEV_ILK
+
+#define CALL(subFunc, skipInst) add (1) RETURN_REG<1>:ud ip:ud (1+skipInst)*INST_SIZE \n\
+ jmpi (1) subFunc
+
+#define CALL_1(subFunc, skipInst) add (1) RETURN_REG1<1>:ud ip:ud (1+skipInst)*INST_SIZE \n\
+ jmpi (1) subFunc
+
+#define RETURN mov (1) ip:ud RETURN_REG<0;1,0>:ud // Return to calling module
+#define RETURN_1 mov (1) ip:ud RETURN_REG1<0;1,0>:ud // Return to second-level calling module
+ // To support iterative calling
+#ifdef SW_SCOREBOARD
+
+#ifdef DEV_CTG_A
+ #define LEADING_THREAD 1 // For CTG A, no SRT is needed. Only PRT is necessary
+#else
+ #define LEADING_THREAD 0 // For CTG B0 and beyond, PRT doesn't take into debug count
+ #define DOUBLE_SB // Scoreboard size needs to be doubled
+#endif
+
+#ifdef DOUBLE_SB // Scoreboard size needs to be doubled
+ #define SB_MASK 0x1ff // Scoreboard wrap-around mask (for 512 entries)
+#else
+ #define SB_MASK 0xff // Scoreboard wrap-around mask (for 256 entries)
+#endif // defined(DOUBLE_SB)
+
+// Scoreboard related definitions
+
+#define TEMPX r50
+#define TEMPY r51
+#define DELTA r52
+
+#define M05_STORE r0.13 // :uw, reuse r0.6:ud upper-word to store M0.5 header information for scoreboard
+
+
+#endif // SW_SCOREBOARD
+
+// End of header.inc
+
+#endif // !defined(__HEADER__)
+
diff --git a/shaders/h264/mc/initialize_MBPara.asm b/shaders/h264/mc/initialize_MBPara.asm new file mode 100644 index 0000000..bd651cf --- /dev/null +++ b/shaders/h264/mc/initialize_MBPara.asm @@ -0,0 +1,125 @@ +/*
+ * Initialize parameters
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Initialize_MBPara.asm
+//
+
+
+//#if !defined(__INITIALIZE_MBPARA__) // Make sure this is only included once
+//#define __INITIALIZE_MBPARA__
+
+
+// WA for weighted prediction - 2007/09/06 // shlee
+// mov (1) guwW128(0)<1> guwR1(0)<0;1,0> // Copy the unique number indicating weight/offset=(128,0)
+
+
+
+// MB Type Category
+// 1 B_L0_16x16
+// 2 B_L1_16x16
+// 3 B_Bi_16x16
+// 4 B_L0_L0_16x8
+// 5 B_L0_L0_8x16
+// 6 B_L1_L1_16x8
+// 7 B_L1_L1_8x16
+// 8 B_L0_L1_16x8
+// 9 B_L0_L1_8x16
+// 10 B_L1_L0_16x8
+// 11 B_L1_L0_8x16
+// 12 B_L0_Bi_16x8
+// 13 B_L0_Bi_8x16
+// 14 B_L1_Bi_16x8
+// 15 B_L1_Bi_8x16
+// 16 B_Bi_L0_16x8
+// 17 B_Bi_L0_8x16
+// 18 B_Bi_L1_16x8
+// 19 B_Bi_L1_8x16
+// 20 B_Bi_Bi_16x8
+// 21 B_Bi_Bi_8x16
+// 22 B_8x8
+
+ // TODO:
+ // Initialize interpolation area to eliminate uninitialized registers making the results of mac instructions XX.
+ // This issue was reported by Sharath on 5/25/2006, and why multiplication by zero still yields XX has not been understood yet.
+#if 0
+ mov (16) gudINTPY0(0)<1> 0:ud {Compr}
+ mov (16) gudINTPY0(2)<1> 0:ud {Compr}
+ mov (16) gudINTPY1(0)<1> 0:ud {Compr}
+ mov (16) gudINTPY1(2)<1> 0:ud {Compr}
+ mov (16) gudINTPC0(0)<1> 0:ud {Compr}
+ mov (16) gudINTPC1(0)<1> 0:ud {Compr}
+#endif
+
+ mov (1) gMVSTEP:w 0:w // Address increament for MV read
+
+ cmp.e.f0.0 (1) null:w gwMBTYPE<0;1,0> 22:w
+ (-f0.0) jmpi INTERLABEL(NOT_8x8_MODE)
+
+ //--- 8x8 mode
+
+ // Starting address of error data blocks
+ cmp.e.f0.1 (2) null<1>:w gSUBMB_SHAPE<0;1,0>:ub 0:w
+ (f0.1) jmpi INTERLABEL(CONVERT_MVS)
+
+ // Note: MVs and Weights/Offsets are already expanded by HW or driver
+
+ // MV conversion - Convert each MV to absolute coord. (= MV + MB org. + block offset)
+ shl (16) gwTEMP(0)<1> gX<0;2,1>:w 2:w // Convert MB origin to 1/4-pel unit
+ mov (1) gMVSTEP:w 24:w // Address increament for MV read
+ add (2) gwTEMP(0,4)<2> gwTEMP(0,4)<4;2,2> 16:w
+ add (2) gwTEMP(0,9)<2> gwTEMP(0,9)<4;2,2> 16:w
+ add (4) gwTEMP(0,12)<1> gwTEMP(0,12)<4;4,1> 16:w
+
+ add (16) gMV<1>:w gMV<16;16,1>:w gwTEMP(0)<16;16,1>
+ add (8) gwTEMP(0)<2> gwTEMP(0)<16;8,2> 32:w
+ add (16) gwMV(1,0)<1> gwMV(1,0)<16;16,1> gwTEMP(0)<16;16,1>
+ add (8) gwTEMP(0,1)<2> gwTEMP(0,1)<16;8,2> 32:w
+ add (16) gwMV(3,0)<1> gwMV(3,0)<16;16,1> gwTEMP(0)<16;16,1>
+ add (8) gwTEMP(0)<2> gwTEMP(0)<16;8,2> -32:w
+ add (16) gwMV(2,0)<1> gwMV(2,0)<16;16,1> gwTEMP(0)<16;16,1>
+
+ jmpi INTERLABEL(INIT_ADDRESS_REGS)
+
+INTERLABEL(NOT_8x8_MODE):
+
+ //--- !8x8 mode (16x16, 16x8, 8x16)
+
+ // MVs and Weights/Offsets are expanded
+ cmp.le.f0.1 (8) null<1>:w gwMBTYPE<0;1,0> 3:w // Check 16x16
+ mov (1) gSUBMB_SHAPE:ub 0:uw // subMB shape
+ (f0.1) mov (8) gMV<1>:d gMV<0;2,1>:d
+ (f0.1) mov (8) gdWGT(1,0)<1> gWGT<0;4,1>:d
+ (f0.1) mov (4) gdWGT(0,4)<1> gWGT<4;4,1>:d
+
+INTERLABEL(CONVERT_MVS):
+ // MV conversion - Convert each MV to absolute coord. (= MV + MB org. + block offset)
+ shl (2) gwTEMP(0)<1> gX<2;2,1>:w 2:w // Convert MB origin to 1/4-pel unit
+ add (16) gMV<1>:w gMV<16;16,1>:w gwTEMP(0)<0;2,1>
+ add (2) gwMV(0,4)<2> gwMV(0,4)<4;2,2> 32:w //{NoDDClr}
+ add (2) gwMV(0,9)<2> gwMV(0,9)<4;2,2> 32:w //{NoDDChk,NoDDClr}
+ add (4) gwMV(0,12)<1> gwMV(0,12)<4;4,1> 32:w //{NoDDChk}
+
+INTERLABEL(INIT_ADDRESS_REGS):
+ // Initialize the address registers
+ mov (2) pERRORYC:ud nOFFSET_ERROR:ud {NoDDClr} // Address of Y and C error blocks
+ mov (1) pRECON_MV:ud nOFFSET_RECON_MV:ud {NoDDChk,NoDDClr} // Address of recon area and motion vectors
+ mov (1) pWGT_BIDX:ud nOFFSET_WGT_BIDX:ud {NoDDChk} // Address of weights/offsets and binding tbl idx
+
+ // Read the parity of the current field (gPARITY - 0:top, 1:bottom, 3:frame)
+ // and set message descriptor for frame/field write
+#if defined(MBAFF)
+ and.nz.f0.0 (1) null:uw gFIELDMBFLAG:ub nFIELDMB_MASK:uw
+ (f0.0) and (1) gPARITY:uw gMBPARITY:ub nMBPARITY_MASK:uw
+ (-f0.0) mov (1) gPARITY:uw 3:uw
+#elif defined(FIELD)
+ and (1) gPARITY:uw gMBPARITY:ub nMBPARITY_MASK:uw
+#endif
+
+
+//#endif // !defined(__INITIALIZE_MBPARA__)
diff --git a/shaders/h264/mc/inter_Header.inc b/shaders/h264/mc/inter_Header.inc new file mode 100644 index 0000000..bd10c22 --- /dev/null +++ b/shaders/h264/mc/inter_Header.inc @@ -0,0 +1,371 @@ +/*
+ * Header file for all AVC INTER prediction kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__INTER_HEADER__) // Make sure this file is only included once
+#define __INTER_HEADER__
+
+// Module name: inter_header.inc
+//
+// Header file for all AVC INTER prediction kernels
+//
+
+#define INTER_KERNEL
+
+//-------------------------------------------------------------------------------------------
+// TODO: The followings will be merged with the above definitions later
+//-------------------------------------------------------------------------------------------
+
+
+//------------ Input parameters & bit masks
+
+// SW WA for weighted prediction - 2007/09/06
+//.declare guwR1 Base=r1 ElementSize=2 Type=uw
+//.declare guwW128 Base=r63.13 ElementSize=2 Type=uw
+
+#ifdef DEV_ILK
+// #define SW_W_128 // Enable SW WA for special Weight=128 case. Can be commented to disable it
+#else // Pre DEV_ILK
+#define SW_W_128 // Enable SW WA for special Weight=128 case.
+#endif // DEV_ILK
+
+#ifdef SW_W_128
+.declare gudW128 Base=r1.0 ElementSize=4 Type=ud
+#else
+#endif // SW_W_128
+
+#define gORIX r3.4 // :ub, X origin
+#define gORIY r3.5 // :ub, Y origin
+
+#define gCBP r3.9 // :ub, CBP (0, 0, Y0, Y1, Y2, Y3, Cb, Cr)
+#define nCBPY_MASK 0x3c
+#define nCBPU_MASK 0x2
+#define nCBPV_MASK 0x1
+
+#define gFIELDFLAGS r3.1 // :uw - To compute message descriptor for write
+
+#define gMBTYPE r3.1 // :ub, MB type
+#define nMBTYPE_MASK 0x1f
+#define gFIELDMBFLAG r3.1 // :ub, Field MB flag
+#define nFIELDMB_MASK 0x40
+#define gMBPARITY r3.3 // :ub, Bottom field flag
+#define nMBPARITY_MASK 0x01
+
+#define gWPREDFLAG r3.0 // :ub, Weighted pred flag
+#define nWBIDIR_MASK 0xc0
+
+#define gSUBMB_SHAPE r3.12 // :ub, Sub-MB shape
+#define gSUBMB_MODE r3.13 // :ub, Sub-MB prediction mode
+.declare guwSUBMB_SHAPE_MODE Base=r3.6 ElementSize=2 Type=uw
+
+#define gYWDENOM r3.14 // :ub, Luma log2 weight denom
+#define gCWDENOM r3.15 // :ub, Chroma log2 weight denom
+
+#define gADDR r3.24 // :ub, Register addresses of error data / MV
+
+.declare gubBIDX Base=r3.16 ElementSize=1 Type=ub
+
+#define gWGT r8 // Weights/offsets
+.declare gdWGT Base=r8 ElementSize=4 Type=d
+.declare gwWGT Base=r8 ElementSize=2 Type=w
+#define gMV r4 // MVs
+.declare gwMV Base=r4 ElementSize=2 Type=w
+.declare gdMV Base=r4 ElementSize=4 Type=d
+
+.declare gwERRORY Base=r10 ElementSize=2 Type=w // 16 GRFs
+.declare gubERRORY Base=r10 ElementSize=1 Type=ub
+.declare gwERRORC Base=r26 ElementSize=2 Type=w // 8 GRFs
+.declare gubERRORC Base=r26 ElementSize=2 Type=ub
+
+//------------ Address registers
+#define pMSGDSC a0.0 // ud: Must be the leading dword of the address register
+#define pREF a0.0
+
+#define pBIDX a0.2
+#define pWGT a0.3
+#define pERRORYC a0.2 // :ud
+#define pERRORY a0.4
+#define pERRORC a0.5
+#define pMV a0.6
+
+#define pWGT_BIDX a0.1 // :ud, WGT & BIDX
+#define pRECON_MV a0.3 // :ud, RECON & MV
+
+#define pREF0 a0.0 // :uw
+#define pREF0D a0.0 // :ud
+#define pREF1 a0.1
+#define pREF2 a0.2
+#define pREF2D a0.1 // :ud
+#define pREF3 a0.3
+#define pREF4 a0.4
+#define pREF4D a0.2 // :ud
+#define pREF5 a0.5
+#define pREF6 a0.6
+#define pREF6D a0.3 // :ud
+#define pREF7 a0.7
+
+#define pRES a0.6
+#define pRESD a0.3 // :ud
+#define pRESULT a0.7
+
+#define p0 a0.0
+#define p1 a0.1
+
+//------------ Constants for static/inline/indirect
+#define nOFFSET_BIDX 112 // = 32*3+4*4
+
+#define nOFFSET_WGT 256 // = 32*8
+#define nOFFSET_WGT_BIDX 0x01000070 // = (256<<16)+112
+#define nOFFSET_ERROR 0x03400140 // = (320+128*4)<<16+320=0x03400140
+#define nOFFSET_ERRORY 0x0140
+#define nOFFSET_ERRORC 0x0340
+#define nOFFSET_MV 128 // = 32*4
+#define nOFFSET_RECON_MV 0x04400080 // = (1088<<16)+128 // TODO: OFFSET_RECON is obsolete
+
+//------------ Constants for kernel internal variables
+#define nOFFSET_INTPY0 0x0640 // = 32*50
+#define nOFFSET_INTPY1 0x0780 // = 32*60
+#define nOFFSET_INTPC0 0x06c0 // = 32*54
+#define nOFFSET_INTPC1 0x0480 // = 32*36
+#define nOFFSET_INTP0 0x06c00640
+#define nOFFSET_INTP1 0x04800780
+
+#define nOFFSET_INTERIM 0x0480 // = 32*36
+#define nOFFSET_INTERIM2 0x04A00480 // = ((32*37)<<16)|(32*36)
+#define nOFFSET_INTERIM3 0x04A00480 // = ((32*36+32)<<16)|(32*36)
+#define nOFFSET_INTERIM4 0x04A00490 // = ((32*37)<<16)|(32*36+16)
+
+#define nOFFSET_INTERIM4x4 0x04C0 // = 32*38
+#define nOFFSET_INTERIM4x4_4 0x04E004D0 // = ((32*38+32)<<16)|(32*38+16)
+#define nOFFSET_INTERIM4x4_5 0x04D004C0 // = ((32*38+16)<<16)|(32*38)
+#define nOFFSET_INTERIM4x4_6 0x04E004C0 // = ((32*38+32)<<16)|(32*38)
+#define nOFFSET_INTERIM4x4_7 0x04D004C8 // = ((32*38+16)<<16)|(32*38+8)
+#define nOFFSET_INTERIM4x4_8 0x04E004D8 // = ((32*38+32)<<16)|(32*38+24)
+#define nOFFSET_INTERIM4x4_9 0x04F004E8 // = ((32*38+48)<<16)|(32*38+40)
+
+#define nOFFSET_RES 0x540 // = 32*42
+#define nOFFSET_REF 0x560 // = 32*43
+#define nOFFSET_REFC 0x700 // = 32*56
+
+ // Binding table index
+#define nBDIX_DESTY 0
+#define nBDIX_DESTC 1
+#define nBI_LC_DIFF 0x10 // Binding table index diff between luma and chroma
+
+#define nGRFWIB 32
+#define nGRFHWIB 16
+
+//------------ Regions
+
+.declare gudREF Base=r43 ElementSize=4 SrcRegion=<16;16,1> Type=ud
+.declare gubREF Base=r43 ElementSize=1 Type=ub
+.declare gudREFC Base=r56 ElementSize=4 SrcRegion=<16;16,1> Type=ud
+
+// 16x16 handling
+.declare gudREF21x21 Base=r58 ElementSize=4 SrcRegion=<16;16,1> Type=ud
+.declare gudREF18x10 Base=r66 ElementSize=4 SrcRegion=<16;16,1> Type=ud
+.declare gubREF18x10 Base=r66 ElementSize=1 SrcRegion=<16;16,1> Type=ub
+
+
+
+.declare gudREF16x16 Base=r38 ElementSize=4 Type=ud // 8 GRFs
+.declare gubREF16x16 Base=r38 ElementSize=1 Type=ub
+.declare gudREFC16x8 Base=r46 ElementSize=4 Type=ud // 4 GRFs
+.declare gubREFC16x8 Base=r46 ElementSize=1 Type=ub
+
+// TODO
+.declare gubAVG Base=r56 ElementSize=1 Type=ub
+.declare gubREFY_BWD Base=r64 ElementSize=1 Type=ub
+.declare gubREFC_BWD Base=r72 ElementSize=1 Type=ub
+
+
+.declare guwINTPY0 Base=r50 ElementSize=2 SrcRegion=<16;16,1> Type=uw
+.declare gudINTPY0 Base=r50 ElementSize=4 Type=ud
+.declare gubINTPY0 Base=r50 ElementSize=1 SrcRegion=<32;16,2> Type=ub
+.declare guwINTPY1 Base=r60 ElementSize=2 SrcRegion=<16;16,1> Type=uw
+.declare gudINTPY1 Base=r60 ElementSize=4 Type=ud
+.declare gubINTPY1 Base=r60 ElementSize=1 SrcRegion=<32;16,2> Type=ub
+.declare guwYPRED Base=r50 ElementSize=2 SrcRegion=<8;8,1> Type=uw
+.declare gubYPRED Base=r50 ElementSize=1 SrcRegion=<32;16,2> Type=ub
+
+.declare guwINTPC0 Base=r54 ElementSize=2 SrcRegion=<16;16,1> Type=uw
+.declare gwINTPC0 Base=r54 ElementSize=2 SrcRegion=<16;16,1> Type=w
+.declare gudINTPC0 Base=r54 ElementSize=4 Type=ud
+.declare gubINTPC0 Base=r54 ElementSize=1 SrcRegion=<32;16,2> Type=ub
+.declare guwINTPC1 Base=r36 ElementSize=2 SrcRegion=<16;16,1> Type=uw
+.declare gudINTPC1 Base=r36 ElementSize=4 Type=ud
+.declare gubINTPC1 Base=r36 ElementSize=1 SrcRegion=<32;16,2> Type=ub
+.declare guwCPRED Base=r54 ElementSize=2 SrcRegion=<16;8,2> Type=uw
+.declare gubCPRED Base=r54 ElementSize=1 SrcRegion=<32;8,4> Type=ub
+
+#define gINTERIM r36
+.declare gubINTERIM_BUF Base=r36 ElementSize=1 SrcRegion=<32;16,2> Type=ub
+#define gINTERIM4x4 r38
+.declare gubINTERIM4x4_BUF Base=r38 ElementSize=1 SrcRegion=<32;16,2> Type=ub
+.declare gwINTERIM4x4_BUF Base=r38 ElementSize=2 Type=w
+
+.declare gubINTERIM_BUF2 Base=r42 ElementSize=1 SrcRegion=<8;4,2> Type=ub
+.declare gwINTERIM_BUF2 Base=r42 ElementSize=2 SrcRegion=<16;16,1> Type=w
+.declare guwINTERIM_BUF2 Base=r42 ElementSize=2 Type=uw
+
+.declare gwINTERIM_BUF3 Base=r38 ElementSize=2 SrcRegion=<16;16,1> Type=w // 2 GRFs
+.declare gubINTERIM_BUF3 Base=r38 ElementSize=1 Type=ub
+
+.declare gwTEMP Base=r42 ElementSize=2 SrcRegion=<16;16,1> Type=w
+
+//------------ General registers
+
+#define gX r3.2 // w
+#define gY r3.3 // w
+
+#define gMSGDSC_R r3.6 // ud
+#define gMSGDSC_W r3.7 // ud
+
+#ifdef SW_W_128
+.declare gwMBTYPE Base=r8.6 ElementSize=2 Type=w // Shared with gLOOP_SUBMB
+
+// TODO
+#define gLOOP_SUBMB r8.6
+#define gLOOP_SUBMBPT r8.7
+#define gLOOP_DIR r9.6
+#define gLOOPCNT r9.7 // Loop counter for submodules
+#else
+.declare gwMBTYPE Base=r1.0 ElementSize=2 Type=w // Shared with gLOOP_SUBMB
+
+// TODO
+#define gLOOP_SUBMB r1.0
+#define gLOOP_SUBMBPT r1.1
+#define gLOOP_DIR r8.7
+#define gLOOPCNT r9.7 // Loop counter for submodules
+#endif // SW_W_128
+
+#define gW0 r34.6 // Temporary WORD
+#define gW1 r34.7 // Temporary WORD
+#define gW2 r34.8 // Temporary WORD
+#define gW3 r34.9 // Temporary WORD
+#define gD0 r34.3 // Temporary DWORD
+
+#define gW4 r34.15
+
+//
+
+#define gMVX_INT r34.0 // :w
+#define gMVY_INT r34.1 // :w
+#define gMVX_FRAC r34.2 // :w
+#define gMVY_FRAC r34.3 // :w
+#define gMVX_FRACC r34.4 // :w
+#define gMVY_FRACC r34.5 // :w
+
+#define gpINTPY r34.10
+#define gpINTPC r34.11
+#define gpINTP r34.5 // DW
+
+#define gPREDFLAG r34.12
+#define gBIDX r34.13
+#define gREFPARITY r34.14
+#define gCHRMVADJ r1.14
+#define gPARITY r1.15
+#define gCBP_MASK r1.1
+
+#define gMVSTEP r1.13
+
+#define gpADDR r1.2 // :uw (8 words)
+
+#define gSHAPETEMP r8.15 // :uw
+
+#define gCOEFA r42.0
+#define gCOEFB r42.1
+#define gCOEFC r42.2
+#define gCOEFD r42.3
+
+// Weighted prediction
+#define gPREDFLAG0 r46.0
+#define gPREDFLAG1 r46.2
+
+#define gWEIGHTFLAG r43.2
+#define gBIPRED r43.3
+#define gYADD r43.4
+#define gCADD r43.5
+#define gYSHIFT r43.6
+#define gCSHIFT r43.7
+
+#define gOFFSET r44.0
+#define gUOFFSET r44.1
+#define gVOFFSET r44.2
+
+#define gWT0 r45.0
+#define gO0 r45.1
+#define gWT1 r45.2
+#define gO1 r45.3
+#define gUW0 r45.4
+#define gUO0 r45.5
+#define gUW1 r45.6
+#define gUO1 r45.7
+#define gVW0 r45.8
+#define gVO0 r45.9
+#define gVW1 r45.10
+#define gVO1 r45.11
+
+#define gWT0_D r45.0
+#define gUW0_D r45.2
+
+//------------ Message-related Registers & constants
+#define gMSGSRC r2 // Message Source
+
+#define mMSGHDR m1
+#define mMSGHDRY m1
+#define mMSGHDRC m2
+#define mMSGHDR1 m1
+#define mMSGHDR2 m2
+#define mMSGHDR3 m3
+#define mMSGHDR4 m4
+#define mMSGHDRYW m1
+#define mMSGHDRCW m10
+
+#ifdef DEV_ILK
+ // 0000 0100(read) 0001(msg len) xxxx(resp len) 1010 (sampler cache) xxxx (field/frame) xxxx xxxx (bidx)
+#define nDWBRMSGDSC_SC 0x0208A002 // DWORD Block Read Message Descriptor through Data Port, Sampler Cache
+#define nDWBRMSGDSC_SC_TF 0x0208E602 // DWORD Block Read Message Descriptor through Data Port, Sampler Cache
+#define nDWBRMSGDSC_SC_BF 0x0208E702 // DWORD Block Read Message Descriptor through Data Port, Sampler Cache
+ // 0000 0101(write) 0001(msg len) xxxx(resp len) 0010 (render cache) xxxx (field/frame) xxxx xxxx (bidx)
+#define nDWBWMSGDSC 0x02082000 // DWORD Block Write Message Descriptor through Data Port, Render Cache
+#define nDWBWMSGDSC_TF 0x02082600 // DWORD Block Write Message Descriptor through Data Port, Render Cache
+#define nDWBWMSGDSC_BF 0x02082700 // DWORD Block Write Message Descriptor through Data Port, Render Cache
+
+#else // Pre DEV_ILK
+ // 0000 0100(read) 0001(msg len) xxxx(resp len) 1010 (sampler cache) xxxx (field/frame) xxxx xxxx (bidx)
+#define nDWBRMSGDSC_SC 0x0410A002 // DWORD Block Read Message Descriptor through Data Port, Sampler Cache
+#define nDWBRMSGDSC_SC_TF 0x0410A602 // DWORD Block Read Message Descriptor through Data Port, Sampler Cache
+#define nDWBRMSGDSC_SC_BF 0x0410A702 // DWORD Block Read Message Descriptor through Data Port, Sampler Cache
+ // 0000 0101(write) 0001(msg len) xxxx(resp len) 0010 (render cache) xxxx (field/frame) xxxx xxxx (bidx)
+#define nDWBWMSGDSC 0x05102000 // DWORD Block Write Message Descriptor through Data Port, Render Cache
+#define nDWBWMSGDSC_TF 0x05102600 // DWORD Block Write Message Descriptor through Data Port, Render Cache
+#define nDWBWMSGDSC_BF 0x05102700 // DWORD Block Write Message Descriptor through Data Port, Render Cache
+#endif // DEV_ILK
+
+#define nDWB_FIELD_MASK 0x0600
+
+// message data payload
+.declare mbMSGPAYLOADY Base=m2 ElementSize=1 SrcRegion=REGION(16,1) Type=b
+.declare mbMSGPAYLOADC Base=m11 ElementSize=1 SrcRegion=REGION(16,1) Type=b
+
+// Destination registers for write commit
+#define gREG_WRITE_COMMIT_Y r10.0
+#define gREG_WRITE_COMMIT_UV r11.0
+
+#define RETURN_REG_INTER r1.5 // Return pointer for all sub-routine calls (type DWORD)
+
+#define CALL_INTER(subFunc, skipInst) add (1) RETURN_REG_INTER<1>:ud ip:ud 1+skipInst*INST_SIZE \n\
+ jmpi (1) subFunc
+#define RETURN_INTER mov (1) ip:ud RETURN_REG_INTER<0;1,0>:ud // Return to calling module
+
+
+// End of inter_header.inc
+
+#endif // !defined(__INTER_HEADER__)
+
diff --git a/shaders/h264/mc/interpolate_C_2x2.asm b/shaders/h264/mc/interpolate_C_2x2.asm new file mode 100644 index 0000000..ffa65cf --- /dev/null +++ b/shaders/h264/mc/interpolate_C_2x2.asm @@ -0,0 +1,57 @@ +/*
+ * Interpolation kernel for chrominance 2x2 motion compensation
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Interpolate_C_2x2.asm
+//
+// Interpolation kernel for chrominance 2x2 motion compensation
+//
+// $Revision: 8 $
+// $Date: 10/09/06 4:00p $
+//
+
+
+//#if !defined(__Interpolate_C_2x2__) // Make sure this is only included once
+//#define __Interpolate_C_2x2__
+
+
+ // (8-xFrac) and (8-yFrac)
+ add (2) gW0<1>:w gMVX_FRACC<2;2,1>:w -0x08:w
+
+ // Compute the GRF address of the starting position of the reference area
+ mov (1) pREF0:w nOFFSET_REFC:w {NoDDClr}
+ mov (1) pRESULT:uw gpINTPC:uw {NoDDChk}
+
+ // gCOEFA = (8-xFrac)*(8-yFrac)
+ // gCOEFB = xFrac*(8-yFrac)
+ // gCOEFC = (8-xFrac)*yFrac
+ // gCOEFD = xFrac*yFrac
+ mul (1) gCOEFD:w gMVX_FRACC:w gMVY_FRACC:w {NoDDClr}
+ mul (1) gCOEFA:w -gW0:w -gW1:uw {NoDDClr,NoDDChk}
+ mul (1) gCOEFB:w gMVX_FRACC:w -gW1:uw {NoDDClr,NoDDChk}
+ mul (1) gCOEFC:w -gW0:w gMVY_FRACC:w {NoDDChk}
+
+ // (8-xFrac)*(8-yFrac)*A
+ // ---------------------
+ mul (8) acc0<1>:uw r[pREF0,0]<8;4,1>:ub gCOEFA:uw
+
+ // xFrac*(8-yFrac)*B
+ // -------------------
+ mac (8) acc0<1>:uw r[pREF0,2]<8;4,1>:ub gCOEFB:uw
+
+ // (8-xFrac)*yFrac*C
+ // -------------------
+ mac (8) acc0<1>:uw r[pREF0,8]<8;4,1>:ub gCOEFC:uw
+
+ // xFrac*yFrac*D
+ // -----------------
+ mac (8) gwINTERIM_BUF2(0)<1> r[pREF0,10]<8;4,1>:ub gCOEFD:uw
+ mov (4) r[pRESULT]<1>:uw gwINTERIM_BUF2(0)<4;4,1> {NoDDClr}
+ mov (4) r[pRESULT,16]<1>:uw gwINTERIM_BUF2(0,4)<4;4,1> {NoDDChk}
+
+//#endif // !defined(__Interpolate_C_2x2__)
diff --git a/shaders/h264/mc/interpolate_C_4x4.asm b/shaders/h264/mc/interpolate_C_4x4.asm new file mode 100644 index 0000000..ea23b11 --- /dev/null +++ b/shaders/h264/mc/interpolate_C_4x4.asm @@ -0,0 +1,67 @@ +/*
+ * Interpolation kernel for chrominance 4x4 motion compensation
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Interpolate_C_4x4_Func.asm
+//
+// Interpolation kernel for chrominance 4x4 motion compensation
+//
+// $Revision: 8 $
+// $Date: 10/09/06 4:00p $
+//
+
+
+//#if !defined(__Interpolate_C_4x4_Func__) // Make sure this is only included once
+//#define __Interpolate_C_4x4_Func__
+
+
+INTERLABEL(Interpolate_C_4x4_Func):
+
+
+ // (8-xFrac) and (8-yFrac)
+ add (2) gW0<1>:w gMVX_FRACC<2;2,1>:w -0x08:w
+
+ // Compute the GRF address of the starting position of the reference area
+ mov (1) pREF0:w nOFFSET_REFC:w {NoDDClr}
+ mov (1) pREF1:uw nOFFSET_REFC+16:w {NoDDChk,NoDDClr}
+ mov (1) pRESULT:uw gpINTPC:uw {NoDDChk}
+
+ // gCOEFA = (8-xFrac)*(8-yFrac)
+ // gCOEFB = xFrac*(8-yFrac)
+ // gCOEFC = (8-xFrac)*yFrac
+ // gCOEFD = xFrac*yFrac
+ mul (1) gCOEFD:w gMVX_FRACC:w gMVY_FRACC:w {NoDDClr}
+ mul (1) gCOEFA:w -gW0:w -gW1:uw {NoDDClr,NoDDChk}
+ mul (1) gCOEFB:w gMVX_FRACC:w -gW1:uw {NoDDClr,NoDDChk}
+ mul (1) gCOEFC:w -gW0:w gMVY_FRACC:w {NoDDChk}
+
+ add (2) gW0<1>:uw pREF0<2;2,1>:uw 16:w
+
+ // (8-xFrac)*(8-yFrac)*A
+ // ---------------------
+ mul (16) acc0<1>:uw r[pREF0,0]<16;8,1>:ub gCOEFA:uw
+ mul (16) acc1<1>:uw r[pREF0,nGRFWIB]<16;8,1>:ub gCOEFA:uw
+
+ // xFrac*(8-yFrac)*B
+ // -------------------
+ mac (16) acc0<1>:uw r[pREF0,2]<16;8,1>:ub gCOEFB:uw
+ mac (16) acc1<1>:uw r[pREF0,nGRFWIB+2]<16;8,1>:ub gCOEFB:uw
+
+ // (8-xFrac)*yFrac*C
+ // -------------------
+ mov (2) pREF0<1>:uw gW0<2;2,1>:uw
+ mac (16) acc0<1>:uw r[pREF0,0]<8,1>:ub gCOEFC:uw
+ mac (16) acc1<1>:uw r[pREF0,nGRFWIB]<8,1>:ub gCOEFC:uw
+
+ // xFrac*yFrac*D
+ // -----------------
+ mac (16) r[pRESULT]<1>:uw r[pREF0,2]<8,1>:ub gCOEFD:uw
+ mac (16) r[pRESULT,GRFWIB]<1>:uw r[pREF0,nGRFWIB+2]<8,1>:ub gCOEFD:uw {SecHalf}
+
+
+//#endif // !defined(__Interpolate_C_4x4_Func__)
diff --git a/shaders/h264/mc/interpolate_Y_4x4.asm b/shaders/h264/mc/interpolate_Y_4x4.asm new file mode 100644 index 0000000..dbb5733 --- /dev/null +++ b/shaders/h264/mc/interpolate_Y_4x4.asm @@ -0,0 +1,217 @@ +/*
+ * Interpolation kernel for luminance motion compensation
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Interpolate_Y_4x4.asm
+//
