/* * 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 * */ #include #include #include #include #include "intel_batchbuffer.h" #include "intel_driver.h" #include "i965_defines.h" #include "i965_drv_video.h" #include "gen6_mfd.h" #define DMV_SIZE 0x88000 /* 557056 bytes for a frame */ static const uint32_t zigzag_direct[64] = { 0, 1, 8, 16, 9, 2, 3, 10, 17, 24, 32, 25, 18, 11, 4, 5, 12, 19, 26, 33, 40, 48, 41, 34, 27, 20, 13, 6, 7, 14, 21, 28, 35, 42, 49, 56, 57, 50, 43, 36, 29, 22, 15, 23, 30, 37, 44, 51, 58, 59, 52, 45, 38, 31, 39, 46, 53, 60, 61, 54, 47, 55, 62, 63 }; static void gen6_mfd_avc_frame_store_index(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, struct gen6_mfd_context *gen6_mfd_context) { struct i965_driver_data *i965 = i965_driver_data(ctx); int i, j; assert(ARRAY_ELEMS(gen6_mfd_context->reference_surface) == ARRAY_ELEMS(pic_param->ReferenceFrames)); for (i = 0; i < ARRAY_ELEMS(gen6_mfd_context->reference_surface); i++) { int found = 0; if (gen6_mfd_context->reference_surface[i].surface_id == VA_INVALID_ID) continue; for (j = 0; j < ARRAY_ELEMS(pic_param->ReferenceFrames); j++) { VAPictureH264 *ref_pic = &pic_param->ReferenceFrames[j]; if (ref_pic->flags & VA_PICTURE_H264_INVALID) continue; if (gen6_mfd_context->reference_surface[i].surface_id == ref_pic->picture_id) { found = 1; break; } } if (!found) { struct object_surface *obj_surface = SURFACE(gen6_mfd_context->reference_surface[i].surface_id); obj_surface->flags &= ~SURFACE_REFERENCED; if ((obj_surface->flags & SURFACE_ALL_MASK) == SURFACE_DISPLAYED) { dri_bo_unreference(obj_surface->bo); obj_surface->bo = NULL; obj_surface->flags &= ~SURFACE_REF_DIS_MASK; } if (obj_surface->free_private_data) obj_surface->free_private_data(&obj_surface->private_data); gen6_mfd_context->reference_surface[i].surface_id = VA_INVALID_ID; gen6_mfd_context->reference_surface[i].frame_store_id = -1; } } for (i = 0; i < ARRAY_ELEMS(pic_param->ReferenceFrames); i++) { VAPictureH264 *ref_pic = &pic_param->ReferenceFrames[i]; int found = 0; if (ref_pic->flags & VA_PICTURE_H264_INVALID) continue; for (j = 0; j < ARRAY_ELEMS(gen6_mfd_context->reference_surface); j++) { if (gen6_mfd_context->reference_surface[j].surface_id == VA_INVALID_ID) continue; if (gen6_mfd_context->reference_surface[j].surface_id == ref_pic->picture_id) { found = 1; break; } } if (!found) { int frame_idx; struct object_surface *obj_surface = SURFACE(ref_pic->picture_id); assert(obj_surface); i965_check_alloc_surface_bo(ctx, obj_surface, 1, VA_FOURCC('N', 'V', '1', '2'), SUBSAMPLE_YUV420); for (frame_idx = 0; frame_idx < ARRAY_ELEMS(gen6_mfd_context->reference_surface); frame_idx++) { for (j = 0; j < ARRAY_ELEMS(gen6_mfd_context->reference_surface); j++) { if (gen6_mfd_context->reference_surface[j].surface_id == VA_INVALID_ID) continue; if (gen6_mfd_context->reference_surface[j].frame_store_id == frame_idx) break; } if (j == ARRAY_ELEMS(gen6_mfd_context->reference_surface)) break; } assert(frame_idx < ARRAY_ELEMS(gen6_mfd_context->reference_surface)); for (j = 0; j < ARRAY_ELEMS(gen6_mfd_context->reference_surface); j++) { if (gen6_mfd_context->reference_surface[j].surface_id == VA_INVALID_ID) { gen6_mfd_context->reference_surface[j].surface_id = ref_pic->picture_id; gen6_mfd_context->reference_surface[j].frame_store_id = frame_idx; break; } } } } /* sort */ for (i = 0; i < ARRAY_ELEMS(gen6_mfd_context->reference_surface) - 1; i++) { if (gen6_mfd_context->reference_surface[i].surface_id != VA_INVALID_ID && gen6_mfd_context->reference_surface[i].frame_store_id == i) continue; for (j = i + 1; j < ARRAY_ELEMS(gen6_mfd_context->reference_surface); j++) { if (gen6_mfd_context->reference_surface[j].surface_id != VA_INVALID_ID && gen6_mfd_context->reference_surface[j].frame_store_id == i) { VASurfaceID id = gen6_mfd_context->reference_surface[i].surface_id; int frame_idx = gen6_mfd_context->reference_surface[i].frame_store_id; gen6_mfd_context->reference_surface[i].surface_id = gen6_mfd_context->reference_surface[j].surface_id; gen6_mfd_context->reference_surface[i].frame_store_id = gen6_mfd_context->reference_surface[j].frame_store_id; gen6_mfd_context->reference_surface[j].surface_id = id; gen6_mfd_context->reference_surface[j].frame_store_id = frame_idx; break; } } } } static void gen6_mfd_free_avc_surface(void **data) { struct gen6_avc_surface *gen6_avc_surface = *data; if (!gen6_avc_surface) return; dri_bo_unreference(gen6_avc_surface->dmv_top); gen6_avc_surface->dmv_top = NULL; dri_bo_unreference(gen6_avc_surface->dmv_bottom); gen6_avc_surface->dmv_bottom = NULL; free(gen6_avc_surface); *data = NULL; } static void gen6_mfd_init_avc_surface(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, struct object_surface *obj_surface) { struct i965_driver_data *i965 = i965_driver_data(ctx); struct gen6_avc_surface *gen6_avc_surface = obj_surface->private_data; obj_surface->free_private_data = gen6_mfd_free_avc_surface; if (!gen6_avc_surface) { gen6_avc_surface = calloc(sizeof(struct gen6_avc_surface), 1); assert((obj_surface->size & 0x3f) == 0); obj_surface->private_data = gen6_avc_surface; } gen6_avc_surface->dmv_bottom_flag = (pic_param->pic_fields.bits.field_pic_flag && !pic_param->seq_fields.bits.direct_8x8_inference_flag); if (gen6_avc_surface->dmv_top == NULL) { gen6_avc_surface->dmv_top = dri_bo_alloc(i965->intel.bufmgr, "direct mv w/r buffer", DMV_SIZE, 0x1000); } if (gen6_avc_surface->dmv_bottom_flag && gen6_avc_surface->dmv_bottom == NULL) { gen6_avc_surface->dmv_bottom = dri_bo_alloc(i965->intel.bufmgr, "direct mv w/r buffer", DMV_SIZE, 0x1000); } } static void gen6_mfd_pipe_mode_select(VADriverContextP ctx, struct decode_state *decode_state, int standard_select, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; assert(standard_select == MFX_FORMAT_MPEG2 || standard_select == MFX_FORMAT_AVC || standard_select == MFX_FORMAT_VC1); BEGIN_BCS_BATCH(batch, 4); OUT_BCS_BATCH(batch, MFX_PIPE_MODE_SELECT | (4 - 2)); OUT_BCS_BATCH(batch, (MFD_MODE_VLD << 16) | /* VLD mode */ (0 << 10) | /* disable Stream-Out */ (gen6_mfd_context->post_deblocking_output.valid << 9) | /* Post Deblocking Output */ (gen6_mfd_context->pre_deblocking_output.valid << 8) | /* Pre Deblocking Output */ (0 << 7) | /* disable TLB prefectch */ (0 << 5) | /* not in stitch mode */ (MFX_CODEC_DECODE << 4) | /* decoding mode */ (standard_select << 0)); OUT_BCS_BATCH(batch, (0 << 20) | /* round flag in PB slice */ (0 << 19) | /* round flag in Intra8x8 */ (0 << 7) | /* expand NOA bus flag */ (1 << 6) | /* must be 1 */ (0 << 5) | /* disable clock gating for NOA */ (0 << 4) | /* terminate if AVC motion and POC table error occurs */ (0 << 3) | /* terminate if AVC mbdata error occurs */ (0 << 2) | /* terminate if AVC CABAC/CAVLC decode error occurs */ (0 << 1) | /* AVC long field motion vector */ (1 << 0)); /* always calculate AVC ILDB boundary strength */ OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_surface_state(VADriverContextP ctx, struct decode_state *decode_state, int standard_select, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; struct i965_driver_data *i965 = i965_driver_data(ctx); struct object_surface *obj_surface = SURFACE(decode_state->current_render_target); assert(obj_surface); BEGIN_BCS_BATCH(batch, 