+// Interpolation kernel for luminance motion compensation
+//
+// $Revision: 10 $
+// $Date: 10/09/06 4:00p $
+//
+
+
+ // Compute the GRF address of the starting position of the reference area
+#if 1
+ (-f0.1) mov (1) pREF:w nOFFSET_REF+2+nGRFWIB:w
+ (f0.1) mov (1) pREF:w nOFFSET_REF+2:w
+ mov (1) pRESULT:uw gpINTPY:uw
+#else
+ mov (1) pREF:w nOFFSET_REF+2+nGRFWIB:w {NoDDClr}
+ mov (1) pRESULT:uw gpINTPY:uw {NoDDChk}
+#endif
+
+ /*
+ * | |
+ * - - 0 1 2 3 + -
+ * 4 5 6 7
+ * 8 9 A B
+ * C D E F
+ * - - + - - - + -
+ * | |
+ */
+
+ // Case 0
+ or.z.f0.1 (16) null:w gMVY_FRAC<0;1,0>:w gMVX_FRAC<0;1,0>:w
+ (f0.1) mov (4) r[pRESULT]<1>:uw r[pREF0]<4;4,1>:ub
+ (f0.1) mov (4) r[pRESULT,16]<1>:uw r[pREF0,16]<4;4,1>:ub
+ (f0.1) mov (4) r[pRESULT,32]<1>:uw r[pREF0,32]<4;4,1>:ub
+ (f0.1) mov (4) r[pRESULT,48]<1>:uw r[pREF0,48]<4;4,1>:ub
+ (f0.1) jmpi INTERLABEL(Exit_Interpolate_Y_4x4)
+
+ // Store all address registers
+ mov (8) gpADDR<1>:w a0<8;8,1>:w
+
+ mul.z.f0.0 (1) gW4:w gMVY_FRAC:w gMVX_FRAC:w
+ and.nz.f0.1 (1) null gW4:w 1:w
+
+ add (1) pREF1:uw pREF0:uw nGRFWIB/2:uw
+ add (2) pREF2<1>:uw pREF0<2;2,1>:uw nGRFWIB:uw
+ mov (4) gW0<1>:uw pREF0<4;4,1>:uw
+
+ (f0.0) jmpi INTERLABEL(Interpolate_Y_H_4x4)
+ (f0.1) jmpi INTERLABEL(Interpolate_Y_H_4x4)
+
+ //-----------------------------------------------------------------------
+ // CASE: A69BE (H/V interpolation)
+ //-----------------------------------------------------------------------
+
+ // Compute interim horizontal intepolation
+ add (1) pREF0<1>:uw pREF0<0;1,0>:uw -34:w
+ add (1) pREF1<1>:uw pREF1<0;1,0>:uw -18:w {NoDDClr}
+ mov (1) pRESD:ud nOFFSET_INTERIM4x4_5:ud {NoDDChk} // Case 69be
+
+ // Check whether this position is 'A'
+ cmp.e.f0.0 (1) null gW4:w 4:w
+
+ $for(0;<2;1) {
+ add (16) acc0<1>:w r[pREF0,nGRFWIB*2*%1]<16;4,1>:ub r[pREF0,nGRFWIB*2*%1+5]<16;4,1>:ub {Compr}
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB*2*%1+1]<16;4,1>:ub -5:w {Compr}
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB*2*%1+2]<16;4,1>:ub 20:w {Compr}
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB*2*%1+3]<16;4,1>:ub 20:w {Compr}
+ mac (16) r[pRES,nGRFWIB*%1]<1>:w r[pREF0,nGRFWIB*2*%1+4]<16;4,1>:ub -5:w {Compr}
+ }
+ // last line
+ add (4) acc0<1>:w r[pREF0,nGRFWIB*2*2]<4;4,1>:ub r[pREF0,nGRFWIB*2*2+5]<4;4,1>:ub
+ mac (4) acc0<1>:w r[pREF0,nGRFWIB*2*2+1]<4;4,1>:ub -5:w
+ mac (4) acc0<1>:w r[pREF0,nGRFWIB*2*2+2]<4;4,1>:ub 20:w
+ mac (4) acc0<1>:w r[pREF0,nGRFWIB*2*2+3]<4;4,1>:ub 20:w
+ mac (4) r[pRES,nGRFWIB*2]<1>:w r[pREF0,nGRFWIB*2*2+4]<4;4,1>:ub -5:w
+
+ // Compute interim/output vertical interpolation
+ mov (1) pREF6D:ud nOFFSET_INTERIM4x4_4:ud {NoDDClr}
+ mov (1) pREF0D:ud nOFFSET_INTERIM4x4_7:ud {NoDDChk,NoDDClr}
+ mov (1) pREF2D:ud nOFFSET_INTERIM4x4_8:ud {NoDDChk,NoDDClr}
+ mov (1) pREF4D:ud nOFFSET_INTERIM4x4_9:ud {NoDDChk}
+
+ add (16) acc0<1>:w gwINTERIM4x4_BUF(0)<16;16,1> 512:w
+ mac (16) acc0<1>:w gwINTERIM4x4_BUF(1)<16;16,1> -5:w
+ mac (16) acc0<1>:w r[pREF6,0]<8,1>:w 20:w
+
+ (f0.0) mov (1) pRES:uw nOFFSET_RES:uw // Case a
+ (-f0.0) mov (1) pRES:uw nOFFSET_INTERIM4x4:uw // Case 69be
+
+ mac (16) acc0<1>:w r[pREF0,0]<4,1>:w -5:w
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB]<4,1>:w 1:w
+ mac (16) acc0<1>:w r[pREF2,0]<4,1>:w 20:w
+ asr.sat (16) r[pRES]<2>:ub acc0<16;16,1>:w 10:w
+
+ (f0.0) jmpi INTERLABEL(Return_Interpolate_Y_4x4)
+
+INTERLABEL(Interpolate_Y_H_4x4):
+
+ cmp.e.f0.0 (1) null gMVX_FRAC:w 0:w
+ cmp.e.f0.1 (1) null gMVY_FRAC:w 2:w
+ (f0.0) jmpi INTERLABEL(Interpolate_Y_V_4x4)
+ (f0.1) jmpi INTERLABEL(Interpolate_Y_V_4x4)
+
+ //-----------------------------------------------------------------------
+ // CASE: 123567DEF (H interpolation)
+ //-----------------------------------------------------------------------
+
+ add (4) pREF0<1>:uw gW0<4;4,1>:uw -2:w
+ cmp.g.f0.0 (4) null:w gMVY_FRAC<0;1,0>:w 2:w
+ cmp.e.f0.1 (1) null gMVX_FRAC:w 2:w
+ (f0.0) add (4) pREF0<1>:uw pREF0<4;4,1>:uw nGRFWIB/2:uw
+
+ cmp.e.f0.0 (1) null:w gMVY_FRAC<0;1,0>:w 0:w
+
+ (f0.1) mov (1) pRESULT:uw nOFFSET_RES:uw // Case 26E
+ (-f0.1) mov (1) pRESULT:uw nOFFSET_INTERIM4x4:uw // Case 1357DF
+
+ // Compute interim/output horizontal interpolation
+ add (16) acc0<1>:w r[pREF0]<4,1>:ub 16:w
+ mac (16) acc0<1>:w r[pREF0,1]<4,1>:ub -5:w
+ mac (16) acc0<1>:w r[pREF0,2]<4,1>:ub 20:w
+ mac (16) acc0<1>:w r[pREF0,3]<4,1>:ub 20:w
+ mac (16) acc0<1>:w r[pREF0,4]<4,1>:ub -5:w
+ mac (16) acc0<1>:w r[pREF0,5]<4,1>:ub 1:w
+ asr.sat (16) r[pRESULT]<2>:ub acc0<16;16,1>:w 5:w
+
+ (-f0.1) jmpi INTERLABEL(Interpolate_Y_V_4x4)
+ (-f0.0) jmpi INTERLABEL(Average_4x4)
+
+ jmpi INTERLABEL(Return_Interpolate_Y_4x4)
+
+INTERLABEL(Interpolate_Y_V_4x4):
+
+ cmp.e.f0.0 (1) null gMVY_FRAC:w 0:w
+ (f0.0) jmpi INTERLABEL(Interpolate_Y_I_4x4)
+
+ //-----------------------------------------------------------------------
+ // CASE: 48C59D7BF (V interpolation)
+ //-----------------------------------------------------------------------
+
+ cmp.g.f0.1 (8) null:w gMVX_FRAC<0;1,0>:w 2:w
+
+ mov (4) pREF0<1>:uw gW0<4;4,1>:uw {NoDDClr}
+ add (4) pREF4<1>:w gW0<4;4,1>:uw 16:w {NoDDChk}
+
+ (f0.1) add (8) pREF0<1>:uw pREF0<4;4,1>:uw 1:uw
+
+ cmp.e.f0.0 (1) null:w gMVX_FRAC<0;1,0>:w 0:w
+ cmp.e.f0.1 (1) null gMVY_FRAC:w 2:w
+
+ // Compute interim/output vertical interpolation
+ add (16) acc0<1>:w r[pREF0,-nGRFWIB]<4,1>:ub 16:w
+ mac (16) acc0<1>:w r[pREF0]<4,1>:ub 20:w
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB]<4,1>:ub -5:w
+ mac (16) acc0<1>:w r[pREF4,-nGRFWIB]<4,1>:ub -5:w
+ mac (16) acc0<1>:w r[pREF4]<4,1>:ub 20:w
+ mac (16) acc0<1>:w r[pREF4,nGRFWIB]<4,1>:ub 1:w
+
+ mov (1) pRESULT:uw nOFFSET_RES:uw
+ (-f0.0) jmpi INTERLABEL(VFILTER_4x4)
+ (-f0.1) mov (1) pRESULT:uw nOFFSET_INTERIM4x4:uw
+
+ INTERLABEL(VFILTER_4x4):
+
+ asr.sat (16) r[pRESULT]<2>:ub acc0<16;16,1>:w 5:w
+
+ (-f0.0) jmpi INTERLABEL(Average_4x4)
+ (f0.1) jmpi INTERLABEL(Return_Interpolate_Y_4x4 )
+
+INTERLABEL(Interpolate_Y_I_4x4):
+
+ //-----------------------------------------------------------------------
+ // CASE: 134C (Integer position)
+ //-----------------------------------------------------------------------
+
+ mov (4) pREF0<1>:uw gW0<4;4,1>:uw
+
+ cmp.e.f0.0 (4) null:w gMVX_FRAC<0;1,0>:w 3:w
+ cmp.e.f0.1 (4) null:w gMVY_FRAC<0;1,0>:w 3:w
+ (f0.0) add (4) pREF0<1>:uw pREF0<4;4,1>:uw 1:uw
+ (f0.1) add (4) pREF0<1>:uw pREF0<4;4,1>:uw nGRFWIB/2:uw
+
+ mov (16) guwINTERIM_BUF2(0)<1> r[pREF0]<4,1>:ub
+
+INTERLABEL(Average_4x4):
+
+ //-----------------------------------------------------------------------
+ // CASE: 13456789BCDEF (Average)
+ //-----------------------------------------------------------------------
+
+ // Average two interim results
+ avg.sat (16) gubINTERIM_BUF2(0)<2> gubINTERIM_BUF2(0)<32;16,2> gINTERIM4x4<32;16,2>:ub
+
+INTERLABEL(Return_Interpolate_Y_4x4):
+ // Move result
+ mov (1) pRES:uw gpINTPY:uw
+ mov (4) r[pRES,0]<2>:ub gubINTERIM_BUF2(0,0)
+ mov (4) r[pRES,16]<2>:ub gubINTERIM_BUF2(0,8)
+ mov (4) r[pRES,32]<2>:ub gubINTERIM_BUF2(0,16)
+ mov (4) r[pRES,48]<2>:ub gubINTERIM_BUF2(0,24)
+
+ // Restore all address registers
+ mov (8) a0<1>:w gpADDR<8;8,1>:w
+
+INTERLABEL(Exit_Interpolate_Y_4x4):
+
+
+// end of file
diff --git a/shaders/h264/mc/interpolate_Y_8x8.asm b/shaders/h264/mc/interpolate_Y_8x8.asm new file mode 100644 index 0000000..e7e3ff9 --- /dev/null +++ b/shaders/h264/mc/interpolate_Y_8x8.asm @@ -0,0 +1,262 @@ +/*
+ * Interpolation kernel for luminance motion compensation
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Interpolate_Y_8x8.asm
+//
+// Interpolation kernel for luminance motion compensation
+//
+// $Revision: 13 $
+// $Date: 10/09/06 4:00p $
+//
+
+
+//---------------------------------------------------------------
+// In: pMV => Source address of MV
+// In: gMVX_FRAC<2;2,1>:w => MV fractional components
+// In: f0.1 (1) => If 1, vertical MV is integer
+// In: gpINTPY:uw => Destination address for interpolated result
+// In: Reference area staring from R43
+// If horizontal/vertical MVs are all integer, 8x8 pixels are on R43~R44 (2 GRFs)
+// If only horz MV is integer, 8x13 pixels are on R43~R46 (4 GRFs)
+// If only vert MV is integer, 13x8 pixels are on R43~R46 (4 GRFs)
+// If no MVs are integer, 13x13 pixels are on R43~R49 (7 GRFs)
+//---------------------------------------------------------------
+
+
+INTERLABEL(Interpolate_Y_8x8_Func):
+
+
+
+ // Check whether MVX is integer MV
+ and.z.f0.0 (1) null:w r[pMV,0]<0;1,0>:w 0x3:w
+ (-f0.0) jmpi (1) INTERLABEL(Interpolate_Y_8x8_Func2)
+
+ // TODO: remove this back-to-back read - huge latency..
+ mov (8) gubREF(6,2)<1> gubREF(3,0)<8;8,1>
+ mov (8) gubREF(5,18)<1> gubREF(2,24)<8;8,1> {NoDDClr}
+ mov (8) gubREF(5,2)<1> gubREF(2,16)<8;8,1> {NoDDChk}
+ mov (8) gubREF(4,18)<1> gubREF(2,8)<8;8,1> {NoDDClr}
+ mov (8) gubREF(4,2)<1> gubREF(2,0)<8;8,1> {NoDDChk}
+ mov (8) gubREF(3,18)<1> gubREF(1,24)<8;8,1> {NoDDClr}
+ mov (8) gubREF(3,2)<1> gubREF(1,16)<8;8,1> {NoDDChk}
+ mov (8) gubREF(2,18)<1> gubREF(1,8)<8;8,1> {NoDDClr}
+ mov (8) gubREF(2,2)<1> gubREF(1,0)<8;8,1> {NoDDChk}
+ mov (8) gubREF(1,18)<1> gubREF(0,24)<8;8,1> {NoDDClr}
+ mov (8) gubREF(1,2)<1> gubREF(0,16)<8;8,1> {NoDDChk}
+ mov (8) gubREF(0,18)<1> gubREF(0,8)<8;8,1>
+ mov (8) gubREF(0,2)<1> gubREF(0,0)<8;8,1>
+
+INTERLABEL(Interpolate_Y_8x8_Func2):
+
+ // Compute the GRF address of the starting position of the reference area
+ (-f0.1) mov (1) pREF:w nOFFSET_REF+2+nGRFWIB:w
+ (f0.1) mov (1) pREF:w nOFFSET_REF+2:w
+ mov (1) pRESULT:uw gpINTPY:uw
+
+ /*
+ * | |
+ * - - 0 1 2 3 + -
+ * 4 5 6 7
+ * 8 9 A B
+ * C D E F
+ * - - + - - - + -
+ * | |
+ */
+
+ // Case 0
+ or.z.f0.1 (16) null:w gMVY_FRAC<0;1,0>:w gMVX_FRAC<0;1,0>:w
+ (f0.1) mov (16) r[pRESULT]<1>:uw r[pREF]<16;8,1>:ub
+ (f0.1) mov (16) r[pRESULT,nGRFWIB]<1>:uw r[pREF,nGRFWIB]<16;8,1>:ub
+ (f0.1) mov (16) r[pRESULT,nGRFWIB*2]<1>:uw r[pREF,nGRFWIB*2]<16;8,1>:ub
+ (f0.1) mov (16) r[pRESULT,nGRFWIB*3]<1>:uw r[pREF,nGRFWIB*3]<16;8,1>:ub
+ (f0.1) jmpi INTERLABEL(Exit_Interpolate_Y_8x8)
+
+ // Store all address registers
+ mov (8) gpADDR<1>:w a0<8;8,1>:w
+
+ mul.z.f0.0 (1) gW4:w gMVY_FRAC:w gMVX_FRAC:w
+ add (1) pREF1:uw pREF0:uw nGRFWIB/2:uw
+ and.nz.f0.1 (1) null gW4:w 1:w
+ add (2) pREF2<1>:uw pREF0<2;2,1>:uw nGRFWIB:uw
+ mov (4) gW0<1>:uw pREF0<4;4,1>:uw
+
+ (f0.0) jmpi INTERLABEL(Interpolate_Y_H_8x8)
+ (f0.1) jmpi INTERLABEL(Interpolate_Y_H_8x8)
+
+ //-----------------------------------------------------------------------
+ // CASE: A69BE (H/V interpolation)
+ //-----------------------------------------------------------------------
+
+ // Compute interim horizontal intepolation of 12 lines (not 9 lines)
+// add (1) pREF0<1>:ud pREF0<0;1,0>:ud 0xffeeffde:ud // (-18<<16)|(-34)
+ add (1) pREF0<1>:uw pREF0<0;1,0>:uw -34:w
+ add (1) pREF1<1>:uw pREF1<0;1,0>:uw -18:w {NoDDClr}
+ mov (1) pRESD:ud nOFFSET_INTERIM3:ud {NoDDChk}
+
+ // Check whether this position is 'A'
+ cmp.e.f0.0 (1) null gW4:w 4:w
+
+ $for(0;<6;2) {
+ add (32) acc0<1>:w r[pREF,nGRFWIB*%1]<16;8,1>:ub r[pREF0,nGRFWIB*%1+5]<16;8,1>:ub {Compr}
+ mac (32) acc0<1>:w r[pREF,nGRFWIB*%1+1]<16;8,1>:ub -5:w {Compr}
+ mac (32) acc0<1>:w r[pREF,nGRFWIB*%1+2]<16;8,1>:ub 20:w {Compr}
+ mac (32) acc0<1>:w r[pREF,nGRFWIB*%1+3]<16;8,1>:ub 20:w {Compr}
+ mac (32) r[pRES,nGRFWIB*%1]<1>:w r[pREF,nGRFWIB*%1+4]<16;8,1>:ub -5:w {Compr}
+ }
+ // last line
+ add (8) acc0<1>:w r[pREF,nGRFWIB*6]<8;8,1>:ub r[pREF,nGRFWIB*6+5]<8;8,1>:ub
+ mac (8) acc0<1>:w r[pREF,nGRFWIB*6+1]<8;8,1>:ub -5:w
+ mac (8) acc0<1>:w r[pREF,nGRFWIB*6+2]<8;8,1>:ub 20:w
+ mac (8) acc0<1>:w r[pREF,nGRFWIB*6+3]<8;8,1>:ub 20:w
+ mac (8) r[pRES,nGRFWIB*6]<1>:w r[pREF,nGRFWIB*6+4]<8;8,1>:ub -5:w
+
+ // Compute interim/output vertical interpolation
+ mov (1) pREF0:ud nOFFSET_INTERIM2:ud {NoDDClr} // set pREF0 and pREF1 at the same time
+ mov (1) pREF2D:ud nOFFSET_INTERIM4:ud {NoDDChk,NoDDClr} // set pREF2 and pREF3 at the same time
+ (f0.0) sel (1) pRES:uw gpINTPY:uw nOFFSET_INTERIM:uw {NoDDChk} // Case A vs. 69BE
+
+ $for(0;<4;2) {
+ add (32) acc0<1>:w r[pREF0,nGRFWIB*%1]<16;16,1>:w 512:w {Compr}
+ mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1]<8,1>:w -5:w
+ mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:w -5:w
+ mac (32) acc0<1>:w r[pREF0,nGRFWIB*%1+nGRFWIB]<16;16,1>:w 20:w {Compr}
+ mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:w 20:w
+ mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB+nGRFWIB]<8,1>:w 20:w
+ mac (32) acc0<1>:w r[pREF0,(2+%1)*nGRFWIB]<16;16,1>:w -5:w {Compr}
+ mac (16) acc0<1>:w r[pREF2,(2+%1)*nGRFWIB]<8,1>:w 1:w
+ mac (16) acc1<1>:w r[pREF2,(2+%1)*nGRFWIB+nGRFWIB]<8,1>:w 1:w
+ asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub acc0<16;16,1>:w 10:w
+ asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub acc1<16;16,1>:w 10:w {SecHalf}
+ }
+
+ (f0.0) jmpi INTERLABEL(Return_Interpolate_Y_8x8)
+
+INTERLABEL(Interpolate_Y_H_8x8):
+
+ cmp.e.f0.0 (1) null gMVX_FRAC:w 0:w
+ cmp.e.f0.1 (1) null gMVY_FRAC:w 2:w
+ (f0.0) jmpi INTERLABEL(Interpolate_Y_V_8x8)
+ (f0.1) jmpi INTERLABEL(Interpolate_Y_V_8x8)
+
+ //-----------------------------------------------------------------------
+ // CASE: 123567DEF (H interpolation)
+ //-----------------------------------------------------------------------
+
+ add (4) pREF0<1>:uw gW0<4;4,1>:uw -2:w
+ cmp.g.f0.0 (4) null:w gMVY_FRAC<0;1,0>:w 2:w
+ cmp.e.f0.1 (1) null gMVX_FRAC:w 2:w
+ (f0.0) add (4) pREF0<1>:uw pREF0<4;4,1>:uw nGRFWIB/2:uw
+
+ cmp.e.f0.0 (1) null:w gMVY_FRAC<0;1,0>:w 0:w
+
+ (f0.1) sel (1) pRES:uw gpINTPY:uw nOFFSET_INTERIM:uw // Case 26E vs. 1357DF
+
+ // Compute interim/output horizontal interpolation
+ $for(0;<4;2) {
+ add (16) acc0<1>:w r[pREF0,nGRFWIB*%1]<8,1>:ub 16:w
+ add (16) acc1<1>:w r[pREF0,nGRFWIB*%1+nGRFWIB]<8,1>:ub 16:w
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+1]<8,1>:ub -5:w
+ mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+1+nGRFWIB]<8,1>:ub -5:w
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+2]<8,1>:ub 20:w
+ mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+2+nGRFWIB]<8,1>:ub 20:w
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+3]<8,1>:ub 20:w
+ mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+3+nGRFWIB]<8,1>:ub 20:w
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+4]<8,1>:ub -5:w
+ mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+4+nGRFWIB]<8,1>:ub -5:w
+ mac (16) acc0<1>:w r[pREF0,nGRFWIB*%1+5]<8,1>:ub 1:w
+ mac (16) acc1<1>:w r[pREF0,nGRFWIB*%1+5+nGRFWIB]<8,1>:ub 1:w
+ asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub acc0<16;16,1>:w 5:w
+ asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub acc1<16;16,1>:w 5:w {SecHalf}
+ }
+
+ (-f0.1) jmpi INTERLABEL(Interpolate_Y_V_8x8)
+ (-f0.0) jmpi INTERLABEL(Average_8x8)
+
+ jmpi INTERLABEL(Return_Interpolate_Y_8x8)
+
+INTERLABEL(Interpolate_Y_V_8x8):
+
+ cmp.e.f0.0 (1) null gMVY_FRAC:w 0:w
+ (f0.0) jmpi INTERLABEL(Interpolate_Y_I_8x8)
+
+ //-----------------------------------------------------------------------
+ // CASE: 48C59D7BF (V interpolation)
+ //-----------------------------------------------------------------------
+
+ mov (2) pREF0<1>:uw gW0<4;2,2>:uw {NoDDClr}
+ mov (2) pREF2<1>:uw gW1<2;2,1>:uw {NoDDChk,NoDDClr}
+ mov (1) pRES:uw gpINTPY:uw {NoDDChk}
+
+ cmp.g.f0.1 (4) null:w gMVX_FRAC<0;1,0>:w 2:w
+ cmp.e.f0.0 (1) null:w gMVX_FRAC<0;1,0>:w 0:w
+ (f0.1) add (4) pREF0<1>:uw pREF0<4;4,1>:uw 1:uw
+
+ cmp.e.f0.1 (1) null gMVY_FRAC:w 2:w
+ (-f0.0) jmpi INTERLABEL(VFILTER_8x8)
+ (-f0.1) mov (1) pRES:uw nOFFSET_INTERIM:uw
+
+ INTERLABEL(VFILTER_8x8):
+
+ // Compute interim/output vertical interpolation
+ $for(0;<4;2) {
+ add (32) acc0<1>:w r[pREF0,nGRFWIB*%1-nGRFWIB]<16;8,1>:ub 16:w {Compr}
+ mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1-nGRFWIB]<8,1>:ub -5:w
+ mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1]<8,1>:ub -5:w
+ mac (32) acc0<1>:w r[pREF0,nGRFWIB*%1]<16;8,1>:ub 20:w {Compr}
+ mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1]<8,1>:ub 20:w
+ mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:ub 20:w
+ mac (32) acc0<1>:w r[pREF0,nGRFWIB*%1+nGRFWIB]<16;8,1>:ub -5:w {Compr}
+ mac (16) acc0<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB]<8,1>:ub 1:w
+ mac (16) acc1<1>:w r[pREF2,nGRFWIB*%1+nGRFWIB+nGRFWIB]<8,1>:ub 1:w
+ asr.sat (16) r[pRES,nGRFWIB*%1]<2>:ub acc0<16;16,1>:w 5:w
+ asr.sat (16) r[pRES,nGRFWIB*%1+nGRFWIB]<2>:ub acc1<16;16,1>:w 5:w {SecHalf}
+ }
+
+ (-f0.0) jmpi INTERLABEL(Average_8x8)
+ (f0.1) jmpi INTERLABEL(Return_Interpolate_Y_8x8)
+
+INTERLABEL(Interpolate_Y_I_8x8):
+
+ //-----------------------------------------------------------------------
+ // CASE: 134C (Integer position)
+ //-----------------------------------------------------------------------
+
+ mov (2) pREF0<1>:uw gW0<2;2,1>:uw {NoDDClr}
+
+ mov (1) pRES:uw gpINTPY:uw {NoDDChk}
+
+ cmp.e.f0.0 (2) null:w gMVX_FRAC<0;1,0>:w 3:w
+ cmp.e.f0.1 (2) null:w gMVY_FRAC<0;1,0>:w 3:w
+ (f0.0) add (2) pREF0<1>:uw pREF0<2;2,1>:uw 1:uw
+ (f0.1) add (2) pREF0<1>:uw pREF0<2;2,1>:uw nGRFWIB/2:uw
+
+ mov (16) r[pRES]<1>:uw r[pREF0]<8,1>:ub
+ mov (16) r[pRES,nGRFWIB]<1>:uw r[pREF0,nGRFWIB]<8,1>:ub
+ mov (16) r[pRES,nGRFWIB*2]<1>:uw r[pREF0,nGRFWIB*2]<8,1>:ub
+ mov (16) r[pRES,nGRFWIB*3]<1>:uw r[pREF0,nGRFWIB*3]<8,1>:ub
+
+INTERLABEL(Average_8x8):
+
+ //-----------------------------------------------------------------------
+ // CASE: 13456789BCDEF (Average)
+ //-----------------------------------------------------------------------
+
+ // Average two interim results
+ avg.sat (16) r[pRES,0]<2>:ub r[pRES,0]<32;16,2>:ub gubINTERIM_BUF(0)
+ avg.sat (16) r[pRES,nGRFWIB]<2>:ub r[pRES,nGRFWIB]<32;16,2>:ub gubINTERIM_BUF(1)
+ avg.sat (16) r[pRES,nGRFWIB*2]<2>:ub r[pRES,nGRFWIB*2]<32;16,2>:ub gubINTERIM_BUF(2)
+ avg.sat (16) r[pRES,nGRFWIB*3]<2>:ub r[pRES,nGRFWIB*3]<32;16,2>:ub gubINTERIM_BUF(3)
+
+INTERLABEL(Return_Interpolate_Y_8x8):
+ // Restore all address registers
+ mov (8) a0<1>:w gpADDR<8;8,1>:w
+
+INTERLABEL(Exit_Interpolate_Y_8x8):
+
+// end of file
diff --git a/shaders/h264/mc/intra_Header.inc b/shaders/h264/mc/intra_Header.inc new file mode 100644 index 0000000..501c7a8 --- /dev/null +++ b/shaders/h264/mc/intra_Header.inc @@ -0,0 +1,276 @@ +/*
+ * Header file for all AVC intra prediction kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__INTRA_HEADER__) // Make sure this file is only included once
+#define __INTRA_HEADER__
+
+// Module name: intra_header.inc
+//
+// Header file for all AVC intra prediction kernels
+//
+// This header file defines everything that's specific to intra macroblock kernels
+
+
+// ----------- Various data buffers and pointers ------------
+//
+// I_PCM data buffer
+//
+#define I_PCM_BUF_Y 4
+#define I_PCM_BUF_UV 12
+
+#define REG_I_PCM_BUF_Y r4
+#define REG_I_PCM_BUF_UV r12
+
+.declare I_PCM_Y Base=REG_I_PCM_BUF_Y ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8-bit I_PCM Y data
+.declare I_PCM_UV Base=REG_I_PCM_BUF_UV ElementSize=1 SrcRegion=REGION(16,1) Type=ub // 8-bit I_PCM U/V data
+
+// Intra macroblock error data blocks
+//
+#define ERRBUF 4 // Starting GRF index for error data
+#define REG_ERRBUF r4
+.declare MBBLOCKW Base=REG_ERRBUF ElementSize=2 SrcRegion=REGION(16,1) Type=w // For 16-bit inter MB
+.declare MBBLOCKD Base=REG_ERRBUF ElementSize=2 SrcRegion=REGION(16,1) Type=uw // For use in "send" command
+
+#define PERROR a0.2 // Pointer to macroblock error data
+#define PERROR1 a0.3 // Pointer to macroblock error data used by instruction compression
+#define PERROR_UD a0.1 // Pointer to macroblock error data in DWORD unit
+
+// Intra macroblock reference data
+//
+#define REG_INTRA_REF_TOP r49 // Must be an odd numbered GRF register
+.declare INTRA_REF_TOP0 Base=REG_INTRA_REF_TOP ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare INTRA_REF_TOP Base=REG_INTRA_REF_TOP.4 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+ // Actual top row reference data start at offset 4 in BYTE
+.declare INTRA_REF_TOP_W Base=REG_INTRA_REF_TOP.2 ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+ // Actual top row reference data start at offset 2 in WORD
+.declare INTRA_REF_TOP_D Base=REG_INTRA_REF_TOP ElementSize=4 DstRegion=<1> Type=ud // Only used in "send" instruction
+
+#define INTRA_REF_LEFT_ID 50
+#define REG_INTRA_REF_LEFT r50
+.declare INTRA_REF_LEFT0 Base=REG_INTRA_REF_LEFT ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+.declare INTRA_REF_LEFT Base=REG_INTRA_REF_LEFT.3 ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+ // Actual left column reference data are located at offset 3 in BYTE
+.declare INTRA_REF_LEFT_UV Base=REG_INTRA_REF_LEFT.2 ElementSize=1 SrcRegion=REGION(8,4) Type=ub
+ // Actual left column U/V reference data are located at offset 2 in BYTE
+.declare INTRA_REF_LEFT_W Base=REG_INTRA_REF_LEFT.1 ElementSize=2 SrcRegion=REGION(8,2) Type=uw
+ // Actual left column reference data are located at offset 1 in WORD
+.declare INTRA_REF_LEFT_D Base=REG_INTRA_REF_LEFT ElementSize=4 DstRegion=<1> Type=ud // Only used in "send" instruction
+
+#define PREF_LEFT a0.4 // Pointer to left reference data
+#define PREF_LEFT_UD a0.2 // Pointer in DWORD to left reference data
+
+#define INTRA_TEMP_0 52
+#define INTRA_TEMP_1 53
+#define INTRA_TEMP_2 54
+#define INTRA_TEMP_3 55
+#define INTRA_TEMP_4 56
+#define INTRA_TEMP_5 57
+#define INTRA_TEMP_6 58
+
+#define REG_INTRA_TEMP_0 r52
+#define REG_INTRA_TEMP_1 r53
+#define REG_INTRA_TEMP_2 r54
+#define REG_INTRA_TEMP_3 r55
+#define REG_INTRA_TEMP_4 r56
+#define REG_INTRA_TEMP_5 r57
+#define REG_INTRA_TEMP_6 r58
+#define REG_INTRA_TEMP_7 r59
+#define REG_INTRA_TEMP_8 r60
+
+// Destination registers for write commit
+#define REG_WRITE_COMMIT_Y r60.0
+#define REG_WRITE_COMMIT_UV r61.0
+
+// ----------- Various data buffers and pointers ------------
+// R32 - R47 for predicted picture buffer (for both Y and U/V blocks)
+//
+#define PREDBUF 32 // Starting GRF index for predicted buffer
+#define REG_PREDBUF r32
+
+.declare PRED_Y Base=REG_PREDBUF ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Predicted Y picture
+.declare PRED_YW Base=REG_PREDBUF ElementSize=2 SrcRegion=REGION(16,1) Type=uw // Predicted Y picture stored in WORD
+.declare PRED_Y_FM Base=REG_PREDBUF ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Predicted Y picture frame
+.declare PRED_Y_TF Base=REG_PREDBUF ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Predicted Y picture Top field
+
+.declare PRED_UV Base=REG_PREDBUF ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Predicted U/V picture
+.declare PRED_UVW Base=REG_PREDBUF ElementSize=2 SrcRegion=REGION(16,1) Type=uw // Predicted U/V picture stored in WORD
+.declare PRED_UV_FM Base=REG_PREDBUF ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Predicted U/V picture frame
+.declare PRED_UV_TF Base=REG_PREDBUF ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Predicted U/V picture top field
+.declare PRED_UV_BF Base=REG_PREDBUF.16 ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Predicted U/V picture bottom field
+
+// The same region will also be used as finally decoded Y blocks shared with U/V blocks
+//
+#define DECBUF 32
+#define REG_DECBUF r32
+.declare DEC_Y Base=REG_DECBUF ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Decoded Y picture
+.declare DEC_UV Base=REG_DECBUF ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Decoded U/V P-/B-picture
+.declare DEC_UD Base=REG_DECBUF ElementSize=4 SrcRegion=REGION(8,1) Type=ud // Decoded buffer in UD type
+
+// Reference buffer for intra_NxN prediction
+//
+#define PRED_MODE REG_INTRA_TEMP_0
+.declare REF_TOP0 Base=REG_INTRA_TEMP_5 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare REF_TOP Base=REG_INTRA_TEMP_5.4 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+ // Actual top reference data start from offset 3,i.e. p[-1,-1]
+.declare REF_TOP_W Base=REG_INTRA_TEMP_5 ElementSize=2 SrcRegion=REGION(16,1) Type=uw
+.declare REF_TOP_D Base=REG_INTRA_TEMP_5 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+.declare REF_LEFT Base=REG_INTRA_TEMP_6 ElementSize=1 SrcRegion=REGION(16,1) Type=ub
+.declare REF_LEFT_D Base=REG_INTRA_TEMP_6 ElementSize=4 SrcRegion=REGION(8,1) Type=ud
+
+// For intra prediction plane mode
+//
+.declare H1 Base=REG_INTRA_TEMP_0 ElementSize=2 SrcRegion=REGION(8,1) Type=w // Make sure it's an even GRF
+.declare H2 Base=REG_INTRA_TEMP_0.8 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+.declare V1 Base=REG_INTRA_TEMP_1 ElementSize=2 SrcRegion=REGION(8,1) Type=w // Make sure it's the following odd GRF
+.declare V2 Base=REG_INTRA_TEMP_1.8 ElementSize=2 SrcRegion=REGION(8,1) Type=w
+
+.declare CP Base=REG_INTRA_TEMP_2 ElementSize=2 SrcRegion=REGION(16,1) Type=w
+
+#define PINTRAPRED_Y a0.7 // Used as luma intra prediction mode pointer
+#define PINTRAPRED_UV a0.7 // Used as chroma intra prediction mode pointer
+#define PINTRA4X4_Y a0.4 // Used as luma intra_4x4 prediction mode pointer
+
+#define PBWDCOPY_4 a0.4 // a0.4 - a0.7 used in intra_4x4 prediction for moving data backward
+#define PBWDCOPY_8 a0.6 // a0.6 - a0.7 used in intra_8x8 prediction for moving data backward
+
+// For Intra_4x4 prediction mode
+//
+.declare INTRA_4X4_MODE Base=REG_INTRA_TEMP_1 ElementSize=4 SrcRegion=REGION(1,0) DstRegion=<1> Type=d // Actually only need 1 DWORD
+
+// ----------- Intra CURBE constants ------------
+//
+#define REG_CURBE1 r1
+#define REG_CURBE2 r2
+#define INTRA_4X4_OFFSET 1*GRFWIB // 9 Bytes
+#define INTRA_8X8_OFFSET 1*GRFWIB+12 // 9 Bytes starting sub-register r1.3:ud
+#define INTRA_16X16_OFFSET 1*GRFWIB+24 // 4 Bytes starting sub-register r1.6:ud
+#define INTRA_CHROMA_OFFSET 1*GRFWIB+28 // 4 Bytes starting sub-register r1.7:ud
+
+#define TOP_REF_OFFSET REG_CURBE1.10 // r1.5:w
+
+// Constants used in plane intra prediction mode
+#define XY_3 REG_CURBE2.4 // Stored BYTE constants x-3 for x=0...7, i.e. -3,-2,...3,4 for U/V, need duplicate to every other byte
+#define XY_3_1 REG_CURBE2.5 // Stored BYTE constants x-3 for x=0...7, i.e. -3,-2,...3,4 for 2nd instruction in {Comp}
+#define XY_7 REG_CURBE2.0 // Stored BYTE constants x-7 for x=0...15, i.e. -7,-6,...7,8 for Y
+#define XY_7_1 REG_CURBE2.1 // Stored BYTE constants x-7 for x=0...15, i.e. -7,-6,...7,8 for 2nd instruction in {Comp}
+
+#define INV_SHIFT REG_CURBE2.16
+
+#define INV_TRANS4 REG_CURBE2.20 // For reverse data transfer for intra_4x4 (0x00020406)
+#define INV_TRANS48 REG_CURBE2.22 // For reverse data transfer for intra_4x4 (0x0002)
+#define INV_TRANS8 REG_CURBE1.22 // For reverse data transfer for intra_8x8 (0x0001)
+
+#define INTRA_MODE REG_CURBE2.24 // Offset to intra_Pred_4x4_Y from each sub-block
+
+// ----------- In-line parameters ------------
+//
+#define REG_INLINE r3
+
+#define INLINE_DW0 REG_INLINE.0<0;1,0>:ud
+#define INLINE_DW1 REG_INLINE.1<0;1,0>:ud
+#define INLINE_DW2 REG_INLINE.2<0;1,0>:ud
+#define INLINE_DW3 REG_INLINE.3<0;1,0>:ud
+#define INLINE_DW4 REG_INLINE.4<0;1,0>:ud
+#define INLINE_DW5 REG_INLINE.5<0;1,0>:ud
+#define INLINE_DW6 REG_INLINE.6<0;1,0>:ud
+#define INLINE_DW7 REG_INLINE.7<0;1,0>:ud
+
+// Intra macroblock in-line data
+//
+// In-line DWORD 0
+#define REG_MBAFF_FIELD REG_INLINE.1 // :uw, can be added directly to lower-word of MSGDSC
+#define MBAFF_FIELD BIT26+BIT25 // Bits 26:25 - MBAFF field macroblock flag
+ // 00 = Current macroblock is not an MBAFF field macroblock
+ // 11 = Current macroblock is an MBAFF field macroblock
+
+#define REG_FIELD_PARITY INLINE_DW0
+#define FIELD_PARITY BIT24 // Bit 24 - Macroblock field parity flag
+ // 0 = Current field is a top field
+ // 1 = Current field is a bottom field
+
+#define REG_FIELD_MACROBLOCK_FLAG INLINE_DW0
+#define FIELD_MACROBLOCK_FLAG BIT14 // Bit 14 - Field macroblock flag
+ // 0 = Current macroblock is not a field macroblock
+ // 1 = Current macroblock is a field macroblock
+#define REG_MACROBLOCK_TYPE INLINE_DW0
+#define MACROBLOCK_TYPE BIT12+BIT11+BIT10+BIT9+BIT8 // Bit 12:8 - Intra macroblock flag
+
+#define REG_CHROMA_FORMAT_IDC INLINE_DW0
+#define CHROMA_FORMAT_IDC BIT3+BIT2 // Bit 3:2 - Chroma format
+ // 00 = Luma only (Monochrome)
+ // 01 = YUV420
+ // 10 = YUV422
+ // 11 = YUV444
+#define REG_MBAFF_PIC INLINE_DW0
+#define MBAFF_PIC BIT1 // Bit 1 - MBAFF Frame picture
+ // 0 = Not an MBAFF frame picture
+ // 1 = An MBAFF frame picture
+#define REG_INTRA_PRED_8X8_BLK2_AVAIL_FLAG INLINE_DW0
+#define INTRA_PRED_8X8_BLK2_AVAIL_FLAG BIT4 // Bit 4: Pixel available for block 2 in an intra_8x8 MB.
+
+// In-line DWORD 1
+#define ORIX REG_INLINE.4 // :ub, H. origin of the macroblock in macroblock unit
+#define ORIY REG_INLINE.5 // :ub, V. origin of the macroblock in macroblock unit
+
+// In-line DWORD 2
+#define REG_CBPCYB REG_INLINE.9 // :ub, Coded block pattern
+#define REG_CBPCY INLINE_DW2 // Bits 13:8 - Coded block pattern
+ // reflect Y0, Y1, Y2, Y3, Cb4, Cr5
+ // Bit 13 - Y0
+ // Bit 12 - Y1
+ // Bit 11 - Y2
+ // Bit 10 - Y3
+ // Bit 9 - U4
+ // Bit 8 - V5
+#define CBP_MASK 0x3F00:ud // Bit mask for all CBP bits
+#define CBP_Y_MASK 0x3C00:ud // Bit mask for CBP Y bits
+#define CBP_UV_MASK 0x0300:ud // Bit mask for CBP U/V bits
+
+#define CBP_Y0_MASK BIT13:ud // Bit mask for CBP Y0 bit
+#define CBP_Y1_MASK BIT12:ud // Bit mask for CBP Y1 bit
+#define CBP_Y2_MASK BIT11:ud // Bit mask for CBP Y2 bit
+#define CBP_Y3_MASK BIT10:ud // Bit mask for CBP Y3 bit
+#define CBP_U_MASK BIT9:ud // Bit mask for CBP U bit
+#define CBP_V_MASK BIT8:ud // Bit mask for CBP V bit
+
+// In-line DWORD 3
+#define REG_INTRA_CHROMA_PRED_MODE REG_INLINE.12 // :ub - Intra chroma prediction mode
+#define INTRA_CHROMA_PRED_MODE BIT7+BIT6 // Bit 7:6 - Intra chroma prediction mode
+ // 00 = Intra DC prediction
+ // 01 = Intra horizontal prediction
+ // 10 = Intra vertical prediction
+ // 11 = Intra plane prediction
+#define INTRA_CHROMA_PRED_MODE_SHIFT 6 // Intra chroma prediction mode shift
+
+#define REG_INTRA_PRED_AVAIL_FLAG INLINE_DW3
+#define INTRA_PRED_AVAIL_FLAG BIT4+BIT3+BIT2+BIT1+BIT0 // Bits 4:0 - Intra prediction available flag
+ // Bit 0: Macroblock A (the left neighbor) entire or top half
+ // Bit 1: Macroblock B (the upper neighbor)
+ // Bit 2: Macroblock C (the above-right neighbor)
+ // Bit 3: Macroblock D (the above-left neighbor)
+ // Bit 4: Macroblock A (the left neighbor) bottom half
+ // Each bit is defined below
+ // 0 = The macroblock is not available for intra prediction
+ // 1 = The macroblock is available for intra prediction
+#define INTRA_PRED_LEFT_TH_AVAIL_FLAG BIT0 // Bit 0: Macroblock A (the left neighbor) entire or top half
+#define INTRA_PRED_UP_AVAIL_FLAG BIT1 // Bit 1: Macroblock B (the upper neighbor)
+#define INTRA_PRED_UP_RIGHT_AVAIL_FLAG BIT2 // Bit 2: Macroblock C (the above-right neighbor)
+#define INTRA_PRED_UP_LEFT_AVAIL_FLAG BIT3 // Bit 3: Macroblock D (the above-left neighbor)
+#define INTRA_PRED_LEFT_BH_AVAIL_FLAG BIT4 // Bit 4: Macroblock A (the left neighbor) bottom half
+//#define INTRA_PRED_8X8_BLK2_AVAIL_FLAG BIT5 // Bit 5: Pixel available for block 2 in an intra_8x8 MB.
+#define REG_INTRA_PRED_AVAIL_FLAG_BYTE REG_INLINE.12 // Byte location of Intra_Pred_Avail_Flag
+#define REG_INTRA_PRED_AVAIL_FLAG_WORD REG_INLINE.6 // Word location of Intra_Pred_Avail_Flag
+
+
+.declare INTRA_PRED_MODE Base=REG_INLINE.16 ElementSize=1 SrcRegion=REGION(16,1) Type=ub // Intra prediction mode
+
+// End of intra_header.inc
+
+#endif // !defined(__INTRA_HEADER__)
+
diff --git a/shaders/h264/mc/intra_Pred_4x4_Y_4.asm b/shaders/h264/mc/intra_Pred_4x4_Y_4.asm new file mode 100644 index 0000000..584d012 --- /dev/null +++ b/shaders/h264/mc/intra_Pred_4x4_Y_4.asm @@ -0,0 +1,240 @@ +/*
+ * Intra predict 4 Intra_4x4 luma blocks
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__INTRA_PRED_4X4_Y_4__) // Make sure this is only included once
+#define __INTRA_PRED_4X4_Y_4__
+
+// Module name: intra_Pred_4x4_Y_4.asm
+//
+// Intra predict 4 Intra_4x4 luma blocks
+//
+//--------------------------------------------------------------------------
+// Input data:
+//
+// REF_TOP: Top reference data stored in BYTE with p[-1,-1] at REF_TOP(0,-1)
+// REF_LEFT: Left reference data stored in BYTE with p[-1,0] at REF_LEFT(0,0)
+// PRED_MODE: Intra prediction mode stored in 4 words (4 LSB)
+// REG_INTRA_PRED_AVAIL: Top/Left available flag, (Bit0: Left, Bit1: Top)
+//
+//--------------------------------------------------------------------------
+
+#undef INTRA_PRED_AVAIL
+#undef INTRA_REF
+#undef REF_LEFT_BACK
+#undef REF_TMP
+#undef REF_TMP1
+
+#define INTRA_PRED_AVAIL REG_INTRA_TEMP_2.8
+#define INTRA_REF REG_INTRA_TEMP_2
+#define REF_LEFT_BACK REG_INTRA_TEMP_8
+#define REF_TMP REG_INTRA_TEMP_3
+#define REF_TMP1 REG_INTRA_TEMP_4
+
+intra_Pred_4x4_Y_4:
+
+ mov (8) REF_LEFT_BACK<1>:ub REF_LEFT(0)REGION(8,1) // Store left referece data
+// Set up pointers to each intra_4x4 prediction mode
+//
+ and (4) PINTRA4X4_Y<1>:w PRED_MODE<4;4,1>:w 0x0F:w
+ add (4) INTRA_4X4_MODE(0) r[PINTRA4X4_Y, INTRA_4X4_OFFSET]<1,0>:ub INTRA_MODE<4;4,1>:ub
+
+// Sub-block 0 *****************
+ mov (1) INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w // Top/Left neighbor available flags
+ CALL_1(INTRA_4X4_MODE(0),1)
+
+// Add error data to predicted intra data
+ADD_ERROR_SB0:
+ add.sat (8) r[PPREDBUF_Y,PREDSUBBLK0]<2>:ub r[PERROR,ERRBLK0]<8;4,1>:w REG_INTRA_4X4_PRED<8;8,1>:w // Too bad indexed src can't
+ add.sat (8) r[PPREDBUF_Y,PREDSUBBLK0+16]<2>:ub r[PERROR,ERRBLK0+32]<8;4,1>:w REG_INTRA_4X4_PRED.8<8;8,1>:w // cross 2 GRFs
+
+// Sub-block 1 *****************
+ mov (16) REF_TOP0(0)<1> REF_TOP0(0,4)REGION(8,1) // Top reference data
+ mov (4) REF_LEFT(0)<1> r[PPREDBUF_Y,PREDSUBBLK0+6]<8;1,0>:ub // New left referece data from sub-block 0
+ or (1) INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w 1:w // Left neighbor is available
+ CALL_1(INTRA_4X4_MODE(0,1),1)
+
+// Add error data to predicted intra data
+ADD_ERROR_SB1:
+ add.sat (8) r[PPREDBUF_Y,PREDSUBBLK1]<2>:ub r[PERROR,ERRBLK1]<8;4,1>:w REG_INTRA_4X4_PRED<8;8,1>:w // Too bad indexed src can't
+ add.sat (8) r[PPREDBUF_Y,PREDSUBBLK1+16]<2>:ub r[PERROR,ERRBLK1+32]<8;4,1>:w REG_INTRA_4X4_PRED.8<8;8,1>:w // cross 2 GRFs
+
+// Sub-block 2 *****************
+ mov (1) REF_TOP0(0,3)<1> REF_LEFT_BACK.3<0;1,0>:ub // Top-left reference data from stored left referece data
+ mov (4) REF_TOP0(0,4)<1> r[PPREDBUF_Y,PREDSUBBLK0+24]REGION(4,2):ub // Top reference data
+ mov (4) REF_TOP0(0,8)<1> r[PPREDBUF_Y,PREDSUBBLK0+24+32]REGION(4,2):ub // Too bad indexed src can't cross 2 GRFs
+ mov (4) REF_TOP0(0,12)<1> r[PPREDBUF_Y,PREDSUBBLK0+30+32]REGION(1,0):ub // Extended top-right reference data
+ mov (4) REF_LEFT(0)<1> REF_LEFT_BACK.4<4;4,1>:ub // From stored left referece data
+ or (1) INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w 2:w // Top neighbor is available
+ CALL_1(INTRA_4X4_MODE(0,2),1)
+
+// Add error data to predicted intra data
+ADD_ERROR_SB2:
+ add.sat (8) r[PPREDBUF_Y,PREDSUBBLK2]<2>:ub r[PERROR,ERRBLK2]<8;4,1>:w REG_INTRA_4X4_PRED<8;8,1>:w // Too bad indexed src can't
+ add.sat (8) r[PPREDBUF_Y,PREDSUBBLK2+16]<2>:ub r[PERROR,ERRBLK2+32]<8;4,1>:w REG_INTRA_4X4_PRED.8<8;8,1>:w // cross 2 GRFs
+
+// Sub-block 3 *****************
+ mov (16) REF_TOP0(0)<1> REF_TOP0(0,4)REGION(8,1) // Top reference data
+ mov (8) REF_TOP0(0,8)<1> REF_TOP0(0,7)<0;1,0> // Extended top-right reference data
+ mov (4) REF_LEFT(0)<1> r[PPREDBUF_Y,PREDSUBBLK2+6]<8;1,0>:ub // Left referece data from sub-block 0
+ or (1) INTRA_PRED_AVAIL<1>:w REG_INTRA_PRED_AVAIL<0;1,0>:w 3:w // Top/Left neighbor are available
+ CALL_1(INTRA_4X4_MODE(0,3),1)
+
+// Add error data to predicted intra data
+ADD_ERROR_SB3:
+ add.sat (8) r[PPREDBUF_Y,PREDSUBBLK3]<2>:ub r[PERROR,ERRBLK3]<8;4,1>:w REG_INTRA_4X4_PRED<8;8,1>:w // Too bad indexed src can't
+ add.sat (8) r[PPREDBUF_Y,PREDSUBBLK3+16]<2>:ub r[PERROR,ERRBLK3+32]<8;4,1>:w REG_INTRA_4X4_PRED.8<8;8,1>:w // cross 2 GRFs
+
+ RETURN
+
+//--------------------------------------------------------------------------
+// Actual module that performs Intra_4x4 prediction and construction
+//
+// REF_TOP: Top reference data stored in BYTE with p[-1,-1] at REF_TOP(0,-1)
+// REF_LEFT: Left reference data stored in BYTE with p[-1,0] at REF_LEFT(0,0)
+// PINTRA4X4_Y: Intra prediction mode
+// INTRA_PRED_AVAIL: Top/Left available flag, (Bit0: Left, Bit1: Top)
+//
+// Output data:
+//
+// REG_INTRA_4X4_PRED: Predicted 4x4 block data stored in 1 GRF register
+//--------------------------------------------------------------------------
+intra_Pred_4x4_Y:
+// Mode 0
+INTRA_4X4_VERTICAL:
+ mov (16) REG_INTRA_4X4_PRED<1>:w REF_TOP(0)<0;4,1>
+ RETURN_1
+
+// Mode 1
+INTRA_4X4_HORIZONTAL:
+ mov (16) REG_INTRA_4X4_PRED<1>:w REF_LEFT(0)<1;4,0>
+ RETURN_1
+
+// Mode 2
+INTRA_4X4_DC:
+// Rearrange reference samples for unified DC prediction code
+//
+ and.nz.f0.0 (16) NULLREG INTRA_PRED_AVAIL<0;1,0>:w 2:w {Compr}
+ and.nz.f0.1 (16) NULLREG INTRA_PRED_AVAIL<0;1,0>:w 1:w {Compr}
+ (-f0.0.any16h) mov (16) REF_TOP_W(0)<1> 0x8080:uw // Top macroblock not available for intra prediction
+ (-f0.1.any8h) mov (8) REF_LEFT(0)<1> REF_TOP(0)REGION(8,1) // Left macroblock not available for intra prediction
+ (-f0.0.any8h) mov (8) REF_TOP(0)<1> REF_LEFT(0)REGION(8,1) // Top macroblock not available for intra prediction
+// Perform DC prediction
+//
+ add (4) PRED_YW(15)<1> REF_TOP(0)REGION(4,1) REF_LEFT(0)REGION(4,1)
+ add (2) PRED_YW(15)<1> PRED_YW(15)REGION(2,1) PRED_YW(15,2)REGION(2,1)
+ add (16) acc0<1>:w PRED_YW(15)REGION(1,0) PRED_YW(15,1)REGION(1,0)
+ add (16) acc0<1>:w acc0:w 4:w
+ shr (16) REG_INTRA_4X4_PRED<1>:w acc0:w 3:w
+ RETURN_1
+
+// Mode 3
+INTRA_4X4_DIAG_DOWN_LEFT:
+ mov (8) INTRA_REF<1>:ub REF_TOP(0)REGION(8,1) // Keep REF_TOP untouched for future use
+ mov (4) INTRA_REF.8<1>:ub REF_TOP(0,7)REGION(4,1) // p[8,-1] = p[7,-1]
+ add (8) acc0<1>:w INTRA_REF.2<8;8,1> 2:w // p[x+2]+2
+ mac (8) acc0<1>:w INTRA_REF.1<8;8,1> 2:w // 2*p[x+1]+p[x+2]+2
+ mac (8) PRED_YW(15)<1> INTRA_REF.0<8;8,1> 1:w // p[x]+2*p[x+1]+p[x+2]+2
+
+ shr (16) REG_INTRA_4X4_PRED<1>:w PRED_YW(15)<1;4,1> 2:w // (p[x]+2*p[x+1]+p[x+2]+2)>>2
+ RETURN_1
+
+// Mode 4
+INTRA_4X4_DIAG_DOWN_RIGHT:
+
+// Set inverse shift count
+ shl (4) REF_TMP<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b
+ mov (8) INTRA_REF.4<1>:ub REF_TOP(0,-1)REGION(8,1) // INTRA_REF holds all reference data
+ mov (4) INTRA_REF<1>:ub REF_TMP.3<16;4,4>:ub
+
+ add (8) acc0<1>:w INTRA_REF.2<8;8,1>:ub 2:w // p[x+2]+2
+ mac (8) acc0<1>:w INTRA_REF.1<8;8,1>:ub 2:w // 2*p[x+1]+p[x+2]+2
+ mac (8) INTRA_REF<1>:w INTRA_REF<8;8,1>:ub 1:w // p[x]+2*p[x+1]+p[x+2]+2
+
+// Store data in reversed order
+ add (4) PBWDCOPY_4<1>:w INV_TRANS4<4;4,1>:b INTRA_TEMP_2*GRFWIB:w // Must match with INTRA_REF
+ shr (16) REG_INTRA_4X4_PRED<1>:w r[PBWDCOPY_4]<4,1>:w 2:w
+ RETURN_1
+
+// Mode 5
+INTRA_4X4_VERT_RIGHT:
+
+// Set inverse shift count
+ shl (4) REF_TMP<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b
+ mov (8) INTRA_REF.4<1>:ub REF_TOP(0,-1)REGION(8,1) // INTRA_REF holds all reference data
+ mov (4) INTRA_REF<1>:ub REF_TMP.3<16;4,4>:ub
+
+ // Even rows
+ avg (8) PRED_YW(14)<1> INTRA_REF.4<8;8,1> INTRA_REF.5<8;8,1> // avg(p[x-1],p[x])
+ // Odd rows
+ add (8) acc0<1>:w INTRA_REF.3<8;8,1>:ub 2:w // p[x]+2
+ mac (8) acc0<1>:w INTRA_REF.2<8;8,1>:ub 2:w // 2*p[x-1]+p[x]+2
+ mac (8) acc0<1>:w INTRA_REF.1<8;8,1>:ub 1:w // p[x-2]+2*p[x-1]+p[x]+2
+ shr (8) INTRA_REF<1>:w acc0:w 2:w // (p[x-2]+2*p[x-1]+p[x]+2)>>2
+
+ mov (4) INTRA_REF.2<2>:w INTRA_REF.2<4;4,1>:w // Keep zVR = -2,-3 unchanged
+ mov (4) INTRA_REF.3<2>:w PRED_YW(14)REGION(4,1) // Combining even rows
+
+ add (4) PBWDCOPY_4<1>:w INV_TRANS4<4;4,1>:b INTRA_TEMP_2*GRFWIB:w // Must match with INTRA_REF
+ mov (16) REG_INTRA_4X4_PRED<1>:w r[PBWDCOPY_4]<4,2>:w
+ RETURN_1
+
+// Mode 6
+INTRA_4X4_HOR_DOWN:
+// Set inverse shift count
+ shl (4) REF_TMP<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b
+ mov (8) INTRA_REF.4<1>:ub REF_TOP(0,-1)REGION(8,1) // INTRA_REF holds all reference data
+ mov (4) INTRA_REF<1>:ub REF_TMP.3<16;4,4>:ub
+
+ // Even pixels
+ avg (8) PRED_YW(14)<1> INTRA_REF<8;8,1> INTRA_REF.1<8;8,1> // avg(p[y-1],p[y])
+ // Odd pixels
+ add (8) acc0<1>:w INTRA_REF.2<8;8,1>:ub 2:w // p[y]+2
+ mac (8) acc0<1>:w INTRA_REF.1<8;8,1>:ub 2:w // 2*p[y-1]+p[y]+2
+ mac (8) REF_TMP<1>:w INTRA_REF.0<8;8,1>:ub 1:w // p[y-2]+2*p[y-1]+p[y]+2
+ shr (4) INTRA_REF.1<2>:w REF_TMP<4;4,1>:w 2:w // (p[y-2]+2*p[y-1]+p[y]+2)>>2
+
+ shr (2) INTRA_REF.8<1>:w REF_TMP.4<2;2,1>:w 2:w // Keep zVR = -2,-3 unchanged
+ mov (4) INTRA_REF.0<2>:w PRED_YW(14)REGION(4,1) // Combining even pixels
+
+ shl (4) PBWDCOPY_4<1>:w INV_TRANS4<4;4,1>:b 1:w // Convert to WORD offset
+ add (4) PBWDCOPY_4<1>:w PBWDCOPY_4<4;4,1>:w INTRA_TEMP_2*GRFWIB:w // Must match with INTRA_REF
+ mov (16) REG_INTRA_4X4_PRED<1>:w r[PBWDCOPY_4]<4,1>:w
+ RETURN_1
+
+// Mode 7
+INTRA_4X4_VERT_LEFT:
+ // Even rows
+ avg (8) PRED_YW(14)<2> REF_TOP(0)REGION(8,1) REF_TOP(0,1)REGION(8,1) // avg(p[x],p[x+1])
+ // Odd rows
+ add (8) acc0<1>:w REF_TOP(0,2)REGION(8,1) 2:w // p[x+2]+2
+ mac (8) acc0<1>:w REF_TOP(0,1)REGION(8,1) 2:w // 2*p[x+1]+p[x+2]+2
+ mac (8) PRED_YW(15)<1> REF_TOP(0)REGION(8,1) 1:w // p[x]+2*p[x+1]+p[x+2]+2
+ shr (8) PRED_YW(14,1)<2> PRED_YW(15)REGION(8,1) 2:w
+
+ mov (16) REG_INTRA_4X4_PRED<1>:w PRED_YW(14)<1;4,2>
+ RETURN_1
+
+// Mode 8
+INTRA_4X4_HOR_UP:
+// Set extra left reference pixels for unified prediction
+ mov (8) REF_LEFT(0,4)<1> REF_LEFT(0,3)REGION(1,0) // Copy p[-1,3] to p[-1,y],y=4...7
+ // Even pixels
+ avg (8) PRED_YW(14)<2> REF_LEFT(0)REGION(8,1) REF_LEFT(0,1)REGION(8,1) // avg(p[y],p[y+1])
+ // Odd pixels
+ add (8) acc0<1>:w REF_LEFT(0,2)REGION(8,1) 2:w // p[y+2]+2
+ mac (8) acc0<1>:w REF_LEFT(0,1)REGION(8,1) 2:w // 2*p[y+1]+p[y+2]+2
+ mac (8) PRED_YW(15)<1> REF_LEFT(0)REGION(8,1) 1:w // p[y]+2*p[y+1]+p[y+2]+2
+ shr (8) PRED_YW(14,1)<2> PRED_YW(15)REGION(8,1) 2:w // (p[y]+2*p[y+1]+p[y+2]+2)>>2
+
+ mov (16) REG_INTRA_4X4_PRED<1>:w PRED_YW(14)<2;4,1>
+ RETURN_1
+
+// End of intra_Pred_4x4_Y_4
+
+#endif // !defined(__INTRA_PRED_4X4_Y_4__)
diff --git a/shaders/h264/mc/intra_Pred_8x8_Y.asm b/shaders/h264/mc/intra_Pred_8x8_Y.asm new file mode 100644 index 0000000..ce77771 --- /dev/null +++ b/shaders/h264/mc/intra_Pred_8x8_Y.asm @@ -0,0 +1,246 @@ +/*
+ * Intra predict 8X8 luma block
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__INTRA_PRED_8X8_Y__) // Make sure this is only included once
+#define __INTRA_PRED_8X8_Y__
+
+// Module name: intra_Pred_8X8_Y.asm
+//
+// Intra predict 8X8 luma block
+//
+//--------------------------------------------------------------------------
+// Input data:
+//
+// REF_TOP: Top reference data stored in BYTE with p[-1,-1] at REF_TOP(0,-1), p[-1,-1] and [15,-1] adjusted
+// REF_LEFT: Left reference data stored in BYTE with p[-1,0] at REF_LEFT(0,2), REF_LEFT(0,1) (p[-1,-1]) adjusted
+// PRED_MODE: Intra prediction mode stored in 4 LSBs
+// INTRA_PRED_AVAIL: Top/Left available flag, (Bit0: Left, Bit1: Top)
+//
+// Output data:
+//
+// REG_INTRA_8X8_PRED: Predicted 8X8 block data
+//--------------------------------------------------------------------------
+
+#define INTRA_REF REG_INTRA_TEMP_1
+#define REF_TMP REG_INTRA_TEMP_2
+
+intra_Pred_8x8_Y:
+
+// Reference sample filtering
+//
+ // Set up boundary pixels for unified filtering
+ mov (1) REF_TOP(0,16)<1> REF_TOP(0,15)REGION(1,0) // p[16,-1] = p[15,-1]
+ mov (8) REF_LEFT(0,2+8)<1> REF_LEFT(0,2+7)REGION(1,0) // p[-1,8] = p[-1,7]
+
+ // Top reference sample filtering (!!Consider instruction compression later)
+ add (16) acc0<1>:w REF_TOP(0,-1)REGION(16,1) 2:w // p[x-1,-1]+2
+ mac (16) acc0<1>:w REF_TOP(0)REGION(16,1) 2:w // p[x-1,-1]+2*p[x,-1]+2
+ mac (16) acc0<1>:w REF_TOP(0,1)REGION(16,1) 1:w // p[x-1,-1]+2*p[x,-1]+p[x+1,-1]+2
+ shr (16) REF_TMP<1>:w acc0:w 2:w // (p[x-1,-1]+2*p[x,-1]+p[x+1,-1]+2)>>2
+
+ // Left reference sample filtering
+ add (16) acc0<1>:w REF_LEFT(0)REGION(16,1) 2:w // p[-1,y-1]+2
+ mac (16) acc0<1>:w REF_LEFT(0,1)REGION(16,1) 2:w // p[-1,y-1]+2*p[-1,y]+2
+ mac (16) acc0<1>:w REF_LEFT(0,2)REGION(16,1) 1:w // p[-1,y-1]+2*p[-1,y]+p[-1,y+1]+2
+ shr (16) INTRA_REF<1>:w acc0:w 2:w // (p[-1,y-1]+2*p[-1,y]+p[-1,y+1]+2)>>2
+
+ // Re-assign filtered reference samples
+ mov (16) REF_TOP(0)<1> REF_TMP<32;16,2>:ub // p'[x,-1], x=0...15
+ mov (8) REF_LEFT(0)<1> INTRA_REF.2<16;8,2>:ub // p'[-1,y], y=0...7
+ mov (1) REF_TOP(0,-1)<1> INTRA_REF<0;1,0>:ub // p'[-1,-1]
+
+// Select intra_8x8 prediction mode
+//
+ and (1) PINTRAPRED_Y<1>:w PRED_MODE<0;1,0>:w 0x0F:w
+ // WA for "jmpi" restriction
+ mov (1) REG_INTRA_TEMP_1<1>:ud r[PINTRAPRED_Y, INTRA_8X8_OFFSET]:ub
+ jmpi (1) REG_INTRA_TEMP_1<0;1,0>:d
+
+// Mode 0
+#define PTMP a0.6
+#define PTMP_D a0.3
+INTRA_8X8_VERTICAL:
+ $for(0,0; <4; 1,32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PERROR,%2]<16;16,1>:w REF_TOP(0)<0;8,1>
+ }
+ RETURN
+
+// Mode 1
+INTRA_8X8_HORIZONTAL:
+ $for(0,0; <8; 2,32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PERROR,%2]<16;16,1>:w REF_LEFT(0,%1)<1;8,0>
+ }
+ RETURN
+
+// Mode 2
+INTRA_8X8_DC:
+// Rearrange reference samples for unified DC prediction code
+//
+ and.nz.f0.0 (16) NULLREG INTRA_PRED_AVAIL<0;1,0>:w 2:w // Top macroblock available for intra prediction?