6); OUT_BCS_BATCH(batch, MFX_SURFACE_STATE | (6 - 2)); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, ((obj_surface->orig_height - 1) << 19) | ((obj_surface->orig_width - 1) << 6)); OUT_BCS_BATCH(batch, (MFX_SURFACE_PLANAR_420_8 << 28) | /* 420 planar YUV surface */ (1 << 27) | /* must be 1 for interleave U/V, hardware requirement */ (0 << 22) | /* surface object control state, FIXME??? */ ((obj_surface->width - 1) << 3) | /* pitch */ (0 << 2) | /* must be 0 for interleave U/V */ (1 << 1) | /* must be y-tiled */ (I965_TILEWALK_YMAJOR << 0)); /* tile walk, FIXME: must be 1 ??? */ OUT_BCS_BATCH(batch, (0 << 16) | /* must be 0 for interleave U/V */ (obj_surface->height)); /* y offset for U(cb) */ OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_pipe_buf_addr_state(VADriverContextP ctx, struct decode_state *decode_state, int standard_select, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; struct i965_driver_data *i965 = i965_driver_data(ctx); int i; BEGIN_BCS_BATCH(batch, 24); OUT_BCS_BATCH(batch, MFX_PIPE_BUF_ADDR_STATE | (24 - 2)); if (gen6_mfd_context->pre_deblocking_output.valid) OUT_BCS_RELOC(batch, gen6_mfd_context->pre_deblocking_output.bo, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); else OUT_BCS_BATCH(batch, 0); if (gen6_mfd_context->post_deblocking_output.valid) OUT_BCS_RELOC(batch, gen6_mfd_context->post_deblocking_output.bo, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); else OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); /* ignore for decoding */ OUT_BCS_BATCH(batch, 0); /* ignore for decoding */ if (gen6_mfd_context->intra_row_store_scratch_buffer.valid) OUT_BCS_RELOC(batch, gen6_mfd_context->intra_row_store_scratch_buffer.bo, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); else OUT_BCS_BATCH(batch, 0); if (gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.valid) OUT_BCS_RELOC(batch, gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.bo, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); else OUT_BCS_BATCH(batch, 0); /* DW 7..22 */ for (i = 0; i < ARRAY_ELEMS(gen6_mfd_context->reference_surface); i++) { struct object_surface *obj_surface; if (gen6_mfd_context->reference_surface[i].surface_id != VA_INVALID_ID) { obj_surface = SURFACE(gen6_mfd_context->reference_surface[i].surface_id); assert(obj_surface && obj_surface->bo); OUT_BCS_RELOC(batch, obj_surface->bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); } else { OUT_BCS_BATCH(batch, 0); } } OUT_BCS_BATCH(batch, 0); /* ignore DW23 for decoding */ ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_ind_obj_base_addr_state(VADriverContextP ctx, dri_bo *slice_data_bo, int standard_select, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; BEGIN_BCS_BATCH(batch, 11); OUT_BCS_BATCH(batch, MFX_IND_OBJ_BASE_ADDR_STATE | (11 - 2)); OUT_BCS_RELOC(batch, slice_data_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); /* MFX Indirect Bitstream Object Base Address */ OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); /* ignore for VLD mode */ OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); /* ignore for VLD mode */ OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); /* ignore for VLD mode */ OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); /* ignore for VLD mode */ OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_bsp_buf_base_addr_state(VADriverContextP ctx, struct decode_state *decode_state, int standard_select, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; BEGIN_BCS_BATCH(batch, 4); OUT_BCS_BATCH(batch, MFX_BSP_BUF_BASE_ADDR_STATE | (4 - 2)); if (gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.valid) OUT_BCS_RELOC(batch, gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); else OUT_BCS_BATCH(batch, 0); if (gen6_mfd_context->mpr_row_store_scratch_buffer.valid) OUT_BCS_RELOC(batch, gen6_mfd_context->mpr_row_store_scratch_buffer.bo, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); else OUT_BCS_BATCH(batch, 0); if (gen6_mfd_context->bitplane_read_buffer.valid) OUT_BCS_RELOC(batch, gen6_mfd_context->bitplane_read_buffer.bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); else OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } #if 0 static void gen6_mfd_aes_state(VADriverContextP ctx, struct decode_state *decode_state, int standard_select) { /* FIXME */ } static void gen6_mfd_wait(VADriverContextP ctx, struct decode_state *decode_state, int standard_select, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; BEGIN_BCS_BATCH(batch, 1); OUT_BCS_BATCH(batch, MFX_WAIT | (1 << 8)); ADVANCE_BCS_BATCH(batch); } #endif static void gen6_mfd_avc_img_state(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; int qm_present_flag; int img_struct; int mbaff_frame_flag; unsigned int width_in_mbs, height_in_mbs; VAPictureParameterBufferH264 *pic_param; 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 (decode_state->iq_matrix && decode_state->iq_matrix->buffer) qm_present_flag = 1; else qm_present_flag = 0; /* built-in QM matrices */ 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 */ /* MFX 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 */ BEGIN_BCS_BATCH(batch, 13); OUT_BCS_BATCH(batch, MFX_AVC_IMG_STATE | (13 - 2)); OUT_BCS_BATCH(batch, ((width_in_mbs * height_in_mbs) & 0x7fff)); OUT_BCS_BATCH(batch, (height_in_mbs << 16) | (width_in_mbs << 0)); OUT_BCS_BATCH(batch, ((pic_param->second_chroma_qp_index_offset & 0x1f) << 24) | ((pic_param->chroma_qp_index_offset & 0x1f) << 16) | (0 << 14) | /* Max-bit conformance Intra flag ??? FIXME */ (0 << 13) | /* Max Macroblock size conformance Inter flag ??? FIXME */ (1 << 12) | /* always 1, hardware requirement */ (qm_present_flag << 10) | (img_struct << 8) | (16 << 0)); OUT_BCS_BATCH(batch, (pic_param->seq_fields.bits.chroma_format_idc << 10) | (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(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_avc_qm_state(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; 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(batch, cmd_len); OUT_BCS_BATCH(batch, MFX_AVC_QM_STATE | (cmd_len - 2)); if (pic_param->pic_fields.bits.transform_8x8_mode_flag) OUT_BCS_BATCH(batch, (0x0 << 8) | /* don't use default built-in matrices */ (0xff << 0)); /* six 4x4 and two 8x8 scaling matrices */ else OUT_BCS_BATCH(batch, (0x0 << 8) | /* don't use default built-in matrices */ (0x3f << 0)); /* six 4x4 scaling matrices */ intel_batchbuffer_data(batch, &iq_matrix->ScalingList4x4[0][0], 6 * 4 * 4); if (pic_param->pic_fields.bits.transform_8x8_mode_flag) intel_batchbuffer_data(batch, &iq_matrix->ScalingList8x8[0][0], 2 * 16 * 4); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_avc_directmode_state(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, VASliceParameterBufferH264 *slice_param, struct gen6_mfd_context *gen6_mfd_context) { struct i965_driver_data *i965 = i965_driver_data(ctx); struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; struct object_surface *obj_surface; struct gen6_avc_surface *gen6_avc_surface; VAPictureH264 *va_pic; int i, j; BEGIN_BCS_BATCH(batch, 69); OUT_BCS_BATCH(batch, MFX_AVC_DIRECTMODE_STATE | (69 - 2)); /* reference surfaces 0..15 */ for (i = 0; i < ARRAY_ELEMS(gen6_mfd_context->reference_surface); i++) { if (gen6_mfd_context->reference_surface[i].surface_id != VA_INVALID_ID) { obj_surface = SURFACE(gen6_mfd_context->reference_surface[i].surface_id); assert(obj_surface); gen6_avc_surface = obj_surface->private_data; if (gen6_avc_surface == NULL) { OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); } else { OUT_BCS_RELOC(batch, gen6_avc_surface->dmv_top, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); if (gen6_avc_surface->dmv_bottom_flag == 1) OUT_BCS_RELOC(batch, gen6_avc_surface->dmv_bottom, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); else OUT_BCS_RELOC(batch, gen6_avc_surface->dmv_top, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); } } else { OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); } } /* the current decoding frame/field */ va_pic = &pic_param->CurrPic; assert(!(va_pic->flags & VA_PICTURE_H264_INVALID)); obj_surface = SURFACE(va_pic->picture_id); assert(obj_surface && obj_surface->bo && obj_surface->private_data); gen6_avc_surface = obj_surface->private_data; OUT_BCS_RELOC(batch, gen6_avc_surface->dmv_top, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); if (gen6_avc_surface->dmv_bottom_flag == 1) OUT_BCS_RELOC(batch, gen6_avc_surface->dmv_bottom, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); else OUT_BCS_RELOC(batch, gen6_avc_surface->dmv_top, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); /* POC List */ for (i = 0; i < ARRAY_ELEMS(gen6_mfd_context->reference_surface); i++) { if (gen6_mfd_context->reference_surface[i].surface_id != VA_INVALID_ID) { int found = 0; for (j = 0; j < ARRAY_ELEMS(pic_param->ReferenceFrames); j++) { va_pic = &pic_param->ReferenceFrames[j]; if (va_pic->flags & VA_PICTURE_H264_INVALID) continue; if (va_pic->picture_id == gen6_mfd_context->reference_surface[i].surface_id) { found = 1; break; } } assert(found == 1); assert(!(va_pic->flags & VA_PICTURE_H264_INVALID)); OUT_BCS_BATCH(batch, va_pic->TopFieldOrderCnt); OUT_BCS_BATCH(batch, va_pic->BottomFieldOrderCnt); } else { OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); } } va_pic = &pic_param->CurrPic; OUT_BCS_BATCH(batch, va_pic->TopFieldOrderCnt); OUT_BCS_BATCH(batch, va_pic->BottomFieldOrderCnt); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_avc_slice_state(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, VASliceParameterBufferH264 *slice_param, VASliceParameterBufferH264 *next_slice_param, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; int width_in_mbs = pic_param->picture_width_in_mbs_minus1 + 1; int height_in_mbs = pic_param->picture_height_in_mbs_minus1 + 1; int slice_hor_pos, slice_ver_pos, next_slice_hor_pos, next_slice_ver_pos; int num_ref_idx_l0, num_ref_idx_l1; int mbaff_picture = (!pic_param->pic_fields.bits.field_pic_flag && pic_param->seq_fields.bits.mb_adaptive_frame_field_flag); int weighted_pred_idc = 0; int first_mb_in_slice = 0, first_mb_in_next_slice = 0; int slice_type; if (slice_param->slice_type == SLICE_TYPE_I || slice_param->slice_type == SLICE_TYPE_SI) { slice_type = SLICE_TYPE_I; } else if (slice_param->slice_type == SLICE_TYPE_P || slice_param->slice_type == SLICE_TYPE_SP) { slice_type = SLICE_TYPE_P; } else { assert(slice_param->slice_type == SLICE_TYPE_B); slice_type = SLICE_TYPE_B; } if (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_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; weighted_pred_idc = (pic_param->pic_fields.bits.weighted_pred_flag == 1); } 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; weighted_pred_idc = (pic_param->pic_fields.bits.weighted_bipred_idc == 1); } first_mb_in_slice = slice_param->first_mb_in_slice << mbaff_picture; slice_hor_pos = first_mb_in_slice % width_in_mbs; slice_ver_pos = first_mb_in_slice / width_in_mbs; if (next_slice_param) { first_mb_in_next_slice = next_slice_param->first_mb_in_slice << mbaff_picture; next_slice_hor_pos = first_mb_in_next_slice % width_in_mbs; next_slice_ver_pos = first_mb_in_next_slice / width_in_mbs; } else { next_slice_hor_pos = 0; next_slice_ver_pos = height_in_mbs; } BEGIN_BCS_BATCH(batch, 11); /* FIXME: is it 10??? */ OUT_BCS_BATCH(batch, MFX_AVC_SLICE_STATE | (11 - 2)); OUT_BCS_BATCH(batch, slice_type); OUT_BCS_BATCH(batch, (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(batch, (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(batch, (slice_ver_pos << 24) | (slice_hor_pos << 16) | (first_mb_in_slice << 0)); OUT_BCS_BATCH(batch, (next_slice_ver_pos << 16) | (next_slice_hor_pos << 0)); OUT_BCS_BATCH(batch, (next_slice_param == NULL) << 19); /* last slice flag */ OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_avc_phantom_slice_state(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; 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 */ BEGIN_BCS_BATCH(batch, 11); /* FIXME: is it 10??? */ OUT_BCS_BATCH(batch, MFX_AVC_SLICE_STATE | (11 - 2)); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, height_in_mbs << 24 | width_in_mbs * height_in_mbs / (1 + !!pic_param->pic_fields.bits.field_pic_flag)); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_avc_ref_idx_state(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, VASliceParameterBufferH264 *slice_param, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; int i, j, num_ref_list; 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]; if (slice_param->slice_type == SLICE_TYPE_I || slice_param->slice_type == SLICE_TYPE_SI) return; if (slice_param->slice_type == SLICE_TYPE_P || slice_param->slice_type == SLICE_TYPE_SP) { num_ref_list = 1; } else { num_ref_list = 2; } for (i = 0; i < num_ref_list; i++) { VAPictureH264 *va_pic; if (i == 0) { va_pic = slice_param->RefPicList0; } else { va_pic = slice_param->RefPicList1; } BEGIN_BCS_BATCH(batch, 10); OUT_BCS_BATCH(batch, MFX_AVC_REF_IDX_STATE | (10 - 2)); OUT_BCS_BATCH(batch, i); 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 < ARRAY_ELEMS(gen6_mfd_context->reference_surface); frame_idx++) { if (gen6_mfd_context->reference_surface[frame_idx].surface_id != VA_INVALID_ID && va_pic->picture_id == gen6_mfd_context->reference_surface[frame_idx].surface_id) { assert(frame_idx == gen6_mfd_context->reference_surface[frame_idx].frame_store_id); break; } } assert(frame_idx < ARRAY_ELEMS(gen6_mfd_context->reference_surface)); 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(batch, refs, sizeof(refs)); ADVANCE_BCS_BATCH(batch); } } static void gen6_mfd_avc_weightoffset_state(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, VASliceParameterBufferH264 *slice_param, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; int i, j, num_weight_offset_table = 0; short weightoffsets[32 * 6]; if ((slice_param->slice_type == SLICE_TYPE_P || slice_param->slice_type == SLICE_TYPE_SP) && (pic_param->pic_fields.bits.weighted_pred_flag == 1)) { num_weight_offset_table = 1; } if ((slice_param->slice_type == SLICE_TYPE_B) && (pic_param->pic_fields.bits.weighted_bipred_idc == 1)) { num_weight_offset_table = 2; } for (i = 0; i < num_weight_offset_table; i++) { BEGIN_BCS_BATCH(batch, 98); OUT_BCS_BATCH(batch, MFX_AVC_WEIGHTOFFSET_STATE | (98 - 2)); OUT_BCS_BATCH(batch, i); if (i == 0) { for (j = 0; j < 32; j++) { weightoffsets[j * 6 + 0] = slice_param->luma_weight_l0[j]; weightoffsets[j * 6 + 1] = slice_param->luma_offset_l0[j]; weightoffsets[j * 6 + 2] = slice_param->chroma_weight_l0[j][0]; weightoffsets[j * 6 + 3] = slice_param->chroma_offset_l0[j][0]; weightoffsets[j * 6 + 4] = slice_param->chroma_weight_l0[j][1]; weightoffsets[j * 6 + 5] = slice_param->chroma_offset_l0[j][1]; } } else { for (j = 0; j < 32; j++) { weightoffsets[j * 6 + 0] = slice_param->luma_weight_l1[j]; weightoffsets[j * 6 + 1] = slice_param->luma_offset_l1[j]; weightoffsets[j * 6 + 2] = slice_param->chroma_weight_l1[j][0]; weightoffsets[j * 6 + 3] = slice_param->chroma_offset_l1[j][0]; weightoffsets[j * 6 + 4] = slice_param->chroma_weight_l1[j][1]; weightoffsets[j * 6 + 5] = slice_param->chroma_offset_l1[j][1]; } } intel_batchbuffer_data(batch, weightoffsets, sizeof(weightoffsets)); ADVANCE_BCS_BATCH(batch); } } static int gen6_mfd_avc_get_slice_bit_offset(uint8_t *buf, int mode_flag, int in_slice_data_bit_offset) { int out_slice_data_bit_offset; int slice_header_size = in_slice_data_bit_offset / 8; int i, j; for (i = 0, j = 0; i < slice_header_size; i++, j++) { if (!buf[j] && !buf[j + 1] && buf[j + 2] == 3) { i++, j += 2; } } out_slice_data_bit_offset = 8 * j + in_slice_data_bit_offset % 8; if (mode_flag == ENTROPY_CABAC) out_slice_data_bit_offset = ALIGN(out_slice_data_bit_offset, 0x8); return out_slice_data_bit_offset; } static void gen6_mfd_avc_bsd_object(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, VASliceParameterBufferH264 *slice_param, dri_bo *slice_data_bo, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; int slice_data_bit_offset; uint8_t *slice_data = NULL; dri_bo_map(slice_data_bo, 0); slice_data = (uint8_t *)(slice_data_bo->virtual + slice_param->slice_data_offset); slice_data_bit_offset = gen6_mfd_avc_get_slice_bit_offset(slice_data, pic_param->pic_fields.bits.entropy_coding_mode_flag, slice_param->slice_data_bit_offset); dri_bo_unmap(slice_data_bo); BEGIN_BCS_BATCH(batch, 6); OUT_BCS_BATCH(batch, MFD_AVC_BSD_OBJECT | (6 - 2)); OUT_BCS_BATCH(batch, ((slice_param->slice_data_size - (slice_data_bit_offset >> 3)) << 0)); OUT_BCS_BATCH(batch, slice_param->slice_data_offset + (slice_data_bit_offset >> 3)); OUT_BCS_BATCH(batch, (0 << 31) | (0 << 14) | (0 << 12) | (0 << 10) | (0 << 8)); OUT_BCS_BATCH(batch, (0 << 16) | (0 << 6) | ((0x7 - (slice_data_bit_offset & 0x7)) << 0)); OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_avc_phantom_slice_bsd_object(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; BEGIN_BCS_BATCH(batch, 6); OUT_BCS_BATCH(batch, MFD_AVC_BSD_OBJECT | (6 - 2)); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_avc_phantom_slice(VADriverContextP ctx, VAPictureParameterBufferH264 *pic_param, struct gen6_mfd_context *gen6_mfd_context) { gen6_mfd_avc_phantom_slice_state(ctx, pic_param, gen6_mfd_context); gen6_mfd_avc_phantom_slice_bsd_object(ctx, pic_param, gen6_mfd_context); } static void gen6_mfd_avc_decode_init(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { VAPictureParameterBufferH264 *pic_param; VASliceParameterBufferH264 *slice_param; VAPictureH264 *va_pic; struct i965_driver_data *i965 = i965_driver_data(ctx); struct object_surface *obj_surface; dri_bo *bo; int i, j, enable_avc_ildb = 0; for (j = 0; j < decode_state->num_slice_params && enable_avc_ildb == 0; j++) { assert(decode_state->slice_params && decode_state->slice_params[j]->buffer); slice_param = (VASliceParameterBufferH264 *)decode_state->slice_params[j]->buffer; for (i = 0; i < decode_state->slice_params[j]->num_elements; 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_SI) || (slice_param->slice_type == SLICE_TYPE_P) || (slice_param->slice_type == SLICE_TYPE_SP) || (slice_param->slice_type == SLICE_TYPE_B)); if (slice_param->disable_deblocking_filter_idc != 1) { enable_avc_ildb = 1; break; } slice_param++; } } assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; gen6_mfd_avc_frame_store_index(ctx, pic_param, gen6_mfd_context); /* Current decoded picture */ va_pic = &pic_param->CurrPic; assert(!