+ and.nz.f0.1 (8) NULLREG INTRA_PRED_AVAIL<0;1,0>:w 1:w // Left macroblock available for intra prediction?
+ (-f0.0.any16h) mov (16) REF_TOP_W(0)<1> 0x8080:uw
+ (-f0.1.any8h) mov (8) REF_LEFT(0)<1> REF_TOP(0)REGION(8,1)
+ (-f0.0.any8h) mov (8) REF_TOP(0)<1> REF_LEFT(0)REGION(8,1)
+
+// Perform DC prediction
+//
+ add (8) PRED_YW(15)<1> REF_TOP(0)REGION(8,1) REF_LEFT(0)REGION(8,1)
+ add (4) PRED_YW(15)<1> PRED_YW(15)REGION(4,1) PRED_YW(15,4)REGION(4,1)
+ add (2) PRED_YW(15)<1> PRED_YW(15)REGION(2,1) PRED_YW(15,2)REGION(2,1)
+ add (16) acc0<1>:w PRED_YW(15)REGION(1,0) PRED_YW(15,1)REGION(1,0)
+ add (16) acc0<1>:w acc0:w 8:w
+ shr (16) REG_INTRA_TEMP_0<1>:w acc0:w 4:w
+
+ // Add error block
+ $for(0,0; <4; 1,32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PERROR,%2]<16;16,1>:w REG_INTRA_TEMP_0<16;16,1>:w
+ }
+ RETURN
+
+// Mode 3
+INTRA_8X8_DIAG_DOWN_LEFT:
+ mov (8) REF_TOP(0,16)<1> REF_TOP(0,15)REGION(8,1) // p[16,-1] = p[15,-1]
+ add (16) acc0<1>:w REF_TOP(0,2)REGION(16,1) 2:w // p[x+2]+2
+ mac (16) acc0<1>:w REF_TOP(0,1)REGION(16,1) 2:w // 2*p[x+1]+p[x+2]+2
+ mac (16) acc0<1>:w REF_TOP(0)REGION(16,1) 1:w // p[x]+2*p[x+1]+p[x+2]+2
+ shr (16) REG_INTRA_TEMP_0<1>:w acc0<16;16,1>:w 2:w // (p[x]+2*p[x+1]+p[x+2]+2)>>2
+
+ // Add error block
+ $for(0,0; <8; 2,32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PERROR,%2]<16;16,1>:w REG_INTRA_TEMP_0.%1<1;8,1>:w
+ }
+ RETURN
+
+// Mode 4
+INTRA_8X8_DIAG_DOWN_RIGHT:
+#define INTRA_REF REG_INTRA_TEMP_1
+#define REF_TMP REG_INTRA_TEMP_2
+
+// Set inverse shift count
+ shl (4) REF_TMP<1>:ud REF_LEFT_D(0,1)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order bottom 4 pixels of left ref.
+ shl (4) REF_TMP.4<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order top 4 pixels of left ref.
+ mov (8) INTRA_REF<1>:ub REF_TMP.3<32;8,4>:ub
+ mov (16) INTRA_REF.8<1>:ub REF_TOP(0,-1)REGION(16,1) // INTRA_REF holds all reference data
+
+ add (16) acc0<1>:w INTRA_REF.2<16;16,1>:ub 2:w // p[x+2]+2
+ mac (16) acc0<1>:w INTRA_REF.1<16;16,1>:ub 2:w // 2*p[x+1]+p[x+2]+2
+ mac (16) acc0<1>:w INTRA_REF<16;16,1>:ub 1:w // p[x]+2*p[x+1]+p[x+2]+2
+ shr (16) INTRA_REF<1>:w acc0<16;16,1>:w 2:w // (p[x]+2*p[x+1]+p[x+2]+2)>>2
+
+// Store data in reversed order
+ add (2) PBWDCOPY_8<1>:w INV_TRANS48<2;2,1>:b INTRA_TEMP_1*GRFWIB:w // Must match with INTRA_REF
+
+ // Add error block
+ $for(0,96; <8; 2,-32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PBWDCOPY_8,%1*2]<8,1>:w r[PERROR,%2]<16;16,1>:w
+ }
+ RETURN
+
+// Mode 5
+INTRA_8X8_VERT_RIGHT:
+#define INTRA_REF REG_INTRA_TEMP_1
+#define REF_TMP REG_INTRA_TEMP_2
+#define REF_TMP1 REG_INTRA_TEMP_3
+
+// Set inverse shift count
+ shl (4) REF_TMP<1>:ud REF_LEFT_D(0,1)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order bottom 4 pixels of left ref.
+ shl (4) REF_TMP.4<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order top 4 pixels of left ref.
+ mov (8) INTRA_REF<1>:ub REF_TMP.3<32;8,4>:ub
+ mov (16) INTRA_REF.8<1>:ub REF_TOP(0,-1)REGION(16,1) // INTRA_REF holds all reference data
+
+ // Even rows
+ avg (16) PRED_YW(14)<1> INTRA_REF.8<16;16,1> INTRA_REF.9<16;16,1> // avg(p[x-1],p[x])
+ // Odd rows
+ add (16) acc0<1>:w INTRA_REF.3<16;16,1>:ub 2:w // p[x]+2
+ mac (16) acc0<1>:w INTRA_REF.2<16;16,1>:ub 2:w // 2*p[x-1]+p[x]+2
+ mac (16) acc0<1>:w INTRA_REF.1<16;16,1>:ub 1:w // p[x-2]+2*p[x-1]+p[x]+2
+ shr (16) REF_TMP<1>:w acc0:w 2:w // (p[x-2]+2*p[x-1]+p[x]+2)>>2
+
+ mov (8) INTRA_REF<1>:ub REF_TMP<16;8,2>:ub // Keep zVR = -1,-2,-3,-4,-5,-6,-7 sequencially
+ mov (8) INTRA_REF.6<2>:ub REF_TMP.12<16;8,2>:ub // Keep zVR = -1,1,3,5,7,9,11,13 at even byte
+ mov (8) INTRA_REF.7<2>:ub PRED_Y(14)REGION(8,2) // Combining zVR = 0,2,4,6,8,10,12,14 at odd byte
+
+ add (2) PBWDCOPY_8<1>:w INV_TRANS8<2;2,1>:b INTRA_TEMP_1*GRFWIB:w // Must match with INTRA_REF
+
+ // Add error block
+ $for(0,96; <8; 2,-32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PBWDCOPY_8,%1]<8,2>:ub r[PERROR,%2]<16;16,1>:w
+ }
+ RETURN
+
+// Mode 6
+INTRA_8X8_HOR_DOWN:
+// Set inverse shift count
+ shl (4) REF_TMP<1>:ud REF_LEFT_D(0,1)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order bottom 4 pixels of left ref.
+ shl (4) REF_TMP.4<1>:ud REF_LEFT_D(0)REGION(1,0) INV_SHIFT<4;4,1>:b // Reverse order top 4 pixels of left ref.
+ mov (8) INTRA_REF<1>:ub REF_TMP.3<16;4,4>:ub
+ mov (16) INTRA_REF.8<1>:ub REF_TOP(0,-1)REGION(16,1) // INTRA_REF holds all reference data
+
+ // Odd pixels
+ add (16) acc0<1>:w INTRA_REF.2<16;16,1>:ub 2:w // p[y]+2
+ mac (16) acc0<1>:w INTRA_REF.1<16;16,1>:ub 2:w // 2*p[y-1]+p[y]+2
+ mac (16) acc0<1>:w INTRA_REF.0<16;16,1>:ub 1:w // p[y-2]+2*p[y-1]+p[y]+2
+ shr (16) PRED_YW(14)<1> acc0:w 2:w // (p[y-2]+2*p[y-1]+p[y]+2)>>2
+ // Even pixels
+ avg (16) INTRA_REF<1>:w INTRA_REF<16;16,1>:ub INTRA_REF.1<16;16,1>:ub // avg(p[y-1],p[y])
+
+ mov (8) INTRA_REF.1<2>:ub PRED_Y(14)REGION(8,2) // Combining odd pixels to form byte type
+ mov (8) INTRA_REF.16<1>:ub PRED_Y(14,16)REGION(8,2) // Keep zVR = -2,-3,-4,-5,-6,-7 unchanged
+ // Now INTRA_REF.0 - INTRA_REF.21 contain predicted data
+
+ add (2) PBWDCOPY_8<1>:w INV_TRANS48<2;2,1>:b INTRA_TEMP_1*GRFWIB:w // Must match with INTRA_REF
+
+ // Add error block
+ $for(0,96; <13; 4,-32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub r[PBWDCOPY_8,%1]<8,1>:ub r[PERROR,%2]<16;16,1>:w
+ }
+ RETURN
+
+// Mode 7
+INTRA_8X8_VERT_LEFT:
+ // Even rows
+ avg (16) PRED_YW(14)<1> REF_TOP(0)REGION(16,1) REF_TOP(0,1)REGION(16,1) // avg(p[x],p[x+1])
+ // Odd rows
+ add (16) acc0<1>:w REF_TOP(0,2)REGION(16,1) 2:w // p[x+2]+2
+ mac (16) acc0<1>:w REF_TOP(0,1)REGION(16,1) 2:w // 2*p[x+1]+p[x+2]+2
+ mac (16) acc0<1>:w REF_TOP(0)REGION(16,1) 1:w // p[x]+2*p[x+1]+p[x+2]+2
+ shr (16) PRED_YW(15)<1> acc0<1>:w 2:w // (p[x]+2*p[x+1]+p[x+2]+2)>>2
+
+ // Add error block
+ $for(0,0; <4; 1,32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub PRED_YW(14,%1)<16;8,1> r[PERROR,%2]<16;16,1>:w
+ }
+ RETURN
+
+// Mode 8
+INTRA_8X8_HOR_UP:
+// Set extra left reference pixels for unified prediction
+ mov (8) REF_LEFT(0,8)<1> REF_LEFT(0,7)REGION(1,0) // Copy p[-1,7] to p[-1,y],y=8...15
+
+ // Even pixels
+ avg (16) PRED_YW(14)<1> REF_LEFT(0)REGION(16,1) REF_LEFT(0,1)REGION(16,1) // avg(p[y],p[y+1])
+ // Odd pixels
+ add (16) acc0<1>:w REF_LEFT(0,2)REGION(16,1) 2:w // p[y+2]+2
+ mac (16) acc0<1>:w REF_LEFT(0,1)REGION(16,1) 2:w // 2*p[y+1]+p[y+2]+2
+ mac (16) acc0<1>:w REF_LEFT(0)REGION(16,1) 1:w // p[y]+2*p[y+1]+p[y+2]+2
+ shr (16) PRED_YW(15)<1> acc0<1>:w 2:w // (p[y]+2*p[y+1]+p[y+2]+2)>>2
+
+ // Merge even/odd pixels
+ // The predicted data need to be stored in byte type (22 bytes are required)
+ mov (16) PRED_Y(14,1)<2> PRED_Y(15)REGION(16,2)
+
+ // Add error block
+ $for(0,0; <4; 1,32) {
+ add.sat (16) r[PPREDBUF_Y,%2]<2>:ub PRED_Y(14,%1*4)<2;8,1> r[PERROR,%2]<16;16,1>:w
+ }
+ RETURN
+
+// End of intra_Pred_8X8_Y
+
+#endif // !defined(__INTRA_PRED_8X8_Y__)
diff --git a/shaders/h264/mc/intra_Pred_Chroma.asm b/shaders/h264/mc/intra_Pred_Chroma.asm new file mode 100644 index 0000000..a1e1697 --- /dev/null +++ b/shaders/h264/mc/intra_Pred_Chroma.asm @@ -0,0 +1,155 @@ +/*
+ * Intra predict 8x8 chroma block
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__INTRA_PRED_CHROMA__) // Make sure this is only included once
+#define __INTRA_PRED_CHROMA__
+
+// Module name: intra_Pred_Chroma.asm
+//
+// Intra predict 8x8 chroma block
+//
+
+ shr (1) PINTRAPRED_UV<1>:w REG_INTRA_CHROMA_PRED_MODE<0;1,0>:ub INTRA_CHROMA_PRED_MODE_SHIFT:w // Bits 1:0 = intra chroma pred mode
+ // WA for "jmpi" restriction
+ mov (1) REG_INTRA_TEMP_1<1>:d r[PINTRAPRED_UV, INTRA_CHROMA_OFFSET]:b
+ jmpi (1) REG_INTRA_TEMP_1<0;1,0>:d
+
+// Mode 0
+INTRA_CHROMA_DC:
+ and.nz.f0.0 (8) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_UP_AVAIL_FLAG:ud // Top macroblock available for intra prediction?
+
+// Calculate DC values for sub-block 0 and 3
+//
+// Rearrange reference samples for unified DC prediction code
+// Need to check INTRA_PRED_LEFT_TH_AVAIL_FLAG for blk0 and INTRA_PRED_LEFT_BH_AVAIL_FLAG for blk3
+//
+ (-f0.0.any8h) mov (8) INTRA_REF_TOP_W(0)<1> 0x8080:uw // Up not available
+
+ and.nz.f0.1 (4) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_TH_AVAIL_FLAG:ud
+ (-f0.1.any4h) mov (4) INTRA_REF_LEFT_W(0)<2> INTRA_REF_TOP_W(0)REGION(4,1) // Left top half macroblock not available for intra prediction
+ and.nz.f0.1 (4) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_BH_AVAIL_FLAG:ud
+ (-f0.1.any4h) mov (4) INTRA_REF_LEFT_W(0,8)<2> INTRA_REF_TOP_W(0,4)REGION(4,1) // Left bottom half macroblock not available for intra prediction
+
+ (-f0.0.any8h) mov (8) INTRA_REF_TOP_W(0)<1> INTRA_REF_LEFT_W(0)REGION(8,2) // Up not available
+// Calculate DC prediction
+//
+ add (16) PRED_UVW(0)<1> INTRA_REF_TOP(0)REGION(16,1) INTRA_REF_LEFT_UV(0)<4;2,1> // Sum of top and left reference
+ add (8) PRED_UVW(0)<1> PRED_UVW(0)<4;2,1> PRED_UVW(0,2)<4;2,1> // Sum of first half (blk #0) and second half (blk #3)
+
+ add (8) PRED_UVW(9)<1> PRED_UVW(0)<0;2,1> PRED_UVW(0,2)<0;2,1> // Sum of blk #0
+ add (8) PRED_UVW(11,8)<1> PRED_UVW(0,4)<0;2,1> PRED_UVW(0,6)<0;2,1> // Sum of blk #3
+
+// Calculate DC values for sub-block 1 and 2
+//
+// Rearrange reference samples for unified DC prediction code
+//
+ // Blk #2
+ (-f0.0.any4h) mov (4) INTRA_REF_TOP_W(0)<1> 0x8080:uw
+ (f0.1.any4h) mov (4) INTRA_REF_TOP_W(0)<1> INTRA_REF_LEFT_W(0,8)REGION(4,2) // Always use available left reference
+ (-f0.1.any4h) mov (4) INTRA_REF_LEFT_W(0,8)<2> INTRA_REF_TOP_W(0)REGION(4,1)
+
+ // Blk #1
+ and.nz.f0.1 (4) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_TH_AVAIL_FLAG:ud
+ (-f0.1.any4h) mov (4) INTRA_REF_LEFT_W(0)<2> 0x8080:uw
+ (f0.0.any4h) mov (4) INTRA_REF_LEFT_W(0)<2> INTRA_REF_TOP_W(0,4)REGION(4,1) // Always use available top reference
+ (-f0.0.any4h) mov (4) INTRA_REF_TOP_W(0,4)<1> INTRA_REF_LEFT_W(0)REGION(4,2)
+
+// Calculate DC prediction
+//
+ add (8) PRED_UVW(0)<1> INTRA_REF_TOP(0)REGION(8,1) INTRA_REF_LEFT_UV(0,16)<4;2,1> // Sum of top and left reference for blk #2
+ add (8) PRED_UVW(0,8)<1> INTRA_REF_LEFT_UV(0)<4;2,1> INTRA_REF_TOP(0,8)REGION(8,1) // Sum of top and left reference for blk #1
+ add (8) PRED_UVW(0)<1> PRED_UVW(0)<4;2,1> PRED_UVW(0,2)<4;2,1> // Sum of first half (blk #2) and second half (blk #1)
+
+ add (8) PRED_UVW(9,8)<1> PRED_UVW(0,4)<0;2,1> PRED_UVW(0,6)<0;2,1> // Sum of blk #1
+ add (8) PRED_UVW(11)<1> PRED_UVW(0)<0;2,1> PRED_UVW(0,2)<0;2,1> // Sum of blk #2
+
+// Now, PRED_UVW(9) holds sums for blks #0 and #1 and PRED_UVW(11) holds sums for blks #2 and #3
+//
+ add (32) acc0<1>:w PRED_UVW(9)REGION(16,1) 4:w {Compr} // Add rounder
+ $for(0; <4; 2) {
+ shr (32) PRED_UVW(%1)<1> acc0:w 3:w {Compr}
+ }
+
+ add (32) acc0<1>:w PRED_UVW(11)REGION(16,1) 4:w {Compr} // Add rounder
+ $for(4; <8; 2) {
+ shr (32) PRED_UVW(%1)<1> acc0:w 3:w {Compr}
+ }
+ jmpi (1) End_of_intra_Pred_Chroma
+
+// Mode 1
+INTRA_CHROMA_HORIZONTAL:
+ mov (1) PREF_LEFT_UD<1>:ud INTRA_REF_LEFT_ID*GRFWIB*0x00010001+0x00040000:ud // Set address registers for instruction compression
+ $for(0,0; <8; 2,8) {
+ mov (32) PRED_UVW(%1)<1> r[PREF_LEFT,%2+2]<0;2,1>:ub {Compr} // Actual left column reference data start at offset 2
+ }
+ jmpi (1) End_of_intra_Pred_Chroma
+
+// Mode 2
+INTRA_CHROMA_VERTICAL:
+ $for(0; <8; 2) {
+ mov (32) PRED_UVW(%1)<1> INTRA_REF_TOP(0) {Compr}
+ }
+ jmpi (1) End_of_intra_Pred_Chroma
+
+// Mode 3
+INTRA_Chroma_PLANE:
+// Refer to H.264/AVC spec Section 8.3.4.4
+
+#undef C
+
+#define A REG_INTRA_TEMP_2.0 // All are WORD type
+#define B REG_INTRA_TEMP_3.0 // B[U] & B[V]
+#define C REG_INTRA_TEMP_3.2 // C[U] & C[V]
+#define YP REG_INTRA_TEMP_0 // Store intermediate results of c*(y-3). Make sure it's an even GRF
+#define YP1 REG_INTRA_TEMP_1 // Store intermediate results of c*(y-3). Make sure it's an odd GRF
+#define XP REG_INTRA_TEMP_5 // Store intermediate results of a+b*(x-3)+16. Make sure it's an odd GRF
+
+// First Calculate constants H and V
+// H1 = sum((x'+1)*p[4+x',-1]), x'=0,1,2,3
+// H2 = sum((-x'-1)*p[2-x',-1]), x'=3,2,1,0
+// H = H1 + H2
+// The same calculation holds for V
+//
+ mul (8) H1(0)<1> INTRA_REF_TOP(0,8)REGION(8,1) 0x44332211:v
+ mul (8) H2(0)<1> INTRA_REF_TOP(0,-2)REGION(8,1) 0xFFEEDDCC:v
+
+ mul (8) V1(0)<1> INTRA_REF_LEFT_UV(0,4*4)<4;2,1> 0x44332211:v
+ mul (8) V2(0)<1> INTRA_REF_LEFT_UV(0)<4;2,1> 0x00FFEEDD:v
+ mul (2) V2(0,6)<1> INTRA_REF_TOP(0,-2)REGION(2,1) -4:w // Replace 0*p[-1,3] with -4*p[-1,-1]
+ // Now, REG_INTRA_TEMP_0 holds [H2, H1] and REG_INTRA_TEMP_1 holds [V2, V1]
+
+ // Sum up [H2, H1] and [V2, V1] using instruction compression
+ // ExecSize = 16 is restricted by B-spec for instruction compression
+ // Actual intermediate results are in lower sub-registers after each summing step
+ add (16) H1(0)<1> H1(0) H2(0) {Compr} // Results in lower 8 WORDs
+ add (16) H1(0)<1> H1(0) H1(0,4) {Compr} // Results in lower 4 WORDs
+ add (16) H1(0)<1> H1(0) H1(0,2) {Compr} // Results in lower 2 WORDs
+
+// Calculate a, b, c and further derivations
+ mov (16) acc0<1>:w 32:w
+ mac (4) acc0<1>:w H1(0)<16;2,1> 34:w
+ shr (4) B<1>:w acc0:w 6:w // Done b,c
+ mov (16) acc0<1>:w 16:w
+ mac (16) acc0<1>:w INTRA_REF_TOP(0,7*2)<0;2,1> 16:w
+ mac (16) A<1>:w INTRA_REF_LEFT_UV(0,7*4)<0;2,1> 16:w // A = a+16
+ mac (16) XP<1>:w B<0;2,1>:w XY_3<1;2,0>:b // XP = A+b*(x-3)
+ mul (8) YP<1>:w C<0;2,1>:w XY_3<2;2,0>:b // YP = c*(y-3), Even portion
+ mul (8) YP1<1>:w C<0;2,1>:w XY_3_1<2;2,0>:b // YP = c*(y-3), Odd portion
+
+// Finally the intra_Chroma plane prediction
+ $for(0; <8; 2) {
+ add (32) acc0<1>:w XP<16;16,1>:w YP.%1<0;2,1>:w {Compr}
+ shr.sat (32) PRED_UV(%1)<2> acc0<16;16,1>:w 5:w {Compr}
+ }
+
+End_of_intra_Pred_Chroma:
+
+// End of intra_Pred_Chroma
+
+#endif // !defined(__INTRA_PRED_CHROMA__)
diff --git a/shaders/h264/mc/intra_pred_16x16_Y.asm b/shaders/h264/mc/intra_pred_16x16_Y.asm new file mode 100644 index 0000000..94ccb07 --- /dev/null +++ b/shaders/h264/mc/intra_pred_16x16_Y.asm @@ -0,0 +1,111 @@ +/*
+ * Intra predict 16x16 luma block
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: intra_Pred_16x16_Y.asm
+//
+// Intra predict 16x16 luma block
+//
+ and (1) PINTRAPRED_Y<1>:w INTRA_PRED_MODE(0)REGION(1,0) 0x0F:w
+ // WA for "jmpi" restriction
+ mov (1) REG_INTRA_TEMP_1<1>:ud r[PINTRAPRED_Y, INTRA_16X16_OFFSET]:ub
+ jmpi (1) REG_INTRA_TEMP_1<0;1,0>:d
+
+// Mode 0
+INTRA_16x16_VERTICAL:
+ $for(0; <16; 2) {
+ mov (32) PRED_YW(%1)<1> INTRA_REF_TOP(0) {Compr}
+ }
+ jmpi (1) End_intra_Pred_16x16_Y
+
+// Mode 1
+INTRA_16x16_HORIZONTAL:
+ mov (1) PREF_LEFT_UD<1>:ud INTRA_REF_LEFT_ID*GRFWIB*0x00010001+0x00040000:ud // Set address registers for instruction compression
+ $for(0,0; <16; 2,8) {
+ mov (32) PRED_YW(%1)<1> r[PREF_LEFT,%2+3]<0;1,0>:ub {Compr} // Actual left column reference data start at offset 3
+ }
+ jmpi (1) End_intra_Pred_16x16_Y
+
+// Mode 2
+INTRA_16x16_DC:
+ and.nz.f0.0 (8) NULLREG REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_UP_AVAIL_FLAG:ud // Top macroblock available for intra prediction?
+ and (8) acc0<1>:ud REG_INTRA_PRED_AVAIL_FLAG INTRA_PRED_LEFT_TH_AVAIL_FLAG+INTRA_PRED_LEFT_BH_AVAIL_FLAG:ud // Left macroblock available for intra prediction?
+ xor.z.f0.1 (8) NULLREG acc0:ud INTRA_PRED_LEFT_TH_AVAIL_FLAG+INTRA_PRED_LEFT_BH_AVAIL_FLAG:ud // Left macroblock available for intra prediction?
+// Rearrange reference samples for unified DC prediction code
+//
+ (-f0.0.any8h) mov (8) INTRA_REF_TOP_W(0)<1> 0x8080:uw
+ (-f0.1.any8h) mov (8) INTRA_REF_LEFT(0)<4> INTRA_REF_TOP(0)REGION(8,1)
+ (-f0.1.any8h) mov (8) INTRA_REF_LEFT(1)<4> INTRA_REF_TOP(0,8)REGION(8,1)
+ (-f0.0.any8h) mov (8) INTRA_REF_TOP(0)<1> INTRA_REF_LEFT(0)REGION(8,4)
+ (-f0.0.any8h) mov (8) INTRA_REF_TOP(0,8)<1> INTRA_REF_LEFT(1)REGION(8,4) // Split due to HW limitation
+// Perform DC prediction
+//
+ add (16) PRED_YW(15)<1> INTRA_REF_LEFT(0)REGION(8,4) INTRA_REF_TOP(0)REGION(16,1)
+ add (8) PRED_YW(15)<1> PRED_YW(15)REGION(8,1) PRED_YW(15,8)REGION(8,1)
+ add (4) PRED_YW(15)<1> PRED_YW(15)REGION(4,1) PRED_YW(15,4)REGION(4,1)
+ add (2) PRED_YW(15)<1> PRED_YW(15)REGION(2,1) PRED_YW(15,2)REGION(2,1)
+ add (32) acc0<1>:w PRED_YW(15)REGION(1,0) PRED_YW(15,1)REGION(1,0) {Compr} // Set up both acc0 and acc1
+ add (32) acc0<1>:w acc0:w 16:w {Compr}
+
+ $for(0; <16; 2) {
+ shr (32) PRED_YW(%1)<1> acc0:w 5:w {Compr}
+ }
+ jmpi (1) End_intra_Pred_16x16_Y
+
+// Mode 3
+INTRA_16x16_PLANE:
+// Refer to H.264/AVC spec Section 8.3.3.4
+
+#define A REG_INTRA_TEMP_2.0 // All are WORD type
+#define B REG_INTRA_TEMP_3.0
+#define C REG_INTRA_TEMP_3.1
+#define YP REG_INTRA_TEMP_0 // Store intermediate results of c*(y-7). Make sure it's an even GRF
+#define YP1 REG_INTRA_TEMP_1 // Store intermediate results of c*(y-7). Make sure it's an odd GRF, used in {Comp}
+#define XP REG_INTRA_TEMP_5 // Store intermediate results of a+b*(x-7)+16. Make sure it's an odd GRF
+
+// First Calculate constants H and V
+// H1 = sum((-x'-1)*p[8+x',-1]), x'=0,1,...7
+// H2 = sum((-x'-1)*p[6-x',-1]), x'=7,6,...0
+// H = -H1 + H2
+// The same calculation holds for V
+//
+ mul (8) H1(0)<1> INTRA_REF_TOP(0,8)REGION(8,1) 0x89ABCDEF:v
+ mul (8) H2(0)<1> INTRA_REF_TOP(0,-1)REGION(8,1) 0xFEDCBA98:v
+
+ mul (8) V1(0)<1> INTRA_REF_LEFT(0,8*4)REGION(8,4) 0x89ABCDEF:v
+ mul (8) V2(0)<1> INTRA_REF_LEFT(0)REGION(8,4) 0x0FEDCBA9:v
+ mul (1) V2(0,7)<1> INTRA_REF_TOP(0,-1)<0;1,0> -8:w // Replace 0*p[-1,7] with -8*p[-1,-1]
+ // Now, REG_INTRA_TEMP_0 holds [H2, -H1] and REG_INTRA_TEMP_1 holds [V2, -V1]
+
+ // Sum up [H2, -H1] and [V2, -V1] using instruction compression
+ // ExecSize = 16 is restricted by B-spec for instruction compression
+ // Actual intermediate results are in lower sub-registers after each summing step
+ add (16) H1(0)<1> -H1(0) H2(0) {Compr} // Results in lower 8 WORDs
+ add (16) H1(0)<1> H1(0) H1(0,4) {Compr} // Results in lower 4 WORDs
+ add (16) H1(0)<1> H1(0) H1(0,2) {Compr} // Results in lower 2 WORDs
+ add (16) H1(0)<1> H1(0) H1(0,1) {Compr} // Results in lower 1 WORD
+
+// Calculate a, b, c and further derivations
+ mov (16) acc0<1>:w 32:w
+ mac (2) acc0<1>:w H1(0)<16;1,0> 5:w
+ shr (2) B<1>:w acc0:w 6:w // Done b,c
+ mov (16) acc0<1>:w 16:w
+ mac (16) acc0<1>:w INTRA_REF_TOP(0,15)<0;1,0> 16:w
+ mac (16) A<1>:w INTRA_REF_LEFT(0,15*4)<0;1,0> 16:w // A = a+16
+ mac (16) XP<1>:w B<0;1,0>:w XY_7<16;16,1>:b // XP = A+b*(x-7)
+ mul (8) YP<1>:w C<0;1,0>:w XY_7<16;8,2>:b // YP = c*(y-7), even portion
+ mul (8) YP1<1>:w C<0;1,0>:w XY_7_1<16;8,2>:b // YP = c*(y-7), odd portion
+
+// Finally the intra_16x16 plane prediction
+ $for(0,0; <16; 2,1) {
+ add (32) acc0<1>:w XP<16;16,1>:w YP.%2<16;16,0>:w {Compr} // Set Width!= 1 to trick EU to use YP_1.%2 for 2nd instruction
+ shr.sat (32) PRED_Y(%1)<2> acc0<16;16,1>:w 5:w {Compr}
+ }
+
+End_intra_Pred_16x16_Y:
+// End of intra_Pred_16x16_Y
diff --git a/shaders/h264/mc/loadRef_C_10x5.asm b/shaders/h264/mc/loadRef_C_10x5.asm new file mode 100644 index 0000000..3c0e851 --- /dev/null +++ b/shaders/h264/mc/loadRef_C_10x5.asm @@ -0,0 +1,57 @@ +/*
+ * Load reference 10x5 area for chroma NV12 4x4 MC
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: LoadRef_C_10x5.asm
+//
+// Load reference 10x5 area for chroma NV12 4x4 MC
+
+
+//#if !defined(__LOADREF_C_10x5__) // Make sure this is only included once
+//#define __LOADREF_C_10x5__
+
+
+#if 1
+
+ // Compute integer and fractional components of MV
+ asr (2) gMVX_INT<1>:w r[pMV,0]<2;2,1>:w 0x03:w {NoDDClr}
+ and (2) gMVX_FRACC<1>:w r[pMV,0]<2;2,1>:w 0x07:w {NoDDChk}
+
+ // Check whether MVY is integer
+ or.z.f0.0 (8) null:w gMVY_FRACC<0;1,0>:w 0:w
+
+ // Compute top-left corner position to be loaded
+ mov (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w
+ shl (1) gMSGSRC.0:d gMSGSRC.0:d 1:w
+
+ (f0.0) add (1) pMSGDSC:ud gMSGDSC_R:ud RESP_LEN(2)+nBI_LC_DIFF:ud
+ (-f0.0) add (1) pMSGDSC:ud gMSGDSC_R:ud RESP_LEN(3)+nBI_LC_DIFF:ud
+
+ // Read 16x5 pixels - TODO: Reading 12x5 instead of 16x5 took more time on CL. Why?
+ (f0.0) mov (1) gMSGSRC.2:ud 0x00030009:ud //{NoDDChk}
+ (-f0.0) mov (1) gMSGSRC.2:ud 0x00040009:ud //{NoDDChk}
+ send (8) gudREFC(0)<1> mMSGHDRC gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+
+#else
+
+ add (1) pMSGDSC:ud gMSGDSC_R:ud RESP_LEN(3)+nBI_LC_DIFF:ud
+
+ // Compute integer and fractional components of MV
+ asr (2) gMVX_INT<1>:w r[pMV,0]<2;2,1>:w 0x03:w {NoDDClr}
+ and (2) gMVX_FRACC<1>:w r[pMV,0]<2;2,1>:w 0x07:w {NoDDChk}
+
+ // Compute top-left corner position to be loaded
+ mov (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w
+ shl (1) gMSGSRC.0:d gMSGSRC.0:d 1:w
+
+ // Read 16x5 pixels
+ mov (1) gMSGSRC.2:ud 0x00040009:ud {NoDDChk}
+ send (8) gudREFC(0)<1> mMSGHDRC gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+#endif
+
+//#endif // !defined(__LOADREF_C_10x5__)
diff --git a/shaders/h264/mc/loadRef_C_6x3.asm b/shaders/h264/mc/loadRef_C_6x3.asm new file mode 100644 index 0000000..5ed7b69 --- /dev/null +++ b/shaders/h264/mc/loadRef_C_6x3.asm @@ -0,0 +1,38 @@ +/*
+ * Load reference 6x3 area for chroma NV12 4x4 MC
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: LoadRef_C_6x3.asm
+//
+// Load reference 6x3 area for chroma NV12 4x4 MC
+
+
+//#if !defined(__LOADREF_C_6x3__) // Make sure this is only included once
+//#define __LOADREF_C_6x3__
+
+
+#ifdef DEV_ILK
+ add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00100010:ud
+#else
+ add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00010010:ud
+#endif // DEV_ILK
+
+ // Compute integer and fractional components of MV
+ asr (2) gMVX_INT<1>:w r[pMV,0]<2;2,1>:w 0x03:w {NoDDClr}
+ and (2) gMVX_FRACC<1>:w r[pMV,0]<2;2,1>:w 0x07:w {NoDDChk}
+
+ // Compute top-left corner position to be loaded
+ mov (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w
+ shl (1) gMSGSRC.0:d gMSGSRC.0:d 1:w
+
+ // Read 8x3 pixels
+ mov (1) gMSGSRC.2:ud 0x00020005:ud
+ send (8) gudREFC(0)<1> mMSGHDRC gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+
+
+//#endif // !defined(__LOADREF_C_6x3__)
diff --git a/shaders/h264/mc/loadRef_Y_16x13.asm b/shaders/h264/mc/loadRef_Y_16x13.asm new file mode 100644 index 0000000..b233ea1 --- /dev/null +++ b/shaders/h264/mc/loadRef_Y_16x13.asm @@ -0,0 +1,127 @@ +/*
+ * Load reference 16x13 area for luma 8x8 MC
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: LoadRef_Y_16x13.asm
+//
+// Load reference 16x13 area for luma 8x8 MC
+
+
+//#if !defined(__LOADREF_Y_16x13__) // Make sure this is only included once
+//#define __LOADREF_Y_16x13__
+
+#if 1
+
+#if 1
+
+ // Check whether MVX is integer MV
+ and.z.f0.0 (1) null:w r[pMV,0]<0;1,0>:w 0x3:w
+
+ // Compute integer and fractional components of MV
+ and (2) gMVX_FRAC<1>:w r[pMV,0]<2;2,1>:w 0x03:w //{NoDDClr}
+ asr (2) gMVX_INT<1>:w r[pMV,0]<2;2,1>:w 0x02:w //{NoDDChk}
+
+ // Check whether MVY is integer
+ or.z.f0.1 (8) null:w gMVY_FRAC<0;1,0>:w 0:w
+
+ // If MVX is a multiple of 4 (..., -4, 0, 4, ...) integer MV, do special handling
+ (f0.0) jmpi (1) INTERLABEL(LOADREF_MVXZERO)
+
+ // Set message descriptor
+ (f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud RESP_LEN(4):ud
+ (-f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud RESP_LEN(7):ud
+
+ // Compute top-left corner position to be loaded
+ // TODO: sel
+ (-f0.1) add (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w -0x02:d //{NoDDClr}
+ (-f0.1) mov (1) gMSGSRC.2:ud 0x000c000c:ud //{NoDDChk}
+ (f0.1) add (1) gMSGSRC.0<1>:d gMVX_INT<0;1,0>:w -0x02:d //{NoDDClr}
+ (f0.1) mov (1) gMSGSRC.1<1>:d gMVY_INT<0;1,0>:w //{NoDDChk,NoDDClr}
+ (f0.1) mov (1) gMSGSRC.2:ud 0x0007000c:ud //{NoDDChk}
+
+ // Read 16x13 pixels
+ send (8) gudREF(0)<1> mMSGHDRY gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+
+ jmpi INTERLABEL(EXIT_LOADREF_Y_16x13)
+
+INTERLABEL(LOADREF_MVXZERO):
+
+ // Set message descriptor
+#ifdef DEV_ILK
+ (f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00200000:ud
+ (-f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00400000:ud
+#else
+ (f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00020000:ud
+ (-f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00040000:ud
+#endif
+
+ // Compute top-left corner position to be loaded
+ // TODO: sel
+ mov (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w
+ (-f0.1) add (1) gMSGSRC.1<1>:d gMVY_INT<0;1,0>:w -0x02:d
+ (-f0.1) mov (1) gMSGSRC.2:ud 0x000c0007:ud //{NoDDChk}
+ (f0.1) mov (1) gMSGSRC.2:ud 0x00070007:ud //{NoDDChk}
+
+ // Read 16x13 pixels
+ send (8) gudREF(0)<1> mMSGHDRY gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+
+
+#else
+ // Compute integer and fractional components of MV
+ and (2) gMVX_FRAC<1>:w r[pMV,0]<2;2,1>:w 0x03:w //{NoDDClr}
+ asr (2) gMVX_INT<1>:w r[pMV,0]<2;2,1>:w 0x02:w //{NoDDChk}
+
+ // Check whether MVY is integer
+ or.z.f0.1 (8) null:w gMVY_FRAC<0;1,0>:w 0:w
+
+ // Set message descriptor
+#ifdef DEV_ILK
+ (f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00400000:ud
+ (-f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00700000:ud
+#else
+ (f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00040000:ud
+ (-f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00070000:ud
+#endif
+
+ // Compute top-left corner position to be loaded
+ // TODO: sel
+ (-f0.1) add (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w -0x02:d //{NoDDClr}
+ (-f0.1) mov (1) gMSGSRC.2:ud 0x000c000c:ud //{NoDDChk}
+ (f0.1) add (1) gMSGSRC.0<1>:d gMVX_INT<0;1,0>:w -0x02:d //{NoDDClr}
+ (f0.1) mov (1) gMSGSRC.1<1>:d gMVY_INT<0;1,0>:w //{NoDDChk,NoDDClr}
+ (f0.1) mov (1) gMSGSRC.2:ud 0x0007000c:ud //{NoDDChk}
+
+ // Read 16x13 pixels
+ send (8) gudREF(0)<1> mMSGHDRY gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+#endif
+
+#else
+
+ // Compute integer and fractional components of MV
+ and (2) gMVX_FRAC<1>:w r[pMV,0]<2;2,1>:w 0x03:w {NoDDClr} //
+ asr (2) gMVX_INT<1>:w r[pMV,0]<2;2,1>:w 0x02:w {NoDDChk} //
+
+ // Set message descriptor
+#ifdef DEV_ILK
+ add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00700000:ud
+#else
+ add (1) pMSGDSC:ud gMSGDSC_R:ud 0x00070000:ud
+#endif // DEV_ILK
+
+ // Compute top-left corner position to be loaded
+ add (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w -0x02:d {NoDDClr} //
+ mov (1) gMSGSRC.2:ud 0x000c000c:ud {NoDDChk} //
+
+ // Read 16x13 pixels
+ send (8) gudREF(0)<1> mMSGHDRY gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+
+#endif
+
+INTERLABEL(EXIT_LOADREF_Y_16x13):
+
+//#endif // !defined(__LOADREF_Y_16x13__)
diff --git a/shaders/h264/mc/loadRef_Y_16x9.asm b/shaders/h264/mc/loadRef_Y_16x9.asm new file mode 100644 index 0000000..e48151e --- /dev/null +++ b/shaders/h264/mc/loadRef_Y_16x9.asm @@ -0,0 +1,61 @@ +/*
+ * Load reference 16x9 area for luma 4x4 MC
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: LoadRef_Y_16x9.asm
+//
+// Load reference 16x9 area for luma 4x4 MC
+
+
+//#if !defined(__LOADREF_Y_16x9__) // Make sure this is only included once
+//#define __LOADREF_Y_16x9__
+
+#if 1
+
+ // Compute integer and fractional components of MV
+ and (2) gMVX_FRAC<1>:w r[pMV,0]<2;2,1>:w 0x03:w //{NoDDClr}
+ asr (2) gMVX_INT<1>:w r[pMV,0]<2;2,1>:w 0x02:w //{NoDDChk}
+
+ // Check whether MVY is integer
+ or.z.f0.1 (8) null:w gMVY_FRAC<0;1,0>:w 0:w
+
+ // Set message descriptor
+ (f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud RESP_LEN(2):ud
+ (-f0.1) add (1) pMSGDSC:ud gMSGDSC_R:ud RESP_LEN(5):ud
+
+ // Compute top-left corner position to be loaded
+ // TODO: sel
+ (-f0.1) add (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w -0x02:d //{NoDDClr}
+ (-f0.1) mov (1) gMSGSRC.2:ud 0x00080008:ud //{NoDDChk}
+ (f0.1) add (1) gMSGSRC.0<1>:d gMVX_INT<0;1,0>:w -0x02:d //{NoDDClr}
+ (f0.1) mov (1) gMSGSRC.1<1>:d gMVY_INT<0;1,0>:w //{NoDDChk,NoDDClr}
+ (f0.1) mov (1) gMSGSRC.2:ud 0x00030008:ud //{NoDDChk}
+
+ // Read 16x9 pixels
+ send (8) gudREF(0)<1> mMSGHDRY gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+
+#else
+
+ // Compute integer and fractional components of MV
+ and (2) gMVX_FRAC<1>:w r[pMV,0]<2;2,1>:w 0x03:w {NoDDClr} //
+ asr (2) gMVX_INT<1>:w r[pMV,0]<2;2,1>:w 0x02:w {NoDDChk} //
+
+ // Set message descriptor
+ add (1) pMSGDSC:ud gMSGDSC_R:ud RESP_LEN(5):ud
+
+ // Compute top-left corner position to be loaded
+ add (2) gMSGSRC.0<1>:d gMVX_INT<2;2,1>:w -0x02:d {NoDDClr} //
+ mov (1) gMSGSRC.2:ud 0x00080008:ud {NoDDChk} //
+
+ // Read 16x9 pixels
+ send (8) gudREF(0)<1> mMSGHDRY gMSGSRC<8;8,1>:ud DAPREAD pMSGDSC:ud
+
+#endif
+
+
+//#endif // !defined(__LOADREF_Y_16x9__)
diff --git a/shaders/h264/mc/load_Intra_Ref_UV.asm b/shaders/h264/mc/load_Intra_Ref_UV.asm new file mode 100644 index 0000000..34adbe6 --- /dev/null +++ b/shaders/h264/mc/load_Intra_Ref_UV.asm @@ -0,0 +1,44 @@ +/*
+ * Load all reference U/V samples from neighboring macroblocks
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__LOAD_INTRA_REF_UV__) // Make sure this is only included once
+#define __LOAD_INTRA_REF_UV__
+
+// Module name: load_Intra_Ref_UV.asm
+//
+// Load all reference U/V samples from neighboring macroblocks
+//
+// Note: Since loading of U/V data always follows writing of Y, the message descriptor is manipulated
+// to avoid recalculating due to frame/field variztions.
+
+// First load top 20x1 row U/V reference samples
+// 4 from macroblock D (actually use 2), 16 from macroblock B
+//
+ shr (1) I_ORIY<1>:w I_ORIY<0;1,0>:w 1:w // Adjust I_ORIY for NV12 format
+ add (2) MSGSRC.0<1>:d I_ORIX<2;2,1>:w TOP_REF_OFFSET<2;2,1>:b {NoDDClr} // Reference samples positioned at (-4, -1)
+ mov (1) MSGSRC.2:ud 0x00000013:ud {NoDDChk} // Block width and height (20x1)
+
+// Update message descriptor based on previous Y block write
+//
+#ifdef DEV_ILK
+ add (1) MSGDSC MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC+DESTUV-DWBWMSGDSC_WC-0x10000000-DESTY:ud // Set message descriptor
+#else
+ add (1) MSGDSC MSGDSC RESP_LEN(1)+DWBRMSGDSC_RC+DESTUV-DWBWMSGDSC_WC-0x00800000-DESTY:ud // Set message descriptor
+#endif // DEV_ILK
+
+ send (8) INTRA_REF_TOP_D(0) MSGHDR MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// Then load left 4x8 reference samples (actually use 1x8 column)
+//
+ add (1) MSGSRC.1<1>:d MSGSRC.1<0;1,0>:d 1:w {NoDDClr} // Reference samples positioned next row
+ mov (1) MSGSRC.2:ud 0x00070003:ud {NoDDChk} // Block width and height (4x8)
+ send (8) INTRA_REF_LEFT_D(0) MSGHDRUV MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// End of load_Intra_Ref_UV
+#endif // !defined(__LOAD_INTRA_REF_UV__)
diff --git a/shaders/h264/mc/load_Intra_Ref_Y.asm b/shaders/h264/mc/load_Intra_Ref_Y.asm new file mode 100644 index 0000000..de8ec49 --- /dev/null +++ b/shaders/h264/mc/load_Intra_Ref_Y.asm @@ -0,0 +1,37 @@ +/*
+ * Load all reference Y samples from neighboring macroblocks
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__LOAD_INTRA_REF_Y__) // Make sure this is only included once
+#define __LOAD_INTRA_REF_Y__
+
+// Module name: load_Intra_Ref_Y.asm
+//
+// Load all reference Y samples from neighboring macroblocks
+//
+load_Intra_Ref_Y:
+// shl (2) I_ORIX<1>:uw ORIX<2;2,1>:ub 4:w // Convert MB origin to pixel unit
+
+// First load top 28x1 row reference samples
+// 4 from macroblock D (actually use 1), 16 from macroblock B, and 8 from macroblock C
+//
+ add (2) MSGSRC.0<1>:d I_ORIX<2;2,1>:w TOP_REF_OFFSET<2;2,1>:b {NoDDClr} // Reference samples positioned at (-4, -1)
+ mov (1) MSGSRC.2:ud 0x0000001B:ud {NoDDChk} // Block width and height (28x1)
+ add (1) MSGDSC REG_MBAFF_FIELD<0;1,0>:uw RESP_LEN(1)+DWBRMSGDSC_RC+DESTY:ud // Set message descriptor
+ send (8) INTRA_REF_TOP_D(0) MSGHDRY0 MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+// Then load left 4x16 reference samples (actually use 1x16 column)
+//
+ add (1) MSGSRC.1<1>:d MSGSRC.1<0;1,0>:d 1:w {NoDDClr} // Reference samples positioned next row
+ mov (1) MSGSRC.2:ud 0x00F0003:ud {NoDDChk} // Block width and height (4x16)
+ add (1) MSGDSC MSGDSC RESP_LEN(1):ud // Need to read 1 more GRF register
+ send (8) INTRA_REF_LEFT_D(0) MSGHDRY1 MSGSRC<8;8,1>:ud DAPREAD MSGDSC
+
+ RETURN
+// End of load_Intra_Ref_Y
+#endif // !defined(__LOAD_INTRA_REF_Y__)
diff --git a/shaders/h264/mc/null.g4a b/shaders/h264/mc/null.g4a new file mode 100644 index 0000000..f138029 --- /dev/null +++ b/shaders/h264/mc/null.g4a @@ -0,0 +1,43 @@ +/* + * Copyright © 2009 Intel Corporation + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + * + * Author: + * Zou Nan hai <nanhai.zou@intel.com> + * Xiang Haihao <haihao.xiang@intel.com> + * + */ +define(`YUV_color',`0xFFFFFFFFUD') +shl(2) g62.0<1>UD g3.4<2,2,1>UB 4UW {align1}; +mov(1) g62.8<1>UD 0x000f000fUD {align1}; +mov(16) m1<1>UD YUV_color {align1 compr}; +mov(16) m3<1>UD YUV_color {align1 compr}; +mov(16) m5<1>UD YUV_color {align1 compr}; +mov(16) m7<1>UD YUV_color {align1 compr}; +send(16) 0 acc0<1>UW g62<8,8,1>UW write(0, 0, 2, 0) mlen 9 rlen 0 {align1}; + +shr(1) g62.4<1>UD g62.4<1,1,1>UD 1UW {align1}; +mov(1) g62.8<1>UD 0x0007000fUD {align1}; +mov(16) m1<1>UD YUV_color {align1 compr}; +mov(16) m3<1>UD YUV_color {align1 compr}; +send(16) 0 acc0<1>UW g62<8,8,1>UW write(1, 0, 2, 0) mlen 5 rlen 0 {align1}; + +send(16) 0 acc0<1>UW g0<8,8,1>UW thread_spawner(0, 0, 0) mlen 1 rlen 0 {align1 EOT}; diff --git a/shaders/h264/mc/null.g4b b/shaders/h264/mc/null.g4b new file mode 100644 index 0000000..fdd18e6 --- /dev/null +++ b/shaders/h264/mc/null.g4b @@ -0,0 +1,13 @@ + { 0x00200009, 0x27c02e21, 0x00450064, 0x00040004 }, + { 0x00000001, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00802001, 0x20200062, 0x00000000, 0xffffffff }, + { 0x00802001, 0x20600062, 0x00000000, 0xffffffff }, + { 0x00802001, 0x20a00062, 0x00000000, 0xffffffff }, + { 0x00802001, 0x20e00062, 0x00000000, 0xffffffff }, + { 0x00800031, 0x24001d28, 0x008d07c0, 0x05902000 }, + { 0x00000008, 0x27c42c21, 0x002107c4, 0x00010001 }, + { 0x00000001, 0x27c80061, 0x00000000, 0x0007000f }, + { 0x00802001, 0x20200062, 0x00000000, 0xffffffff }, + { 0x00802001, 0x20600062, 0x00000000, 0xffffffff }, + { 0x00800031, 0x24001d28, 0x008d07c0, 0x05502001 }, + { 0x00800031, 0x24001d28, 0x008d0000, 0x87100000 }, diff --git a/shaders/h264/mc/null.g4b.gen5 b/shaders/h264/mc/null.g4b.gen5 new file mode 100644 index 0000000..7ecb90d --- /dev/null +++ b/shaders/h264/mc/null.g4b.gen5 @@ -0,0 +1,13 @@ + { 0x00200009, 0x27c02e21, 0x00450064, 0x00040004 }, + { 0x00000001, 0x27c80061, 0x00000000, 0x000f000f }, + { 0x00802001, 0x20200062, 0x00000000, 0xffffffff }, + { 0x00802001, 0x20600062, 0x00000000, 0xffffffff }, + { 0x00802001, 0x20a00062, 0x00000000, 0xffffffff }, + { 0x00802001, 0x20e00062, 0x00000000, 0xffffffff }, + { 0x00800031, 0x24001d28, 0x508d07c0, 0x12082000 }, + { 0x00000008, 0x27c42c21, 0x002107c4, 0x00010001 }, + { 0x00000001, 0x27c80061, 0x00000000, 0x0007000f }, + { 0x00802001, 0x20200062, 0x00000000, 0xffffffff }, + { 0x00802001, 0x20600062, 0x00000000, 0xffffffff }, + { 0x00800031, 0x24001d28, 0x508d07c0, 0x0a082001 }, + { 0x00800031, 0x24001d28, 0x748d0000, 0x82000000 }, diff --git a/shaders/h264/mc/recon_C_4x4.asm b/shaders/h264/mc/recon_C_4x4.asm new file mode 100644 index 0000000..3a2a921 --- /dev/null +++ b/shaders/h264/mc/recon_C_4x4.asm @@ -0,0 +1,37 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Recon_C_4x4.asm
+//
+// $Revision: 11 $
+// $Date: 10/03/06 5:28p $
+//
+
+
+//#if !defined(__RECON_C_4x4__) // Make sure this is only included once
+//#define __RECON_C_4x4__
+
+
+ // TODO: Use instruction compression
+ add.sat (4) r[pERRORC,0]<2>:ub r[pERRORC,0]<4;4,1>:w gubCPRED(0)<16;4,4>
+ add.sat (4) r[pERRORC,128]<2>:ub r[pERRORC,128]<4;4,1>:w gubCPRED(0,2)<16;4,4>
+ add.sat (4) r[pERRORC,32]<2>:ub r[pERRORC,32]<4;4,1>:w gubCPRED(1)<16;4,4>
+ add.sat (4) r[pERRORC,128+32]<2>:ub r[pERRORC,128+32]<4;4,1>:w gubCPRED(1,2)<16;4,4>
+
+ add.sat (4) r[pERRORC,16]<2>:ub r[pERRORC,16]<4;4,1>:w gubCPRED(0,16)<16;4,4>
+ add.sat (4) r[pERRORC,128+16]<2>:ub r[pERRORC,128+16]<4;4,1>:w gubCPRED(0,18)<16;4,4>
+ add.sat (4) r[pERRORC,48]<2>:ub r[pERRORC,48]<4;4,1>:w gubCPRED(1,16)<16;4,4>
+ add.sat (4) r[pERRORC,128+48]<2>:ub r[pERRORC,128+48]<4;4,1>:w gubCPRED(1,18)<16;4,4>
+
+ // Increase chroma error block offset
+#ifndef MONO
+ add (1) pERRORC:w pERRORC:w 8:w
+#endif
+
+
+//#endif // !defined(__RECON_C_4x4__)
diff --git a/shaders/h264/mc/recon_Y_8x8.asm b/shaders/h264/mc/recon_Y_8x8.asm new file mode 100644 index 0000000..6017790 --- /dev/null +++ b/shaders/h264/mc/recon_Y_8x8.asm @@ -0,0 +1,27 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: Recon_Y_8x8.asm
+//
+// $Revision: 10 $
+// $Date: 9/22/06 2:50p $
+//
+
+
+//#if !defined(__RECON_Y_8x8__) // Make sure this is only included once
+//#define __RECON_Y_8x8__
+
+
+ add.sat (16) r[pERRORY,0]<2>:ub r[pERRORY,0]<16;16,1>:w gubYPRED(0)
+ add.sat (16) r[pERRORY,nGRFWIB]<2>:ub r[pERRORY,nGRFWIB]<16;16,1>:w gubYPRED(1)
+ add.sat (16) r[pERRORY,nGRFWIB*2]<2>:ub r[pERRORY,nGRFWIB*2]<16;16,1>:w gubYPRED(2)
+ add.sat (16) r[pERRORY,nGRFWIB*3]<2>:ub r[pERRORY,nGRFWIB*3]<16;16,1>:w gubYPRED(3)
+
+ add (1) pERRORY:w pERRORY:w 128:w
+
+//#endif // !defined(__RECON_Y_8x8__)
diff --git a/shaders/h264/mc/roundShift_C_4x4.asm b/shaders/h264/mc/roundShift_C_4x4.asm new file mode 100644 index 0000000..c609159 --- /dev/null +++ b/shaders/h264/mc/roundShift_C_4x4.asm @@ -0,0 +1,26 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: RoundShift_C_4x4.asm
+//
+// Do (...+32)>>6 to 4x4 (NV12 8x4) interpolated chrominance data
+//
+
+
+//#if !defined(__RoundShift_C_4x4__) // Make sure this is only included once
+//#define __RoundShift_C_4x4__
+
+
+ // TODO: Optimize using instruction compression
+ add (16) acc0<1>:w r[pRESULT,0]<16;16,1>:w 32:w
+ add (16) acc1<1>:w r[pRESULT,nGRFWIB]<16;16,1>:w 32:w
+ asr.sat (16) r[pRESULT,0]<2>:ub acc0<16;16,1>:w 6:w
+ asr.sat (16) r[pRESULT,nGRFWIB]<2>:ub acc1<16;16,1>:w 6:w
+
+
+//#endif // !defined(__RoundShift_C_4x4__)
diff --git a/shaders/h264/mc/save_16x16_Y.asm b/shaders/h264/mc/save_16x16_Y.asm new file mode 100644 index 0000000..713e12c --- /dev/null +++ b/shaders/h264/mc/save_16x16_Y.asm @@ -0,0 +1,42 @@ +/*
+ * Save decoded Y picture data to frame buffer
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__SAVE_16X16_Y__) // Make sure this is only included once
+#define __SAVE_16X16_Y__
+
+// Module name: save_16x16_Y.asm
+//
+// Save decoded Y picture data to frame buffer
+//
+
+save_16x16_Y:
+
+ mov (1) MSGSRC.2:ud 0x000F000F:ud {NoDDClr} // Block width and height (16x16)
+ mov (2) MSGSRC.0:ud I_ORIX<2;2,1>:w {NoDDChk} // X, Y offset
+#ifdef DEV_ILK
+ add (1) MSGDSC MSGDSC MSG_LEN(8)+DWBWMSGDSC_WC-DWBRMSGDSC_RC-0x00200000:ud // Set message descriptor
+#else
+ add (1) MSGDSC MSGDSC MSG_LEN(8)+DWBWMSGDSC_WC-DWBRMSGDSC_RC-0x00020000:ud // Set message descriptor
+#endif // DEV_ILK
+
+ mov (1) PDECBUF_UD<1>:ud 0x10001*DECBUF*GRFWIB+0x00400000:ud // Pointers to row 0 and 2 of decoded data
+
+ $for(0,0; <8; 2,4) {
+ mov (32) MSGPAYLOAD(%1)<1> r[PDECBUF, %2*GRFWIB]REGION(16,2) {Compr} // Block Y0/Y2
+ mov (32) MSGPAYLOAD(%1,16)<1> r[PDECBUF, (1+%2)*GRFWIB]REGION(16,2) {Compr} // Block Y1/Y3
+ }
+
+// Update message descriptor based on previous read setup
+//
+ send (8) REG_WRITE_COMMIT_Y<1>:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+ RETURN
+// End of save_16x16_Y
+
+#endif // !defined(__SAVE_16X16_Y__)
diff --git a/shaders/h264/mc/save_4x4_Y.asm b/shaders/h264/mc/save_4x4_Y.asm new file mode 100644 index 0000000..415034a --- /dev/null +++ b/shaders/h264/mc/save_4x4_Y.asm @@ -0,0 +1,43 @@ +/*
+ * Save Intra_4x4 decoded Y picture data to frame buffer
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__SAVE_4X4_Y__) // Make sure this is only included once
+#define __SAVE_4X4_Y__
+
+// Module name: save_4x4_Y.asm
+//
+// Save Intra_4x4 decoded Y picture data to frame buffer
+// Note: Each 4x4 block is stored in 1 GRF register in the order of block raster scan order,
+// i.e. 0, 1, 4, 5, 2, 3, 6, 7, 8, 9, 12, 13, 10, 11, 14, 15
+
+save_4x4_Y:
+
+ mov (1) MSGSRC.2:ud 0x000F000F:ud {NoDDClr} // Block width and height (16x16)
+ mov (2) MSGSRC.0:ud I_ORIX<2;2,1>:w {NoDDChk} // X, Y offset
+#ifdef DEV_ILK
+ add (1) MSGDSC MSGDSC MSG_LEN(8)+DWBWMSGDSC_WC-DWBRMSGDSC_RC-0x00200000:ud // Set message descriptor
+#else
+ add (1) MSGDSC MSGDSC MSG_LEN(8)+DWBWMSGDSC_WC-DWBRMSGDSC_RC-0x00020000:ud // Set message descriptor
+#endif // DEV_ILK
+
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOAD(%1)<1> DEC_Y(%1)<16;4,1>
+ mov (16) MSGPAYLOAD(%1,16)<1> DEC_Y(%1,4)<16;4,1>
+ mov (16) MSGPAYLOAD(%1+1)<1> DEC_Y(%1,8)<16;4,1>
+ mov (16) MSGPAYLOAD(%1+1,16)<1> DEC_Y(%1,12)<16;4,1>
+ }
+
+// Update message descriptor based on previous read setup
+//
+ send (8) REG_WRITE_COMMIT_Y<1>:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+ RETURN
+// End of save_4x4_Y
+
+#endif // !defined(__SAVE_4X4_Y__)
diff --git a/shaders/h264/mc/save_8x8_UV.asm b/shaders/h264/mc/save_8x8_UV.asm new file mode 100644 index 0000000..aa76af9 --- /dev/null +++ b/shaders/h264/mc/save_8x8_UV.asm @@ -0,0 +1,51 @@ +/*
+ * Save decoded U/V picture data to frame buffer
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__SAVE_8x8_UV__) // Make sure this is only included once
+#define __SAVE_8x8_UV__
+
+// Module name: save_8x8_UV.asm
+//
+// Save decoded U/V picture data to frame buffer
+//
+
+ mov (1) MSGSRC.2:ud 0x0007000F:ud {NoDDClr} // Block width and height (16x8)
+ mov (2) MSGSRC.0<1>:ud I_ORIX<2;2,1>:w {NoDDChk} // I_ORIX has already been adjusted for NV12
+
+// Update message descriptor based on previous read setup
+//
+#ifdef DEV_ILK
+ add (1) MSGDSC MSGDSC MSG_LEN(4)+DWBWMSGDSC_WC-DWBRMSGDSC_RC-0x00100000:ud // Set message descriptor
+#else
+ add (1) MSGDSC MSGDSC MSG_LEN(4)+DWBWMSGDSC_WC-DWBRMSGDSC_RC-0x00010000:ud // Set message descriptor
+#endif // DEV_ILK
+
+// Write U/V picture data
+//
+#ifndef MONO
+ mov MSGPAYLOAD(0,0)<1> DEC_UV(0)REGION(16,2) // U/V row 0
+ mov MSGPAYLOAD(0,16)<1> DEC_UV(1)REGION(16,2) // U/V row 1
+ mov MSGPAYLOAD(1,0)<1> DEC_UV(2)REGION(16,2) // U/V row 2
+ mov MSGPAYLOAD(1,16)<1> DEC_UV(3)REGION(16,2) // U/V row 3
+ mov MSGPAYLOAD(2,0)<1> DEC_UV(4)REGION(16,2) // U/V row 4
+ mov MSGPAYLOAD(2,16)<1> DEC_UV(5)REGION(16,2) // U/V row 5
+ mov MSGPAYLOAD(3,0)<1> DEC_UV(6)REGION(16,2) // U/V row 6
+ mov MSGPAYLOAD(3,16)<1> DEC_UV(7)REGION(16,2) // U/V row 7
+#else // defined(MONO)
+ $for(0; <4; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> 0x80808080:ud {Compr}
+ }
+
+#endif // !defined(MONO)
+
+ send (8) REG_WRITE_COMMIT_UV<1>:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+// End of save_8x8_UV
+
+#endif // !defined(__SAVE_8x8_UV__)
diff --git a/shaders/h264/mc/save_8x8_Y.asm b/shaders/h264/mc/save_8x8_Y.asm new file mode 100644 index 0000000..3ffca79 --- /dev/null +++ b/shaders/h264/mc/save_8x8_Y.asm @@ -0,0 +1,56 @@ +/*
+ * Save Intra_8x8 decoded Y picture data to frame buffer
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__SAVE_8X8_Y__) // Make sure this is only included once
+#define __SAVE_8X8_Y__
+
+// Module name: save_8x8_Y.asm
+//
+// Save Intra_8x8 decoded Y picture data to frame buffer
+// NotE: Every 4 rows of Y data are interleaved with the horizontal neighboring blcok
+//
+save_8x8_Y:
+
+ mov (1) MSGSRC.2:ud 0x000F000F:ud {NoDDClr} // Block width and height (16x16)
+ mov (2) MSGSRC.0:ud I_ORIX<2;2,1>:w {NoDDChk} // X, Y offset
+
+// Update message descriptor based on previous read setup
+//
+#ifdef DEV_ILK
+ add (1) MSGDSC MSGDSC MSG_LEN(8)+DWBWMSGDSC_WC-DWBRMSGDSC_RC-0x00200000:ud // Set message descriptor
+#else
+ add (1) MSGDSC MSGDSC MSG_LEN(8)+DWBWMSGDSC_WC-DWBRMSGDSC_RC-0x00020000:ud // Set message descriptor
+#endif // DEV_ILK
+
+ mov (16) MSGPAYLOAD(0)<1> DEC_Y(0)<32;8,1>
+ mov (16) MSGPAYLOAD(0,16)<1> DEC_Y(0,8)<32;8,1>
+ mov (16) MSGPAYLOAD(1,0)<1> DEC_Y(0,16)<32;8,1>
+ mov (16) MSGPAYLOAD(1,16)<1> DEC_Y(0,24)<32;8,1>
+
+ mov (16) MSGPAYLOAD(2)<1> DEC_Y(2)<32;8,1>
+ mov (16) MSGPAYLOAD(2,16)<1> DEC_Y(2,8)<32;8,1>
+ mov (16) MSGPAYLOAD(3,0)<1> DEC_Y(2,16)<32;8,1>
+ mov (16) MSGPAYLOAD(3,16)<1> DEC_Y(2,24)<32;8,1>
+
+ mov (16) MSGPAYLOAD(4)<1> DEC_Y(4)<32;8,1>
+ mov (16) MSGPAYLOAD(4,16)<1> DEC_Y(4,8)<32;8,1>
+ mov (16) MSGPAYLOAD(5,0)<1> DEC_Y(4,16)<32;8,1>
+ mov (16) MSGPAYLOAD(5,16)<1> DEC_Y(4,24)<32;8,1>
+
+ mov (16) MSGPAYLOAD(6)<1> DEC_Y(6)<32;8,1>
+ mov (16) MSGPAYLOAD(6,16)<1> DEC_Y(6,8)<32;8,1>
+ mov (16) MSGPAYLOAD(7,0)<1> DEC_Y(6,16)<32;8,1>
+ mov (16) MSGPAYLOAD(7,16)<1> DEC_Y(6,24)<32;8,1>
+
+ send (8) REG_WRITE_COMMIT_Y<1>:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+ RETURN
+// End of save_8x8_Y
+
+#endif // !defined(__SAVE_8X8_Y__)
diff --git a/shaders/h264/mc/save_I_PCM.asm b/shaders/h264/mc/save_I_PCM.asm new file mode 100644 index 0000000..77be35e --- /dev/null +++ b/shaders/h264/mc/save_I_PCM.asm @@ -0,0 +1,56 @@ +/*
+ * Save I_PCM Y samples to Y picture buffer
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: save_I_PCM.asm
+//
+// First save I_PCM Y samples to Y picture buffer
+//
+ mov (1) MSGSRC.2:ud 0x000F000F:ud {NoDDClr} // Block width and height (16x16)
+ shl (2) MSGSRC.0:ud ORIX<2;2,1>:ub 4:w {NoDDChk} // Convert MB origin in pixel unit
+
+ add (1) MSGDSC REG_MBAFF_FIELD<0;1,0>:uw MSG_LEN(8)+DWBWMSGDSC_WC+DESTY:ud // Set message descriptor
+
+ $for(0; <8; 2) {
+ mov (32) MSGPAYLOAD(%1)<1> I_PCM_Y(%1)REGION(16,1) {Compr,NoDDClr}
+ mov (32) MSGPAYLOAD(%1,16)<1> I_PCM_Y(%1,16)REGION(16,1) {Compr,NoDDChk}
+ }
+
+ send (8) REG_WRITE_COMMIT_Y<1>:ud MSGHDR MSGSRC<8;8,1>:ud DAPWRITE MSGDSC
+
+// Then save I_PCM U/V samples to U/V picture buffer
+//
+ mov (1) MSGHDR.2:ud 0x0007000F:ud {NoDDClr} // Block width and height (16x8)
+ asr (1) MSGHDR.1:ud MSGSRC.1<0;1,0>:ud 1:w {NoDDChk} // Y offset should be halved
+ add (1) MSGDSC MSGDSC 0x0-MSG_LEN(4)+0x1:d // Set message descriptor for U/V
+
+#if 0
+ and.z.f0.0 (1) NULLREG REG_CHROMA_FORMAT_IDC CHROMA_FORMAT_IDC:ud
+ (f0.0) jmpi (1) MONOCHROME_I_PCM
+#endif
+
+#ifndef MONO
+// Non-monochrome picture
+//
+ $for(0,0; <4; 2,1) {
+ mov (16) MSGPAYLOAD(%1)<2> I_PCM_UV(%2)REGION(16,1) // U data
+ mov (16) MSGPAYLOAD(%1,1)<2> I_PCM_UV(%2+2)REGION(16,1) // V data
+ mov (16) MSGPAYLOAD(%1+1)<2> I_PCM_UV(%2,16)REGION(16,1) // U data
+ mov (16) MSGPAYLOAD(%1+1,1)<2> I_PCM_UV(%2+2,16)REGION(16,1) // V data
+ }
+#else // defined(MONO)
+MONOCHROME_I_PCM:
+ $for(0; <4; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> 0x80808080:ud {Compr}
+ }
+
+#endif // !defined(MONO)
+
+ send (8) REG_WRITE_COMMIT_UV<1>:ud MSGHDR null:ud DAPWRITE MSGDSC
+
+// End of save_I_PCM
diff --git a/shaders/h264/mc/scoreboard.asm b/shaders/h264/mc/scoreboard.asm new file mode 100644 index 0000000..6fb41cf --- /dev/null +++ b/shaders/h264/mc/scoreboard.asm @@ -0,0 +1,282 @@ +/*
+ * Dependency control scoreboard kernel
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: scoreboard.asm
+//
+// Dependency control scoreboard kernel
+//
+// $Revision: 16 $
+// $Date: 10/18/06 4:10p $
+//
+
+// ----------------------------------------------------
+// Main: scoreboard
+// ----------------------------------------------------
+// ----------------------------------------------------
+// Scoreboard structure
+// ----------------------------------------------------
+//
+// 1 DWORD per thread
+//
+// Bit 31: "Checking" thread, i.e. an intra MB that sends "check dependency" message
+// Bit 30: "Completed" thread. This bit set by an "update" message from intra/inter MB.
+// Bits 29:28: Must set to 0
+// Bits 27:24: EUID
+// Bits 23:18: Reserved
+// Bits 17:16: TID
+// Bits 15:8: X offset of current MB
+// Bits 15:5: Reserved
+// Bits 4:0: 5 bits of available neighbor MB flags
+
+.kernel scoreboard
+SCOREBOARD:
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+mov (1) acc0:ud 0xf0aa55a5:ud
+#endif
+
+#include "header.inc"
+#include "scoreboard_header.inc"
+
+//
+// Now, begin source code....
+//
+
+.code
+
+#ifdef AS_ENABLED
+ and.z.f0.1 (1) NULLREG r0.2<0;1,0>:ud TH_RES // Is this a restarted thread previously interrupted?
+ (f0.1) jmpi (1) Scoreboard_Init
+
+ #include "scoreboard_restore_AS.asm"
+
+ jmpi (1) Scoreboard_OpenGW
+Scoreboard_Init:
+#endif // End AS_ENABLED
+
+// Scoreboard must be initialized to 0xc000ffff, meaning all "completed"
+// And it also avoids message mis-handling for the first MB
+ $for(0; <32; 2) {
+ mov (16) CMD_SB(%1)<1> 0xc000ffff:ud {Compr}
+ }
+#ifdef DOUBLE_SB // Scoreboard size needs to be doubled
+ $for(32; <64; 2) {
+ mov (16) CMD_SB(%1)<1> 0xc000ffff:ud {Compr}
+ }
+#endif // DOUBLE_SB
+
+//----------------------------------------------------------
+// Open message gateway for the scoreboard thread
+//
+// RegBase = r4 (0x04)
+// Gateway Size = 64 GRF registers (0x6)
+// Dispatch ID = r0.20:ub
+// Scoreboard Thread Key = 0
+//----------------------------------------------------------
+Scoreboard_OpenGW:
+ mov (8) MSGHDRY0<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+ // Send a message with register base RegBase=0x04(r4) and Gateway size = 0x6 = 64 GRF reg and Key = 0
+ // 000 00000100 00000 00000 110 00000000 ==> 0000 0000 1000 0000 0000 0110 0000 0000
+#ifdef AS_ENABLED
+ add (1) MSGHDRY0.5<1>:ud r0.20:ub 0x00800700:ud // Allocate 128 GRFs for message gateway - for SIP to send notification MSG
+#else
+ #ifdef DOUBLE_SB
+ add (1) MSGHDRY0.5<1>:ud r0.20:ub 0x00800600:ud // 64 GRF's for CTG-B
+ #else
+ add (1) MSGHDRY0.5<1>:ud r0.20:ub 0x00800500:ud // 32 GRF's for CTG-A
+ #endif // DOUBLE_SB
+#endif
+ send (8) NULLREG MSGHDRY0 null:ud MSG_GW OGWMSGDSC
+
+//------------------------------------------------------------------------
+// Send Thread Spawning Message to start dispatching macroblock threads
+//
+//------------------------------------------------------------------------
+#ifdef AS_ENABLED
+ mov (8) acc0<1>:ud CMD_SB(31)<8;8,1> // Ensure scoreboard data have been completely restored
+#endif // End AS_ENABLED
+ mov (8) MSGHDRY1<1>:ud r0<8;8,1>:ud // Initialize message header payload with R0
+ mov (1) MSGHDRY1.4<1>:ud 0x00000400:ud // Dispatch URB length = 1
+
+ send (8) NULLREG MSGHDRY1 null:ud TS TSMSGDSC
+
+ mov (8) MSGHDRY0<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+//------------------------------------------------------------------------
+// Scoreboard control data initialization
+//------------------------------------------------------------------------
+#ifdef AS_ENABLED
+ or (1) cr0.1:ud cr0.1:ud AS_INT_EN // Enable interrupt
+ (f0.1) jmpi (1) Scoreboard_State_Init // Jump if not restarted thread
+
+ // Restore scoreboard kernel control data to r1 - r3
+ mov (1) m4.1:ud 64:ud // Starting r1
+ mov (1) m4.2:ud 0x0002001f:ud // for 3 registers
+ send (8) r1.0<1>:ud m4 null:ud DWBRMSGDSC_SC+0x00030000+AS_SAVE // Restore r1 - r3
+ mov (8) a0.0<1>:uw AR_SAVE<8;8,1>:uw // Restore all address registers
+
+// Check whether all MBs have been decoded
+ cmp.e.f0.0 (1) NULLREG TotalMB<0;1,0>:w 0:w // Set "Last MB" flag
+ (-f0.0) jmpi (1) Before_First_MB
+ END_THREAD
+
+// Check whether it is before the first MB
+Before_First_MB:
+ cmp.e.f0.0 (1) NULLREG AVAILFLAGD<1>:ud 0x08020401:ud // in ACBD order
+ (f0.0) jmpi (1) Wavefront_Walk
+
+Scoreboard_State_Init:
+#endif // End AS_ENABLED
+ mov (2) WFLen_B<2>:w HEIGHTINMB_1<0;1,0>:w
+ mov (1) AVAILFLAGD<1>:ud 0x08020401:ud // in ACBD order
+ mov (1) CASE00PTR<1>:ud Notify_MSG_IP-No_Message_IP:ud // Inter kernel starts
+ mov (1) CASE10PTR<1>:ud Dependency_Check_IP-No_Message_IP:ud // Intra kernel starts
+#ifdef AS_ENABLED
+ mov (1) CASE11PTR<1>:ud 0:ud // No message
+#else
+ mov (1) CASE11PTR<1>:ud MB_Loop_IP-No_Message_IP:ud // No message
+#endif // End AS_ENABLED
+ mov (1) StartXD<1>:ud 0:ud
+ mov (1) NewWFOffsetD<1>:ud 0x01ffff00:ud
+
+ mov (4) WFStart(0)<1> 0xffff:w
+ mov (1) WFStart(0)<1> 0:w
+
+ mov (8) a0.0<1>:uw 0x0:uw // Initialize all pointers to 0
+
+//------------------------------------------------------------------------
+// Scoreboard message handling loop
+//------------------------------------------------------------------------
+//
+Scoreboard_Loop:
+ // Calculate current wavefront length
+ add.ge.f0.1 (16) acc0<1>:w StartX<0;1,0>:w 0:w // Used for x>2*y check
+ mac.g.f0.0 (16) NULLREGW WFLenY<0;1,0>:w -2:w // X - 2*Y > 0 ??
+ (f0.0) mov (1) WFLen<1>:w WFLenY<0;1,0>:w // Use smaller vertical wavefront length
+ (-f0.0) asr.sat (1) WFLen<1>:uw StartX<0;1,0>:w 1:w // Horizontal wavefront length is smaller
+
+ // Initialize 5-MB group
+#ifdef ONE_MB_WA
+ mov (2) MBINDEX(0)<1> WFStart(0)<2;2,1>
+ (f0.1) add (4) MBINDEX(0,2)<1> WFStart(0,1)<4;4,1> -1:w
+ (-f0.1) add (4) MBINDEX(0,2)<1> WFStart(0,0)<4;4,1> -1:w
+ (-f0.1) mov (1) StartX<1>:w 0:w // WA for 1-MB wide pictures
+#else
+ mov (2) MBINDEX(0)<1> WFStart(0)<2;2,1> {NoDDClr}
+ add (4) MBINDEX(0,2)<1> WFStart(0,1)<4;4,1> -1:w {NoDDChk}
+#endif
+
+ // Update WFStart
+ mov (8) acc0<1>:w WFStart(0)<0;1,0> // Move WFStart(0) to acc0 to remove dependency later
+ mov (4) WFStart(0,1)<1> WFStart(0)<4;4,1> {NoDDClr} // Shift WFStart(0:2) to WFStart(1:3)
+ add (1) WFStart(0)<1> acc0.0<0;1,0>:w WFLen<0;1,0>:w {NoDDChk} // WFStart(0) = WFStart(0) + WFLen
+
+ mul (8) MBINDEX(0)<1> MBINDEX(0)<8;8,1> 4:w // Adjust MB order # to be DWORD aligned
+ and (1) DEPPTR<1>:uw acc0<0;1,0>:w SB_MASK*4:uw {NoDDClr} // Wrap around scoreboard entries for current MB
+ and (4) DEPPTRL<1>:uw acc0.1<4;4,1>:w SB_MASK*4:uw {NoDDChk} // Wrap around scoreboard entries for neighbor MBs
+
+Wavefront_Walk:
+ wait n0:ud
+// Check for combined "checking" or "completed" threads in forwarded message
+// 2 MSB of scoreboard message indicate:
+// 0b00 = "inter start" message
+// 0b10 = "intra start" message
+// 0b11 = "No Message" or "inter complete" message
+// 0b01 = Reserved (should never occur)
+//
+MB_Loop:
+ shr (1) PMSGSEL<1>:uw r[CMDPTR,CMD_SB_REG_OFF*GRFWIB+2]<0;1,0>:uw 12:w // DWORD aligned pointer to message handler
+ and.nz.f0.1 (4) NULLREG r[CMDPTR,CMD_SB_REG_OFF*GRFWIB]<0;1,0>:ub AVAILFLAG<4;4,1>:ub // f0.1 4 LSB will have the available flags in ACBD order
+ mov (1) MSGHDRY0.4<1>:ud r[CMDPTR,CMD_SB_REG_OFF*GRFWIB]<0;1,0>:ud // Copy MB thread info from scoreboard
+ jmpi (1) r[PMSGSEL, INLINE_REG_OFF*GRFWIB+16]<0;1,0>:d
+
+// Now determine whether this is "inter done" or "no message"
+// through checking debug_counter
+//
+No_Message:
+#ifdef AS_ENABLED
+ cmp.z.f0.1 (1) NULLREG n0:ud 0 // Are all messages handled?
+ and.z.f0.0 (1) NULLREG cr0.1:ud AS_INT // Poll interrupt bit
+ (-f0.1) jmpi (1) MB_Loop // Continue polling the remaining message from current thread
+
+// All messages have been handled
+ (f0.0) jmpi (1) Wavefront_Walk // No interrupt occurs. Wait for next one
+
+// Interrupt has been detected
+// Save all contents and terminate the scoreboard
+//
+ #include "scoreboard_save_AS.asm"
+
+ // Save scoreboard control data as well
+ //
+ mov (8) AR_SAVE<1>:uw a0.0<8;8,1>:uw // All address registers needs to be saved
+ mov (1) MSGHDR.1:ud 64:ud
+ mov (1) MSGHDR.2:ud 0x0002001f:ud // for 3 registers
+ $for(0; <3; 1) {
+ mov (8) MSGPAYLOADD(%1)<1> CMD_SB(%1-3)REGION(8,1)
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00300000+AS_SAVE // Save r1 - r3
+
+ send (8) NULLREG MSGHDR r0:ud EOTMSGDSC+TH_INT // Terminate with "Thread Interrupted" bit set
+#endif // End AS_ENABLED
+
+Dependency_Check:
+// Current thread is "checking" but not "completed" (0b10 case).
+// Check for dependency clear using all availability bits
+//
+ (f0.1) and.z.f0.1 (4) NULLREG r[DEPPTRL,CMD_SB_REG_OFF*GRFWIB+3]<1,0>:ub DONEFLAG:uw // f0.1 4 LSB contains dependency clear
+ (f0.1.any4h) jmpi (1) Dependency_Check // Dependency not clear, keep polling..
+
+// "Checking" thread and dependency cleared, send a message to let the thread go
+//
+Notify_MSG:
+ send (8) NULLREG MSGHDRY0 null:ud MSG_GW FWDMSGDSC+NOTIFYMSG
+
+// Current macroblock has been serviced. Update to next macroblock in special zig-zag order
+//
+Update_CurMB:
+#if 0
+ add.ge.f0.0 (1) WFLen<1>:w WFLen<0;1,0>:w -1:w // Set "End of wavefront" flag
+ add (1) TotalMB<1>:w TotalMB<0;1,0>:w -1:w // Decrement "TotalMB"
+#else
+ add.ge.f0.0 (2) TotalMB<2>:w TotalMB<4;2,2>:w -1:w // Set "End of wavefront" flag and decrement "TotalMB"
+#endif
+ add (8) MBINDEX(0)<1> MBINDEX(0)<8;8,1> 4:w // Increment MB indices
+ and (1) DEPPTR<1>:uw acc0<0;1,0>:w SB_MASK*4:uw {NoDDClr} // Wrap around 256 scoreboard entries for current MB
+ and (4) DEPPTRL<1>:uw acc0.1<4;4,1>:w SB_MASK*4:uw {NoDDChk} // Wrap around 256 scoreboard entries for neighbor MBs
+ cmp.e.f0.1 (16) NULLREGW StartX<0;1,0>:uw WIDTHINMB_1<0;1,0>:uw // Set "on picture right boundary" flag
+#if 0
+ (f0.0) jmpi (1) Wavefront_Walk // Continue wavefront walking
+#else
+ (f0.0.all2h) jmpi (1) Wavefront_Walk // Continue wavefront walking
+#endif
+
+// Start new wavefront
+//
+ cmp.e.f0.0 (1) NULLREG TotalMB<0;1,0>:w 0:w // Set "Last MB" flag
+ (f0.1) add (4) WFLen<1>:w WFLen<4;4,1>:w NewWFOffset<4;4,1>:b
+ (f0.1) add (4) WFStart(0)<1> WFStart(0)<4;4,1> 1:w
+ (-f0.1) add (1) StartX<1>:w StartX<0;1,0>:w 1:w // Move to right MB
+ (-f0.1) add (1) WFStart(0)<1> WFStart(0)<0;1,0> 1:w
+
+ (-f0.0) jmpi (1) Scoreboard_Loop // Not last MB, start new wavefront walking
+
+// All MBs have decoded. Terminate the thread now
+//
+ END_THREAD
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
+
+// End of scoreboard
diff --git a/shaders/h264/mc/scoreboard_MBAFF.asm b/shaders/h264/mc/scoreboard_MBAFF.asm new file mode 100644 index 0000000..02a49d8 --- /dev/null +++ b/shaders/h264/mc/scoreboard_MBAFF.asm @@ -0,0 +1,299 @@ +/*
+ * Dependency control scoreboard kernel for MBAFF frame
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: scoreboard_MBAFF.asm
+//
+// Dependency control scoreboard kernel for MBAFF frame
+//
+// $Revision: 16 $
+// $Date: 10/18/06 4:10p $
+//
+
+// ----------------------------------------------------
+// Main: scoreboard_MBAFF
+// ----------------------------------------------------
+// ----------------------------------------------------
+// Scoreboard structure
+// ----------------------------------------------------
+//
+// 1 DWORD per thread
+//
+// Bit 31: "Checking" thread, i.e. an intra MB that sends "check dependency" message
+// Bit 30: "Completed" thread. This bit set by an "update" message from intra/inter MB.
+// Bits 29:28: Must set to 0
+// Bits 27:24: EUID
+// Bits 23:18: Reserved
+// Bits 17:16: TID
+// Bits 15:8: X offset of current MB
+// Bits 15:5: Reserved
+// Bits 4:0: 5 bits of available neighbor MB flags
+
+.kernel scoreboard_MBAFF
+SCOREBOARD_MBAFF:
+
+#ifdef _DEBUG
+// WA for FULSIM so we'll know which kernel is being debugged
+mov (1) acc0:ud 0xffaa55a5:ud
+#endif
+
+#include "header.inc"
+#include "scoreboard_header.inc"
+
+//
+// Now, begin source code....
+//
+
+.code
+
+#ifdef AS_ENABLED
+ and.z.f0.1 (1) NULLREG r0.2<0;1,0>:ud TH_RES // Is this a restarted thread previously interrupted?
+ (f0.1) jmpi (1) MBAFF_Scoreboard_Init
+
+ #include "scoreboard_restore_AS.asm"
+
+ jmpi (1) MBAFF_Scoreboard_OpenGW
+MBAFF_Scoreboard_Init:
+#endif // End AS_ENABLED
+
+// Scoreboard must be initialized to 0xc000ffff, meaning all "completed"
+// And it also avoids message mis-handling for the first MB
+ $for(0; <32; 2) {
+ mov (16) CMD_SB(%1)<1> 0xc000ffff:ud {Compr}
+ }
+#ifdef DOUBLE_SB // Scoreboard size needs to be doubled
+ $for(32; <64; 2) {
+ mov (16) CMD_SB(%1)<1> 0xc000ffff:ud {Compr}
+ }
+#endif // DOUBLE_SB
+
+//----------------------------------------------------------
+// Open message gateway for the scoreboard thread
+//
+// RegBase = r4 (0x04)
+// Gateway Size = 64 GRF registers (0x6)
+// Dispatch ID = r0.20:ub
+// Scoreboard Thread Key = 0
+//----------------------------------------------------------
+MBAFF_Scoreboard_OpenGW:
+ mov (8) MSGHDRY0<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+ // Send a message with register base RegBase=0x04(r4) and Gateway size = 0x6 = 64 GRF reg and Key = 0
+ // 000 00000100 00000 00000 110 00000000 ==> 0000 0000 1000 0000 0000 0110 0000 0000
+#ifdef AS_ENABLED
+ add (1) MSGHDRY0.5<1>:ud r0.20:ub 0x00800700:ud // Allocate 128 GRFs for message gateway - for SIP to send notification MSG
+#else
+ #ifdef DOUBLE_SB
+ add (1) MSGHDRY0.5<1>:ud r0.20:ub 0x00800600:ud // 64 GRF's for CTG-B
+ #else
+ add (1) MSGHDRY0.5<1>:ud r0.20:ub 0x00800500:ud // 32 GRF's for CTG-A
+ #endif // DOUBLE_SB
+#endif
+
+ send (8) NULLREG MSGHDRY0 null:ud MSG_GW OGWMSGDSC
+
+//------------------------------------------------------------------------
+// Send Thread Spawning Message to start dispatching macroblock threads
+//
+//------------------------------------------------------------------------
+#ifdef AS_ENABLED
+ mov (8) acc0<1>:ud CMD_SB(31)<8;8,1> // Ensure scoreboard data have been completely restored
+#endif // End AS_ENABLED
+ mov (8) MSGHDRY1<1>:ud r0<8;8,1>:ud // Initialize message header payload with R0
+ mov (1) MSGHDRY1.4<1>:ud 0x00000400:ud // Dispatch URB length = 1
+
+ send (8) NULLREG MSGHDRY1 null:ud TS TSMSGDSC
+
+ mov (8) MSGHDRY0<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+//------------------------------------------------------------------------
+// Scoreboard control data initialization
+//------------------------------------------------------------------------
+#ifdef AS_ENABLED
+ or (1) cr0.1:ud cr0.1:ud AS_INT_EN // Enable interrupt
+ (f0.1) jmpi (1) MBAFF_Scoreboard_State_Init // Jump if not restarted thread
+
+ // Restore scoreboard kernel control data to r1 - r3
+ mov (1) m4.1:ud 64:ud // Starting r1
+ mov (1) m4.2:ud 0x0002001f:ud // for 3 registers
+ send (8) r1.0<1>:ud m4 null:ud DWBRMSGDSC_SC+0x00030000+AS_SAVE // Restore r1 - r3
+ and (1) CMDPTR<1>:uw MBINDEX(0)<0;1,0> SB_MASK*4:uw // Restore scoreboard entries for current MB
+
+// EOT if all MBs have been decoded
+ cmp.e.f0.0 (1) NULLREG TotalMB<0;1,0>:w 0:w // Set "Last MB" flag
+ (-f0.0) jmpi (1) MBAFF_Before_First_MB
+ END_THREAD
+
+// Check whether it is before the first MB
+MBAFF_Before_First_MB:
+ cmp.e.f0.0 (1) NULLREG AVAILFLAGD<1>:ud 0x08020401:ud // in ACBD order
+ (f0.0) jmpi (1) MBAFF_Wavefront_Walk
+
+MBAFF_Scoreboard_State_Init:
+#endif // End AS_ENABLED
+ mov (2) WFLen_B<2>:w HEIGHTINMB_1<0;1,0>:w
+ mov (1) AVAILFLAGD<1>:ud 0x08020401:ud // in ACBD order
+ mov (1) AVAILFLAG1D<1>:ud 0x08020410:ud // in A_C_B_D_ order
+ mov (1) CASE00PTR<1>:ud MBAFF_Notify_MSG_IP-MBAFF_No_Message_IP:ud // Inter kernel starts
+ mov (1) CASE10PTR<1>:ud MBAFF_Dependency_Check_IP-MBAFF_No_Message_IP:ud // Intra kernel starts
+#ifdef AS_ENABLED
+ mov (1) CASE11PTR<1>:ud 0:ud // No message
+#else
+ mov (1) CASE11PTR<1>:ud MBAFF_MB_Loop_IP-MBAFF_No_Message_IP:ud // No message
+#endif // End AS_ENABLED
+ mov (1) StartXD<1>:ud 0:ud
+ mov (1) NewWFOffsetD<1>:ud 0x01ffff00:ud
+
+ mov (8) WFStart_T(0)<1> 0xffff:w
+ mov (1) WFStart_T(0)<1> 0:w
+
+ mov (8) a0.0<1>:uw 0x0:uw // Initialize all pointers to 0
+
+//------------------------------------------------------------------------
+// Scoreboard message handling loop
+//------------------------------------------------------------------------
+//
+MBAFF_Scoreboard_Loop:
+// Calculate current wavefront length (same for top and bottom MB wavefronts)
+ add.ge.f0.1 (16) acc0<1>:w StartX<0;1,0>:w 0:w // Used for x>2*y check
+ mac.g.f0.0 (16) NULLREGW WFLenY<0;1,0>:w -2:w // X - 2*Y > 0 ??
+ (f0.0) mov (2) WFLen_B<1>:w WFLenY<0;1,0>:w // Use smaller vertical wavefront length
+ (f0.0) mov (1) WFLen_Save<1>:w WFLenY<0;1,0>:w // Save current wave front length
+ (-f0.0) asr.sat (2) WFLen_B<1>:uw StartX<0;1,0>:w 1:w // Horizontal wavefront length is smaller
+ (-f0.0) asr.sat (1) WFLen_Save<1>:uw StartX<0;1,0>:w 1:w // Save current wave front length
+
+// Initialize 9-MB group for top macroblock wavefront
+#ifdef ONE_MB_WA_MBAFF
+ mov (2) MBINDEX(0)<1> WFStart_T(0)<2;2,1>
+ (f0.1) add (4) MBINDEX(0,2)<1> WFStart_B(0,1)<4;4,1> -1:w
+ (-f0.1) add (4) MBINDEX(0,2)<1> WFStart_B(0,0)<4;4,1> -1:w
+ mov (1) MBINDEX(0,5)<1> WFStart_B(0,1)<0;1,0>
+ (-f0.1) mov (1) StartX<1>:w 0:w // WA for 1-MB wide pictures
+#else
+ mov (2) MBINDEX(0)<1> WFStart_T(0)<2;2,1> {NoDDClr}
+ add (4) MBINDEX(0,2)<1> WFStart_B(0,1)<4;4,1> -1:w {NoDDChk,NoDDClr}
+ mov (1) MBINDEX(0,5)<1> WFStart_B(0,1)<0;1,0> {NoDDChk,NoDDClr}
+ add (4) MBINDEX(0,6)<1> WFStart_T(0,1)<4;4,1> -1:w {NoDDChk} // Upper MB group (C_B_D_x)
+#endif
+
+// Update WFStart_B[0]
+ add (8) acc0<1>:w WFLen<0;1,0>:w 1:w // WFLen + 1
+ add (1) WFStart_B(0,0)<1> acc0<0;1,0>:w WFStart_T(0,0)<0;1,0> // WFStart_T[0] + WFLen + 1
+
+MBAFF_Start_Wavefront:
+ mul (16) MBINDEX(0)<1> MBINDEX(0)REGION(16,1) 4:w // Adjust MB order # to be DWORD aligned
+ and (1) CMDPTR<1>:uw acc0<0;1,0>:w SB_MASK*4:uw // Wrap around scoreboard entries for current MB
+
+MBAFF_Wavefront_Walk:
+ wait n0:ud
+
+// Check for combined "checking" or "completed" threads in forwarded message
+// 2 MSB of scoreboard message indicate:
+// 0b00 = "inter start" message
+// 0b10 = "intra start" message
+// 0b11 = "No Message" or "inter complete" message
+// 0b01 = Reserved (should never occur)
+//
+MBAFF_MB_Loop:
+ shr (1) PMSGSEL<1>:uw r[CMDPTR,CMD_SB_REG_OFF*GRFWIB+2]<0;1,0>:uw 12:w // DWORD aligned pointer to message handler
+ and.nz.f0.1 (8) NULLREG r[CMDPTR,CMD_SB_REG_OFF*GRFWIB]<0;1,0>:ub AVAILFLAG<8;8,1>:ub // f0.1 8 LSB will have the available flags in ACBDA_C_B_D_ order
+ mov (1) MSGHDRY0.4<1>:ud r[CMDPTR,CMD_SB_REG_OFF*GRFWIB]<0;1,0>:ud // Copy MB thread info from scoreboard
+ jmpi (1) r[PMSGSEL, INLINE_REG_OFF*GRFWIB+16]<0;1,0>:d
+
+// Now determine whether this is "inter done" or "no message"
+// through checking debug_counter
+//
+MBAFF_No_Message:
+#ifdef AS_ENABLED
+ cmp.z.f0.1 (1) NULLREG n0:ud 0 // Are all messages handled?
+ and.z.f0.0 (1) NULLREG cr0.1:ud AS_INT // Poll interrupt bit
+ (-f0.1) jmpi (1) MBAFF_MB_Loop // Continue polling the remaining message from current thread
+
+// All messages have been handled
+ (f0.0) jmpi (1) MBAFF_Wavefront_Walk // No interrupt occurs. Wait for next one
+
+// Interrupt has been detected
+// Save all contents and terminate the scoreboard
+//
+ #include "scoreboard_save_AS.asm"
+
+ // Save scoreboard control data as well
+ //
+ mov (1) MSGHDR.1:ud 64:ud
+ mov (1) MSGHDR.2:ud 0x0002001f:ud // for 3 registers
+ $for(0; <3; 1) {
+ mov (8) MSGPAYLOADD(%1)<1> CMD_SB(%1-3)REGION(8,1)
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00300000+AS_SAVE // Save r1 - r3
+
+ send (8) NULLREG MSGHDR r0:ud EOTMSGDSC+TH_INT // Terminate with "Thread Interrupted" bit set
+#endif // End AS_ENABLED
+
+MBAFF_Dependency_Check:
+// Current thread is "checking" but not "completed" (0b10 case).
+// Check for dependency clear using all availability bits
+//
+ and (8) DEPPTR<1>:uw MBINDEX(0,1)REGION(8,1) SB_MASK*4:uw // Wrap around scoreboard entries for current MB
+MBAFF_Dependency_Polling:
+ (f0.1) and.z.f0.1 (8) NULLREG r[DEPPTR,CMD_SB_REG_OFF*GRFWIB+3]<1,0>:ub DONEFLAG:uw // f0.1 8 LSB contains dependency clear
+ (f0.1.any8h) jmpi (1) MBAFF_Dependency_Polling // Dependency not clear, keep polling..
+
+// "Checking" thread and dependency cleared, send a message to let the thread go
+//
+MBAFF_Notify_MSG:
+ send (8) NULLREG MSGHDRY0 null:ud MSG_GW FWDMSGDSC+NOTIFYMSG
+
+// Current macroblock has been serviced. Update to next macroblock in special zig-zag order
+//
+MBAFF_Update_CurMB:
+ add.ge.f0.0 (2) TotalMB<2>:w TotalMB<4;2,2>:w -1:w // Set "End of wavefront" flag and decrement "TotalMB"
+ add (16) MBINDEX(0)<1> MBINDEX(0)REGION(16,1) 4:w // Increment MB indices
+ and (1) CMDPTR<1>:uw acc0<0;1,0>:w SB_MASK*4:uw // Wrap around scoreboard entries for current MB
+ (f0.0.all2h) jmpi (1) MBAFF_Wavefront_Walk // Continue wavefront walking
+
+// Top macroblock wavefront walk done, start bottom MB wavefront
+ add.ge.f0.0 (1) WFLen<1>:w WFLen_B<0;1,0>:w 0:w {NoDDClr} // Set bottom MB wavefront length
+ mov (1) WFLen_B<1>:w -1:w {NoDDChk} // Reset bottom MB wavefront length
+
+// Initialize 9-MB group for bottom macroblock wavefront
+ mov (8) MBINDEX(0)<1> WFStart_B(0)<1;4,0> {NoDDClr} // Initialize with WFStart_B[0] and WFStart_B[1]
+ mov (4) MBINDEX(0,1)<1> WFStart_T(0,1)<0;1,0> {NoDDChk,NoDDClr} // Initialize with WFStart_T[1]
+ mov (2) MBINDEX(0,2)<1> WFStart_T(0)<0;1,0> {NoDDChk,NoDDClr} // Initialize with WFStart_T[0]
+ add (4) MBINDEX(0,6)<1> WFStart_B(0,1)<4;4,1> -1:w {NoDDChk} // Upper MB group (C_B_D_x)
+
+ (f0.0) jmpi (1) MBAFF_Start_Wavefront // Start bottom MB wavefront walk
+
+// Start new wavefront
+//
+ cmp.e.f0.1 (16) NULLREGW StartX<0;1,0>:uw WIDTHINMB_1<0;1,0>:uw // Set "on picture right boundary" flag
+
+ // Update WFStart_T and WFStart_B
+ add (8) acc0<1>:w WFStart_T(0)REGION(1,0) 1:w // Move WFStart_T[0]+1 to acc0 to remove dependency later
+ mov (8) WFStart_T(0,1)<1> WFStart_T(0)<8;8,1> {NoDDClr} // Shift WFStart_T(B)[0:2] to WFStart_T(B)[1:3]
+ mac (1) WFStart_T(0,0)<1> WFLen_Save<0;1,0>:w 2:w {NoDDChk} // WFStart_T[0] = WFStart_T[0] + 2*WFLen
+
+ cmp.e.f0.0 (1) NULLREG TotalMB<0;1,0>:w 0:w // Set "Last MB" flag
+
+ (f0.1) add (4) WFLen<1>:w WFLen<4;4,1>:w NewWFOffset<4;4,1>:b // + (0, -1, -1, 1)
+ (f0.1) add (8) WFStart_T(0)<1> WFStart_T(0)REGION(4,1) 1:w
+ (-f0.1) add (1) StartX<1>:w StartX<0;1,0>:w 1:w // Move to right MB
+ (-f0.1) add (1) WFStart_T(0)<1> WFStart_T(0)REGION(1,0) 1:w
+
+ (-f0.0) jmpi (1) MBAFF_Scoreboard_Loop // Not last MB, start new wavefront walking
+
+// All MBs have decoded. Terminate the thread now
+//
+ END_THREAD
+
+#if !defined(COMBINED_KERNEL) // For standalone kernel only
+.end_code
+
+.end_kernel
+#endif
+
+// End of scoreboard_MBAFF
diff --git a/shaders/h264/mc/scoreboard_restore_AS.asm b/shaders/h264/mc/scoreboard_restore_AS.asm new file mode 100644 index 0000000..7d95330 --- /dev/null +++ b/shaders/h264/mc/scoreboard_restore_AS.asm @@ -0,0 +1,54 @@ +/*
+ * Restore previously stored scoreboard data after content switching back
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: scoreboard_restore_AS.asm
+//
+// Restore previously stored scoreboard data after content switching back
+//
+//
+ // Restore scoreboard data to r4 - r67
+ // They are saved in a 2D surface with width of 32 and height of 80.
+ // Each row corresponds to one GRF register in the following order
+ // r4 - r67 : Scoreboard message
+ //
+ mov (8) MSGSRC<1>:ud r0.0<8;8,1>:ud {NoDDClr} // Initialize message header payload with r0
+
+ mov (2) MSGSRC.0:ud 0:ud {NoDDClr, NoDDChk} // Starting r4
+ mov (1) MSGSRC.2:ud 0x0007001f:ud {NoDDChk} // for 8 registers
+ send (8) CMD_SB(0)<1> m1 MSGSRC<8;8,1>:ud DWBRMSGDSC_SC+0x00080000+AS_SAVE // Restore r4 - r11
+
+ mov (8) m2:ud MSGSRC<8;8,1>:ud
+ mov (1) m2.1:ud 8:ud
+ send (8) CMD_SB(8)<1> m2 null:ud DWBRMSGDSC_SC+0x00080000+AS_SAVE // Restore r12 - r19
+
+ mov (8) m3:ud MSGSRC<8;8,1>:ud
+ mov (1) m3.1:ud 16:ud
+ send (8) CMD_SB(16)<1> m3 null:ud DWBRMSGDSC_SC+0x00080000+AS_SAVE // Restore r20 - r27
+
+ mov (8) m4:ud MSGSRC<8;8,1>:ud
+ mov (1) m4.1:ud 24:ud
+ send (8) CMD_SB(24)<1> m4 null:ud DWBRMSGDSC_SC+0x00080000+AS_SAVE // Restore r28 - r35
+
+ mov (8) m5:ud MSGSRC<8;8,1>:ud
+ mov (1) m5.1:ud 32:ud
+ send (8) CMD_SB(32)<1> m5 null:ud DWBRMSGDSC_SC+0x00080000+AS_SAVE // Restore r36 - r43
+
+ mov (8) m6:ud MSGSRC<8;8,1>:ud
+ mov (1) m6.1:ud 40:ud
+ send (8) CMD_SB(40)<1> m6 null:ud DWBRMSGDSC_SC+0x00080000+AS_SAVE // Restore r44 - r51
+
+ mov (8) m7:ud MSGSRC<8;8,1>:ud
+ mov (1) m7.1:ud 48:ud
+ send (8) CMD_SB(48)<1> m7 null:ud DWBRMSGDSC_SC+0x00080000+AS_SAVE // Restore r52 - r59
+
+ mov (8) m8:ud MSGSRC<8;8,1>:ud
+ mov (1) m8.1:ud 56:ud
+ send (8) CMD_SB(56)<1> m8 null:ud DWBRMSGDSC_SC+0x00080000+AS_SAVE // Restore r60 - r67
+
+// End of scoreboard_restore_AS
diff --git a/shaders/h264/mc/scoreboard_save_AS.asm b/shaders/h264/mc/scoreboard_save_AS.asm new file mode 100644 index 0000000..13abe0e --- /dev/null +++ b/shaders/h264/mc/scoreboard_save_AS.asm @@ -0,0 +1,72 @@ +/*
+ * Save scoreboard data before content switching
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Module name: scoreboard_save_AS.asm
+//
+// Save scoreboard data before content switching
+//
+//
+ // r1 - r35 need to be saved
+ // They are saved in a 2D surface with width of 32 and height of 64.
+ // Each row corresponds to one GRF register in the following order
+ // r4 - r35 : Scoreboard message
+ // r1 - r3 : Scoreboard kernel control data
+
+ mov (8) MSGHDR<1>:ud r0.0<8;8,1>:ud // Initialize message header payload with r0
+ mov (1) MSGHDR.2:ud 0x0007001f:ud // for 8 registers
+
+ mov (2) MSGHDR.0:ud 0:ud
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> CMD_SB(%1)REGION(8,1) {Compr}
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00800000+AS_SAVE // Save r4 - r11
+
+ mov (1) MSGHDR.1:ud 8:ud
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> CMD_SB(%1+8)REGION(8,1) {Compr}
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00800000+AS_SAVE // Save r12 - r19
+
+ mov (1) MSGHDR.1:ud 16:ud
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> CMD_SB(%1+16)REGION(8,1) {Compr}
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00800000+AS_SAVE // Save r20 - r27
+
+ mov (1) MSGHDR.1:ud 24:ud
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> CMD_SB(%1+24)REGION(8,1) {Compr}
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00800000+AS_SAVE // Save r28 - r35
+
+ mov (1) MSGHDR.1:ud 32:ud
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> CMD_SB(%1+32)REGION(8,1) {Compr}
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00800000+AS_SAVE // Save r36 - r43
+
+ mov (1) MSGHDR.1:ud 40:ud
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> CMD_SB(%1+40)REGION(8,1) {Compr}
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00800000+AS_SAVE // Save r44 - r51
+
+ mov (1) MSGHDR.1:ud 48:ud
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> CMD_SB(%1+48)REGION(8,1) {Compr}
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00800000+AS_SAVE // Save r52 - r59
+
+ mov (1) MSGHDR.1:ud 56:ud
+ $for(0; <8; 2) {
+ mov (16) MSGPAYLOADD(%1)<1> CMD_SB(%1+56)REGION(8,1) {Compr}
+ }
+ send (8) NULLREG MSGHDR null:ud DWBWMSGDSC+0x00800000+AS_SAVE // Save r60 - r67
+
+// End of scoreboard_save_AS
diff --git a/shaders/h264/mc/scoreboard_sip.asm b/shaders/h264/mc/scoreboard_sip.asm new file mode 100644 index 0000000..6330ea1 --- /dev/null +++ b/shaders/h264/mc/scoreboard_sip.asm @@ -0,0 +1,34 @@ +/*
+ * Scoreboard interrupt handler
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: scoreboard_sip.asm
+//
+// scoreboard interrupt handler
+//
+// Simply send a notification message to scoreboard thread
+
+ mov (8) m0<1>:ud 0x00000000:ud // Initialize message header payload with 0
+#ifdef DOUBLE_SB
+ mov (1) m0.5<1>:ud 0x08000200:ud // Message length = 1 DWORD, sent to GRF offset 64 registers
+#else
+ mov (1) m0.5<1>:ud 0x04000200:ud // Message length = 1 DWORD, sent to GRF offset 32 registers
+#endif
+ send (8) null<1>:ud m0 null:ud 0x03108002 // Send notification message to scoreboard kernel
+
+ and (1) cr0.1:ud cr0.1:ud 0x00800000 // Clear preempt exception bit
+ and (1) cr0.0:ud cr0.0:ud 0x7fffffff:ud // Exit SIP routine
+ nop // Required by B-spec
+
+.end_code
+
+
+
+
+
+
diff --git a/shaders/h264/mc/scoreboard_start_inter.asm b/shaders/h264/mc/scoreboard_start_inter.asm new file mode 100644 index 0000000..831b841 --- /dev/null +++ b/shaders/h264/mc/scoreboard_start_inter.asm @@ -0,0 +1,47 @@ +/*
+ * Scoreboard function for starting inter prediction kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__SCOREBOARD_START_INTER__)
+#define __SCOREBOARD_START_INTER__
+//
+// Module name: scoreboard_start_inter.asm
+//
+// Scoreboard function for starting inter prediction kernels
+// This function is only used by inter prediction kernels to send message to
+// scoreboard in order to announce the inter kernel has started
+//
+// $Revision: 5 $
+// $Date: 10/18/06 4:11p $
+//
+scoreboard_start_inter:
+
+// First open message gateway since intra kernels need wake-up message to resume
+//
+ mov (8) MSGHDRY0<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+ // Send a message with register base RegBase = r0 (0x0) and Size = 0x0
+ // 000 00000000 00000 00000 000 00000000 ==> 0000 0000 0000 0000 0000 0000 0000 0000
+ // ---------------------------------------------------------------------------------
+ send (8) NULLREG MSGHDRY0 null:ud MSG_GW OGWMSGDSC
+
+// Derive the scoreboard location where the inter thread writes to
+//
+ mov (8) MSGHDRY1<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+ // Compose M0.5:ud
+ #include "set_SB_offset.asm"
+
+ // Compose M0.0:ud, i.e. message payload
+ or (1) MSGHDRY1.1<1>:uw sr0.0<0;1,0>:uw 0x0000:uw // Set EUID/TID bits + inter start bit
+
+ send (8) NULLREG MSGHDRY1 null:ud MSG_GW FWDMSGDSC+NOTIFYMSG // Send "Inter start" message to scoreboard kernel
+
+ RETURN
+
+#endif // !defined(__SCOREBOARD_START_INTER__)
diff --git a/shaders/h264/mc/scoreboard_start_intra.asm b/shaders/h264/mc/scoreboard_start_intra.asm new file mode 100644 index 0000000..6d6d916 --- /dev/null +++ b/shaders/h264/mc/scoreboard_start_intra.asm @@ -0,0 +1,52 @@ +/*
+ * Scoreboard function for starting intra prediction kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+#if !defined(__SCOREBOARD_START_INTRA__)
+#define __SCOREBOARD_START_INTRA__
+//
+// Module name: scoreboard_start_intra.asm
+//
+// Scoreboard function for starting intra prediction kernels
+// This function is only used by intra prediction kernels to send message to
+// scoreboard in order to check dependency clearance
+//
+// $Revision: 5 $
+// $Date: 10/18/06 4:11p $
+//
+scoreboard_start_intra:
+
+// First open message gateway since intra kernels need wake-up message to resume
+//
+ mov (8) MSGHDRY0<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+ // Send a message with register base RegBase = r0 (0x0) and Size = 0x0
+ // 000 00000000 00000 00000 000 00000000 ==> 0000 0000 0000 0000 0000 0000 0000 0000
+ // ---------------------------------------------------------------------------------
+ and (1) MSGHDRY0.8<1>:uw REG_INTRA_PRED_AVAIL_FLAG_BYTE<0;1,0>:ub 0x1f:uw // Set lower word of key
+ send (8) NULLREG MSGHDRY0 null:ud MSG_GW OGWMSGDSC
+
+// Send "check dependency" message to scoreboard thread
+// --------------------------
+
+// Derive the scoreboard location where the intra thread writes to
+//
+ mov (8) MSGHDRY1<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+ // Compose M0.5:ud
+ #include "set_SB_offset.asm"
+
+ // Compose M0.0:ud, i.e. message payload
+ and (1) MSGHDRY1.0<1>:uw REG_INTRA_PRED_AVAIL_FLAG_BYTE<0;1,0>:ub 0x1f:uw // Set lower word of message
+ or (1) MSGHDRY1.1<1>:uw sr0.0<0;1,0>:uw 0x8000:uw // Set EUID/TID bits + intra start bit
+
+ send (8) NULLREG MSGHDRY1 null:ud MSG_GW FWDMSGDSC+NOTIFYMSG // Send "Intra start" message to scoreboard kernel
+
+ RETURN
+
+#endif // !defined(__SCOREBOARD_START_INTRA__)
diff --git a/shaders/h264/mc/scoreboard_update.asm b/shaders/h264/mc/scoreboard_update.asm new file mode 100644 index 0000000..f519e4a --- /dev/null +++ b/shaders/h264/mc/scoreboard_update.asm @@ -0,0 +1,41 @@ +/*
+ * Scoreboard update function for decoding kernels
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//
+// Module name: scoreboard_update.asm
+//
+// Scoreboard update function for decoding kernels
+//
+// This module is used by decoding kernels to send message to scoreboard to update the
+// "complete" status, thus the dependency of the MB can be cleared.
+//
+// $Revision: 6 $
+// $Date: 10/16/06 5:19p $
+//
+ mov (8) MSGHDRY1<1>:ud 0x00000000:ud // Initialize message header payload with 0
+
+ // Compose M0.5:ud information
+ add (1) MSGHDRY1.10<1>:uw r0.20:ub 0x0200:uw // Message length = 1 DWORD
+ and (1) MSGHDRY1.11<1>:uw M05_STORE<0;1,0>:uw SB_MASK*4:uw // Retrieve stored value and wrap around scoreboard
+
+ or (1) MSGHDRY1.0<1>:ud M05_STORE<0;1,0>:uw 0xc0000000:ud // Set "Completed" bits
+
+#ifndef BSDRESET_ENABLE
+#ifdef INTER_KERNEL
+ mov (1) gREG_WRITE_COMMIT_Y<1>:ud gREG_WRITE_COMMIT_Y<0;1,0>:ud // Make sure Y write is committed
+ mov (1) gREG_WRITE_COMMIT_UV<1>:ud gREG_WRITE_COMMIT_UV<0;1,0>:ud // Make sure U/V write is committed
+#else
+ mov (1) REG_WRITE_COMMIT_Y<1>:ud REG_WRITE_COMMIT_Y<0;1,0>:ud // Make sure Y write is committed
+ mov (1) REG_WRITE_COMMIT_UV<1>:ud REG_WRITE_COMMIT_UV<0;1,0>:ud // Make sure U/V write is committed
+#endif // INTER_KERNEL
+#endif // BSDRESET_ENABLE
+
+ send (8) NULLREG MSGHDRY1 null:ud MSG_GW FWDMSGDSC
+
+// End of scoreboard_update
diff --git a/shaders/h264/mc/set_SB_offset.asm b/shaders/h264/mc/set_SB_offset.asm new file mode 100644 index 0000000..0b166e4 --- /dev/null +++ b/shaders/h264/mc/set_SB_offset.asm @@ -0,0 +1,26 @@ +/*
+ * Common module to set offset into the scoreboard
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+//
+// Module name: set_SB_offset.asm
+//
+// Common module to set offset into the scoreboard
+// Note: This is to encapsulate the way M0.5:ud in ForwardMsg is filled.
+//
+// $Revision: 2 $
+// $Date: 10/16/06 5:19p $
+//
+ add (1) MSGHDRY1.10<1>:uw r0.20:ub 0x0200:uw // Message length = 1 DWORD
+
+ add (16) acc0<1>:w r0.12<0;1,0>:uw -LEADING_THREAD:w // 0-based thread count derived from r0.6:ud
+ shl (1) M05_STORE<1>:uw acc0<0;1,0>:uw 0x2:uw // Store for future "update" use, in DWORD unit
+ and (16) acc0<1>:w acc0<16;16,1>:uw SB_MASK:uw // Wrap around scoreboard
+ shl (1) MSGHDRY1.11<1>:uw acc0<0;1,0>:uw 0x2:uw // Convert to DWORD offset
+
+// End of set_SB_offset
\ No newline at end of file diff --git a/shaders/h264/mc/weightedPred.asm b/shaders/h264/mc/weightedPred.asm new file mode 100644 index 0000000..76525f9 --- /dev/null +++ b/shaders/h264/mc/weightedPred.asm @@ -0,0 +1,140 @@ +/*
+ * Weighted prediction of luminance and chrominance
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: WeightedPred.asm
+//
+// Weighted prediction of luminance and chrominance
+//
+
+
+//#if !defined(__WeightedPred__) // Make sure this is only included once
+//#define __WeightedPred__
+
+
+ and.z.f0.0 (1) gWEIGHTFLAG:w gWPREDFLAG:ub nWBIDIR_MASK:w
+ cmp.e.f0.1 (1) null:w gPREDFLAG:w 2:w
+ (-f0.0) jmpi INTERLABEL(WeightedPred)
+ (f0.1) jmpi INTERLABEL(DefaultWeightedPred_BiPred)
+
+INTERLABEL(DefaultWeightedPred_UniPred):
+
+ cmp.e.f0.0 (1) null:w gPREDFLAG:w 0:w
+ (f0.0) jmpi INTERLABEL(Return_WeightedPred)
+
+ // luma
+ mov (32) gubYPRED(0)<2> gubINTPY1(0) {Compr}
+ mov (32) gubYPRED(2)<2> gubINTPY1(2) {Compr}
+
+#ifndef MONO
+ // chroma
+ mov (32) gubCPRED(0)<2> gubINTPC1(0) {Compr}
+#endif
+
+ jmpi INTERLABEL(Return_WeightedPred)
+
+INTERLABEL(DefaultWeightedPred_BiPred):
+
+ // luma
+ avg.sat (32) gubYPRED(0)<2> gubINTPY0(0) gubINTPY1(0) {Compr}
+ avg.sat (32) gubYPRED(2)<2> gubINTPY0(2) gubINTPY1(2) {Compr}
+
+#ifndef MONO
+ // chroma
+ avg.sat (32) gubCPRED(0)<2> gubINTPC0(0) gubINTPC1(0) {Compr}
+#endif
+
+ jmpi INTERLABEL(Return_WeightedPred)
+
+INTERLABEL(WeightedPred):
+ cmp.e.f0.1 (1) null:w gWEIGHTFLAG:w 0x80:w
+ (-f0.1) jmpi INTERLABEL(WeightedPred_Explicit)
+
+ cmp.e.f0.0 (1) null:w gPREDFLAG:w 2:w
+ (-f0.0) jmpi INTERLABEL(DefaultWeightedPred_UniPred)
+
+ mov (2) gYADD<1>:w 32:w {NoDDClr}
+ mov (2) gYSHIFT<1>:w 6:w {NoDDChk}
+ mov (4) gOFFSET<1>:w 0:w
+ mov (8) gWT0<2>:w r[pWGT,0]<0;2,1>:w
+
+ jmpi INTERLABEL(WeightedPred_LOOP)
+
+ // Explicit Prediction
+INTERLABEL(WeightedPred_Explicit):
+
+ // WA for weighted prediction - 2007/09/06
+#ifdef SW_W_128 // CTG SW WA
+ cmp.e.f0.1 (8) null:ud r[pWGT,0]<8;8,1>:uw gudW128(0)<0;1,0>
+#else // ILK HW solution
+ and.ne.f0.1 (8) null:uw r[pWGT,12]<0;1,0>:ub 0x88848421:v // Expand W=128 flag to all components. 2 MSB are don't care
+#endif
+ asr.nz.f0.0 (2) gBIPRED<1>:w gPREDFLAG<0;1,0>:w 1:w
+ asr (1) gWEIGHTFLAG:w gWEIGHTFLAG:w 6:w
+ (-f0.0) mov (2) gPREDFLAG1<1>:w gPREDFLAG<0;1,0>:w
+ (f0.0) mov (2) gPREDFLAG0<1>:ud 0x00010001:ud
+ (-f0.0) add (2) gPREDFLAG0<1>:w -gPREDFLAG1<2;2,1>:w 1:w
+
+ // WA for weighted prediction - 2007/09/06
+ (f0.1) mov (8) gWT0<1>:ud 0x00000080:ud
+ (-f0.1) mov (8) gWT0<2>:w r[pWGT,0]<16;8,2>:b
+ (-f0.1) mov (8) gO0<2>:w r[pWGT,1]<16;8,2>:b
+ mul (16) gWT0<1>:w gWT0<16;16,1>:w gPREDFLAG0<0;4,1>:w
+
+ // Compute addition
+ cmp.e.f0.1 (2) null<1>:w gYWDENOM<2;2,1>:ub 0:w
+ (-f0.1) shl (2) gW0<1>:w gWEIGHTFLAG<0;1,0>:w gYWDENOM<2;2,1>:ub
+ (f0.1) mov (2) gW0<1>:w 0:w
+ (-f0.1) asr (2) gW0<1>:w gW0<2;2,1>:w 1:w
+ shl (2) gYADD<1>:w gW0<2;2,1>:w gBIPRED<0;1,0>:w
+ (f0.1) add (2) gYADD<1>:w gYADD<2;2,1>:w gBIPRED<0;1,0>:w
+
+ // Compute shift
+ add (2) gYSHIFT<1>:w gYWDENOM<2;2,1>:ub gBIPRED<0;1,0>:w
+
+ // Compute offset
+ add (4) acc0<1>:w gO0<16;4,4>:w gO1<16;4,4>:w
+ add (4) acc0<1>:w acc0<4;4,1>:w gBIPRED<0;1,0>:w
+ asr (4) gOFFSET<1>:w acc0<4;4,1>:w gBIPRED<0;1,0>:w
+
+INTERLABEL(WeightedPred_LOOP):
+ // luma
+ $for(0;<4;2) {
+ mul (16) acc0<1>:w gubINTPY0(%1) gWT0<0;1,0>:w
+ mul (16) acc1<1>:w gubINTPY0(%1+1) gWT0<0;1,0>:w
+ mac (16) acc0<1>:w gubINTPY1(%1) gWT1<0;1,0>:w
+ mac (16) acc1<1>:w gubINTPY1(%1+1) gWT1<0;1,0>:w
+ add (16) acc0<1>:w acc0<16;16,1>:w gYADD:w
+ add (16) acc1<1>:w acc1<16;16,1>:w gYADD:w
+ // Accumulator cannot be used as destination for ASR
+ asr (16) gwINTERIM_BUF3(0)<1> acc0<16;16,1>:w gYSHIFT:w
+ asr (16) gwINTERIM_BUF3(1)<1> acc1<16;16,1>:w gYSHIFT:w
+ add.sat (16) gubYPRED(%1)<2> gwINTERIM_BUF3(0) gOFFSET:w
+ add.sat (16) gubYPRED(%1+1)<2> gwINTERIM_BUF3(1) gOFFSET:w
+ }
+
+#ifndef MONO
+ // chroma
+ mul (16) acc0<1>:w gubINTPC0(0) gUW0<0;2,4>:w
+ mul (16) acc1<1>:w gubINTPC0(1) gUW0<0;2,4>:w
+ mac (16) acc0<1>:w gubINTPC1(0) gUW1<0;2,4>:w
+ mac (16) acc1<1>:w gubINTPC1(1) gUW1<0;2,4>:w
+ add (16) acc0<1>:w acc0<16;16,1>:w gCADD:w
+ add (16) acc1<1>:w acc1<16;16,1>:w gCADD:w
+ // Accumulator cannot be used as destination for ASR
+ asr (16) gwINTERIM_BUF3(0)<1> acc0<16;16,1>:w gCSHIFT:w
+ asr (16) gwINTERIM_BUF3(1)<1> acc1<16;16,1>:w gCSHIFT:w
+ add.sat (16) gubCPRED(0)<2> gwINTERIM_BUF3(0) gUOFFSET<0;2,1>:w
+ add.sat (16) gubCPRED(1)<2> gwINTERIM_BUF3(1) gUOFFSET<0;2,1>:w
+#endif
+
+
+INTERLABEL(Return_WeightedPred):
+
+
+//#endif // !defined(__WeightedPred__)
diff --git a/shaders/h264/mc/writeRecon_C_8x4.asm b/shaders/h264/mc/writeRecon_C_8x4.asm new file mode 100644 index 0000000..be7585e --- /dev/null +++ b/shaders/h264/mc/writeRecon_C_8x4.asm @@ -0,0 +1,46 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: WriteRecon_C_8x4.asm
+//
+// $Revision: 10 $
+// $Date: 10/03/06 5:28p $
+//
+
+
+//#if !defined(__WRITERECON_C_8x4__) // Make sure this is only included once
+//#define __WRITERECON_C_8x4__
+
+
+ // TODO: Why did I use p0?
+#ifndef MONO
+ add (1) p0:w pERRORC:w -16:w
+ mov (16) mbMSGPAYLOADC(0,0)<2> r[p0,0]<32;16,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADC(0,1)<2> r[p0,128]<32;16,2>:ub {NoDDChk}
+ mov (16) mbMSGPAYLOADC(1,0)<2> r[p0,32]<32;16,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADC(1,1)<2> r[p0,128+32]<32;16,2>:ub {NoDDChk}
+#else // defined(MONO)
+ mov (16) mbMSGPAYLOADC(0)<1> 0x80808080:ud {Compr}
+#endif // !defined(MONO)
+
+ #if defined(MBAFF)
+ add (1) pMSGDSC:ud gFIELDFLAGS:uw MSG_LEN(2)+nDWBWMSGDSC+nBDIX_DESTC+ENWRCOM:ud
+ #elif defined(FIELD)
+ add (1) pMSGDSC:ud gFIELDFLAGS:uw MSG_LEN(2)+nDWBWMSGDSC_TF+nBDIX_DESTC+ENWRCOM:ud
+ #endif
+
+ asr (1) gMSGSRC.1:d gMSGSRC.1:d 1:w {NoDDClr}
+ mov (1) gMSGSRC.2:ud 0x0003000f:ud {NoDDChk} // NV12 (16x4)
+
+#if defined(FRAME)
+ send (8) gREG_WRITE_COMMIT_UV<1>:ud mMSGHDRCW gMSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(2)+nDWBWMSGDSC+nBDIX_DESTC+ENWRCOM
+#else
+ send (8) gREG_WRITE_COMMIT_UV<1>:ud mMSGHDRCW gMSGSRC<8;8,1>:ud DAPWRITE pMSGDSC:ud
+#endif // defined(FRAME)
+
+//#endif // !defined(__WRITERECON_C_8x4__)
diff --git a/shaders/h264/mc/writeRecon_YC.asm b/shaders/h264/mc/writeRecon_YC.asm new file mode 100644 index 0000000..ff84aff --- /dev/null +++ b/shaders/h264/mc/writeRecon_YC.asm @@ -0,0 +1,79 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: WriteRecon_YC.asm
+//
+// $Revision: 10 $
+// $Date: 10/03/06 5:28p $
+//
+
+
+//#if !defined(__WRITERECON_YC__) // Make sure this is only included once
+//#define __WRITERECON_YC__
+
+ // TODO: Merge two inst to one.
+ mov (1) p0:w nOFFSET_ERRORY:w
+ mov (1) p1:w nOFFSET_ERRORY+128:w
+
+ $for(0; <4; 1) {
+ mov (16) mbMSGPAYLOADY(%1,0)<1> r[p0,%1*32+0]<8,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADY(%1,16)<1> r[p0,%1*32+16]<8,2>:ub {NoDDChk}
+ }
+ $for(0; <4; 1) {
+ mov (16) mbMSGPAYLOADY(%1+4,0)<1> r[p0,%1*32+256]<8,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADY(%1+4,16)<1> r[p0,%1*32+16+256]<8,2>:ub {NoDDChk}
+ }
+
+
+ #if defined(MBAFF)
+ add (1) pMSGDSC:ud gFIELDFLAGS:uw MSG_LEN(8)+nDWBWMSGDSC+nBDIX_DESTY+ENWRCOM:ud
+ #elif defined(FIELD)
+ add (1) pMSGDSC:ud gFIELDFLAGS:uw MSG_LEN(8)+nDWBWMSGDSC_TF+nBDIX_DESTY+ENWRCOM:ud
+ #endif
+
+ mov (2) gMSGSRC.0<1>:d gX<2;2,1>:w {NoDDClr}
+ mov (1) gMSGSRC.2:ud 0x000f000f:ud {NoDDChk}
+
+#if defined(FRAME)
+ send (8) gREG_WRITE_COMMIT_Y<1>:ud mMSGHDRYW gMSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(8)+nDWBWMSGDSC+nBDIX_DESTY+ENWRCOM
+#else
+ send (8) gREG_WRITE_COMMIT_Y<1>:ud mMSGHDRYW gMSGSRC<8;8,1>:ud DAPWRITE pMSGDSC:ud
+#endif
+
+#ifndef MONO
+ // TODO: Why did I use p0?
+ mov (1) p0:w nOFFSET_ERRORC:w
+ mov (16) mbMSGPAYLOADC(0,0)<2> r[p0,0]<32;16,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADC(0,1)<2> r[p0,128]<32;16,2>:ub {NoDDChk}
+ mov (16) mbMSGPAYLOADC(1,0)<2> r[p0,32]<32;16,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADC(1,1)<2> r[p0,128+32]<32;16,2>:ub {NoDDChk}
+ mov (16) mbMSGPAYLOADC(2,0)<2> r[p0,64]<32;16,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADC(2,1)<2> r[p0,128+64]<32;16,2>:ub {NoDDChk}
+ mov (16) mbMSGPAYLOADC(3,0)<2> r[p0,96]<32;16,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADC(3,1)<2> r[p0,128+96]<32;16,2>:ub {NoDDChk}
+
+
+ #if defined(MBAFF)
+ add (1) pMSGDSC:ud gFIELDFLAGS:uw MSG_LEN(4)+nDWBWMSGDSC+nBDIX_DESTC+ENWRCOM:ud
+ #elif defined(FIELD)
+ add (1) pMSGDSC:ud gFIELDFLAGS:uw MSG_LEN(4)+nDWBWMSGDSC_TF+nBDIX_DESTC+ENWRCOM:ud
+ #endif
+
+ asr (1) gMSGSRC.1:d gMSGSRC.1:d 1:w {NoDDClr}
+ mov (1) gMSGSRC.2:ud 0x0007000f:ud {NoDDChk} // NV12 (16x4)
+
+#if defined(FRAME)
+ send (8) gREG_WRITE_COMMIT_UV<1>:ud mMSGHDRCW gMSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(4)+nDWBWMSGDSC+nBDIX_DESTC+ENWRCOM
+#else
+ send (8) gREG_WRITE_COMMIT_UV<1>:ud mMSGHDRCW gMSGSRC<8;8,1>:ud DAPWRITE pMSGDSC:ud
+#endif // defined(FRAME)
+
+#endif // !defined(MONO)
+
+
+//#endif // !defined(__WRITERECON_YC__)
diff --git a/shaders/h264/mc/writeRecon_Y_16x8.asm b/shaders/h264/mc/writeRecon_Y_16x8.asm new file mode 100644 index 0000000..509a2ec --- /dev/null +++ b/shaders/h264/mc/writeRecon_Y_16x8.asm @@ -0,0 +1,43 @@ +/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0. The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ *
+ */
+// Kernel name: WriteRecon_Y_16x8.asm
+//
+// $Revision: 10 $
+// $Date: 10/03/06 5:28p $
+//
+
+
+//#if !defined(__WRITERECON_Y_16x8__) // Make sure this is only included once
+//#define __WRITERECON_Y_16x8__
+
+
+ add (1) p0:w pERRORY:w -256:w
+ add (1) p1:w pERRORY:w -128:w
+
+ $for(0; <4; 1) {
+ mov (16) mbMSGPAYLOADY(%1,0)<1> r[p0,%1*32+0]<8,2>:ub {NoDDClr}
+ mov (16) mbMSGPAYLOADY(%1,16)<1> r[p0,%1*32+16]<8,2>:ub {NoDDChk}
+ }
+
+ #if defined(MBAFF)
+ add (1) pMSGDSC:ud gFIELDFLAGS:uw MSG_LEN(4)+nDWBWMSGDSC+nBDIX_DESTY+ENWRCOM:ud
+ #elif defined(FIELD)
+ add (1) pMSGDSC:ud gFIELDFLAGS:uw MSG_LEN(4)+nDWBWMSGDSC_TF+nBDIX_DESTY+ENWRCOM:ud
+ #endif
+
+ mov (2) gMSGSRC.0<1>:d gX<2;2,1>:w {NoDDClr}
+ mov (1) gMSGSRC.2:ud 0x0007000f:ud {NoDDChk}
+
+#if defined(FRAME)
+ send (8) gREG_WRITE_COMMIT_Y<1>:ud mMSGHDRYW gMSGSRC<8;8,1>:ud DAPWRITE MSG_LEN(4)+nDWBWMSGDSC+nBDIX_DESTY+ENWRCOM
+#else
+ send (8) gREG_WRITE_COMMIT_Y<1>:ud mMSGHDRYW gMSGSRC<8;8,1>:ud DAPWRITE pMSGDSC:ud
+#endif
+
+//#endif // !defined(__WRITERECON_Y_16x8__)
diff --git a/shaders/render/exa_wm.g4i b/shaders/render/exa_wm.g4i index 5d3d45b..8163de5 100644 --- a/shaders/render/exa_wm.g4i +++ b/shaders/render/exa_wm.g4i @@ -35,6 +35,9 @@ define(`dst_y_uw', `g1.10<2,4,0>UW') define(`screen_x0', `g1.0<0,1,0>F') define(`screen_y0', `g1.4<0,1,0>F') +/* UV flag */ +define(`interleaved_uv', `g2.0<0,1,0>UW') + /* Source transformation parameters */ define(`src_du_dx', `g3.0<0,1,0>F') define(`src_du_dy', `g3.4<0,1,0>F') diff --git a/shaders/render/exa_wm_src_sample_planar.g4a b/shaders/render/exa_wm_src_sample_planar.g4a index ca77b48..8cbb289 100644 --- a/shaders/render/exa_wm_src_sample_planar.g4a +++ b/shaders/render/exa_wm_src_sample_planar.g4a @@ -34,23 +34,18 @@ include(`exa_wm.g4i') /* use simd16 sampler, param 0 is u, param 1 is v. */ /* 'payload' loading, assuming tex coord start from g4 */ +and.nz (1) null interleaved_uv 0x01UW {align1}; +(f0) jmpi INTERLEAVED_UV; + /* load r */ mov (1) g0.8<1>UD 0x0000e000UD { align1 mask_disable }; /* src_msg will be copied with g0, as it contains send desc */ /* emit sampler 'send' cmd */ -/* sample Y */ -send (16) src_msg_ind /* msg reg index */ - src_sample_g<1>UW /* readback */ - g0<8,8,1>UW /* copy to msg start reg*/ - sampler (1,0,F) /* sampler message description, (binding_table,sampler_index,datatype) - /* here(src->dst) we should use src_sampler and src_surface */ - mlen 5 rlen 2 { align1 }; /* required message len 5, readback len 8 */ - /* sample U (Cr) */ send (16) src_msg_ind /* msg reg index */ - src_sample_r<1>UW /* readback */ + src_sample_g<1>UW /* readback */ g0<8,8,1>UW /* copy to msg start reg*/ sampler (3,2,F) /* sampler message description, (binding_table,sampler_index,datatype) /* here(src->dst) we should use src_sampler and src_surface */ @@ -63,3 +58,30 @@ send (16) src_msg_ind /* msg reg index */ sampler (5,4,F) /* sampler message description, (binding_table,sampler_index,datatype) /* here(src->dst) we should use src_sampler and src_surface */ mlen 5 rlen 2 { align1 }; /* required message len 5, readback len 8 */ + +jmpi SAMPLE_Y; + +INTERLEAVED_UV: +/* load r */ +mov (1) g0.8<1>UD 0x0000c000UD { align1 mask_disable }; + +/* sample UV (CrCb) */ +send (16) src_msg_ind /* msg reg index */ + src_sample_g<1>UW /* readback */ + g0<8,8,1>UW /* copy to msg start reg*/ + sampler (3,2,F) /* sampler message description, (binding_table,sampler_index,datatype) + /* here(src->dst) we should use src_sampler and src_surface */ + mlen 5 rlen 4 { align1 }; /* required message len 5, readback len 8 */ + +/* load r */ +mov (1) g0.8<1>UD 0x0000e000UD { align1 mask_disable }; + +SAMPLE_Y: +/* sample Y */ +send (16) src_msg_ind /* msg reg index */ + src_sample_r<1>UW /* readback */ + g0<8,8,1>UW /* copy to msg start reg*/ + sampler (1,0,F) /* sampler message description, (binding_table,sampler_index,datatype) + /* here(src->dst) we should use src_sampler and src_surface */ + mlen 5 rlen 2 { align1 }; /* required message len 5, readback len 8 */ + diff --git a/shaders/render/exa_wm_src_sample_planar.g4b b/shaders/render/exa_wm_src_sample_planar.g4b index 77a5c23..94f2f3b 100644 --- a/shaders/render/exa_wm_src_sample_planar.g4b +++ b/shaders/render/exa_wm_src_sample_planar.g4b @@ -1,4 +1,10 @@ + { 0x02000005, 0x20002d3c, 0x00000040, 0x00010001 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000004 }, { 0x00000201, 0x20080061, 0x00000000, 0x0000e000 }, - { 0x01800031, 0x22001d29, 0x008d0000, 0x02520001 }, - { 0x01800031, 0x21c01d29, 0x008d0000, 0x02520203 }, + { 0x01800031, 0x22001d29, 0x008d0000, 0x02520203 }, { 0x01800031, 0x22401d29, 0x008d0000, 0x02520405 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000003 }, + { 0x00000201, 0x20080061, 0x00000000, 0x0000c000 }, + { 0x01800031, 0x22001d29, 0x008d0000, 0x02540203 }, + { 0x00000201, 0x20080061, 0x00000000, 0x0000e000 }, + { 0x01800031, 0x21c01d29, 0x008d0000, 0x02520001 }, diff --git a/shaders/render/exa_wm_src_sample_planar.g4b.gen5 b/shaders/render/exa_wm_src_sample_planar.g4b.gen5 index a381e68..c645723 100644 --- a/shaders/render/exa_wm_src_sample_planar.g4b.gen5 +++ b/shaders/render/exa_wm_src_sample_planar.g4b.gen5 @@ -1,4 +1,10 @@ + { 0x02000005, 0x20002d3c, 0x00000040, 0x00010001 }, + { 0x00010020, 0x34001c00, 0x00001400, 0x00000008 }, { 0x00000201, 0x20080061, 0x00000000, 0x0000e000 }, - { 0x01800031, 0x22001d29, 0x208d0000, 0x0a2a0001 }, - { 0x01800031, 0x21c01d29, 0x208d0000, 0x0a2a0203 }, + { 0x01800031, 0x22001d29, 0x208d0000, 0x0a2a0203 }, { 0x01800031, 0x22401d29, 0x208d0000, 0x0a2a0405 }, + { 0x00000020, 0x34001c00, 0x00001400, 0x00000006 }, + { 0x00000201, 0x20080061, 0x00000000, 0x0000c000 }, + { 0x01800031, 0x22001d29, 0x208d0000, 0x0a4a0203 }, + { 0x00000201, 0x20080061, 0x00000000, 0x0000e000 }, + { 0x01800031, 0x21c01d29, 0x208d0000, 0x0a2a0001 }, diff --git a/shaders/render/exa_wm_yuv_rgb.g4a b/shaders/render/exa_wm_yuv_rgb.g4a index c16037e..5b9e625 100644 --- a/shaders/render/exa_wm_yuv_rgb.g4a +++ b/shaders/render/exa_wm_yuv_rgb.g4a @@ -30,25 +30,25 @@ include(`exa_wm.g4i') define(`YCbCr_base', `src_sample_base') -define(`Cr', `src_sample_r') -define(`Cr_01', `src_sample_r_01') -define(`Cr_23', `src_sample_r_23') +define(`Cr', `src_sample_b') +define(`Cr_01', `src_sample_b_01') +define(`Cr_23', `src_sample_b_23') -define(`Y', `src_sample_g') -define(`Y_01', `src_sample_g_01') -define(`Y_23', `src_sample_g_23') +define(`Y', `src_sample_r') +define(`Y_01', `src_sample_r_01') +define(`Y_23', `src_sample_r_23') -define(`Cb', `src_sample_b') -define(`Cb_01', `src_sample_b_01') -define(`Cb_23', `src_sample_b_23') +define(`Cb', `src_sample_g') +define(`Cb_01', `src_sample_g_01') +define(`Cb_23', `src_sample_g_23') -define(`Crn', `mask_sample_r') -define(`Crn_01', `mask_sample_r_01') -define(`Crn_23', `mask_sample_r_23') +define(`Crn', `mask_sample_g') +define(`Crn_01', `mask_sample_g_01') +define(`Crn_23', `mask_sample_g_23') -define(`Yn', `mask_sample_g') -define(`Yn_01', `mask_sample_g_01') -define(`Yn_23', `mask_sample_g_23') +define(`Yn', `mask_sample_r') +define(`Yn_01', `mask_sample_r_01') +define(`Yn_23', `mask_sample_r_23') define(`Cbn', `mask_sample_b') define(`Cbn_01', `mask_sample_b_01') diff --git a/shaders/render/exa_wm_yuv_rgb.g4b b/shaders/render/exa_wm_yuv_rgb.g4b index ce19e1a..6a76da4 100644 --- a/shaders/render/exa_wm_yuv_rgb.g4b +++ b/shaders/render/exa_wm_yuv_rgb.g4b @@ -1,11 +1,11 @@ - { 0x00802040, 0x23007fbd, 0x008d0200, 0xbd808081 }, - { 0x00802041, 0x23007fbd, 0x008d0300, 0x3f94fdf4 }, - { 0x00802040, 0x22c07fbd, 0x008d01c0, 0xbf008084 }, - { 0x00802040, 0x23407fbd, 0x008d0240, 0xbf008084 }, - { 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, - { 0x80802048, 0x21c07fbd, 0x008d02c0, 0x3fcc49ba }, - { 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, - { 0x00802048, 0x24007fbc, 0x008d02c0, 0xbf5020c5 }, + { 0x00802040, 0x22c07fbd, 0x008d01c0, 0xbd808081 }, + { 0x00802041, 0x22c07fbd, 0x008d02c0, 0x3f94fdf4 }, + { 0x00802040, 0x23007fbd, 0x008d0240, 0xbf008084 }, + { 0x00802040, 0x23407fbd, 0x008d0200, 0xbf008084 }, + { 0x00802001, 0x240003bc, 0x008d02c0, 0x00000000 }, + { 0x80802048, 0x21c07fbd, 0x008d0300, 0x3fcc49ba }, + { 0x00802001, 0x240003bc, 0x008d02c0, 0x00000000 }, + { 0x00802048, 0x24007fbc, 0x008d0300, 0xbf5020c5 }, { 0x80802048, 0x22007fbd, 0x008d0340, 0xbec8b439 }, - { 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, + { 0x00802001, 0x240003bc, 0x008d02c0, 0x00000000 }, { 0x80802048, 0x22407fbd, 0x008d0340, 0x40011687 }, diff --git a/shaders/render/exa_wm_yuv_rgb.g4b.gen5 b/shaders/render/exa_wm_yuv_rgb.g4b.gen5 index ce19e1a..6a76da4 100644 --- a/shaders/render/exa_wm_yuv_rgb.g4b.gen5 +++ b/shaders/render/exa_wm_yuv_rgb.g4b.gen5 @@ -1,11 +1,11 @@ - { 0x00802040, 0x23007fbd, 0x008d0200, 0xbd808081 }, - { 0x00802041, 0x23007fbd, 0x008d0300, 0x3f94fdf4 }, - { 0x00802040, 0x22c07fbd, 0x008d01c0, 0xbf008084 }, - { 0x00802040, 0x23407fbd, 0x008d0240, 0xbf008084 }, - { 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, - { 0x80802048, 0x21c07fbd, 0x008d02c0, 0x3fcc49ba }, - { 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, - { 0x00802048, 0x24007fbc, 0x008d02c0, 0xbf5020c5 }, + { 0x00802040, 0x22c07fbd, 0x008d01c0, 0xbd808081 }, + { 0x00802041, 0x22c07fbd, 0x008d02c0, 0x3f94fdf4 }, + { 0x00802040, 0x23007fbd, 0x008d0240, 0xbf008084 }, + { 0x00802040, 0x23407fbd, 0x008d0200, 0xbf008084 }, + { 0x00802001, 0x240003bc, 0x008d02c0, 0x00000000 }, + { 0x80802048, 0x21c07fbd, 0x008d0300, 0x3fcc49ba }, + { 0x00802001, 0x240003bc, 0x008d02c0, 0x00000000 }, + { 0x00802048, 0x24007fbc, 0x008d0300, 0xbf5020c5 }, { 0x80802048, 0x22007fbd, 0x008d0340, 0xbec8b439 }, - { 0x00802001, 0x240003bc, 0x008d0300, 0x00000000 }, + { 0x00802001, 0x240003bc, 0x008d02c0, 0x00000000 }, { 0x80802048, 0x22407fbd, 0x008d0340, 0x40011687 }, |