(va_pic->flags & VA_PICTURE_H264_INVALID)); obj_surface = SURFACE(va_pic->picture_id); assert(obj_surface); obj_surface->flags &= ~SURFACE_REF_DIS_MASK; obj_surface->flags |= (pic_param->pic_fields.bits.reference_pic_flag ? SURFACE_REFERENCED : 0); i965_check_alloc_surface_bo(ctx, obj_surface, 1, VA_FOURCC('N','V','1','2'), SUBSAMPLE_YUV420); gen6_mfd_init_avc_surface(ctx, pic_param, obj_surface); dri_bo_unreference(gen6_mfd_context->post_deblocking_output.bo); gen6_mfd_context->post_deblocking_output.bo = obj_surface->bo; dri_bo_reference(gen6_mfd_context->post_deblocking_output.bo); gen6_mfd_context->post_deblocking_output.valid = enable_avc_ildb; dri_bo_unreference(gen6_mfd_context->pre_deblocking_output.bo); gen6_mfd_context->pre_deblocking_output.bo = obj_surface->bo; dri_bo_reference(gen6_mfd_context->pre_deblocking_output.bo); gen6_mfd_context->pre_deblocking_output.valid = !enable_avc_ildb; dri_bo_unreference(gen6_mfd_context->intra_row_store_scratch_buffer.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "intra row store", 128 * 64, 0x1000); assert(bo); gen6_mfd_context->intra_row_store_scratch_buffer.bo = bo; gen6_mfd_context->intra_row_store_scratch_buffer.valid = 1; dri_bo_unreference(gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "deblocking filter row store", 30720, /* 4 * 120 * 64 */ 0x1000); assert(bo); gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.bo = bo; gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.valid = 1; dri_bo_unreference(gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "bsd mpc row store", 11520, /* 1.5 * 120 * 64 */ 0x1000); assert(bo); gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo = bo; gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.valid = 1; dri_bo_unreference(gen6_mfd_context->mpr_row_store_scratch_buffer.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "mpr row store", 7680, /* 1. 0 * 120 * 64 */ 0x1000); assert(bo); gen6_mfd_context->mpr_row_store_scratch_buffer.bo = bo; gen6_mfd_context->mpr_row_store_scratch_buffer.valid = 1; gen6_mfd_context->bitplane_read_buffer.valid = 0; } static void gen6_mfd_avc_decode_picture(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; VAPictureParameterBufferH264 *pic_param; VASliceParameterBufferH264 *slice_param, *next_slice_param, *next_slice_group_param; dri_bo *slice_data_bo; int i, j; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferH264 *)decode_state->pic_param->buffer; gen6_mfd_avc_decode_init(ctx, decode_state, gen6_mfd_context); intel_batchbuffer_start_atomic_bcs(batch, 0x1000); intel_batchbuffer_emit_mi_flush(batch); gen6_mfd_pipe_mode_select(ctx, decode_state, MFX_FORMAT_AVC, gen6_mfd_context); gen6_mfd_surface_state(ctx, decode_state, MFX_FORMAT_AVC, gen6_mfd_context); gen6_mfd_pipe_buf_addr_state(ctx, decode_state, MFX_FORMAT_AVC, gen6_mfd_context); gen6_mfd_bsp_buf_base_addr_state(ctx, decode_state, MFX_FORMAT_AVC, gen6_mfd_context); gen6_mfd_avc_img_state(ctx, decode_state, gen6_mfd_context); gen6_mfd_avc_qm_state(ctx, decode_state, gen6_mfd_context); for (j = 0; j < decode_state->num_slice_params; j++) { assert(decode_state->slice_params && decode_state->slice_params[j]->buffer); slice_param = (VASliceParameterBufferH264 *)decode_state->slice_params[j]->buffer; slice_data_bo = decode_state->slice_datas[j]->bo; gen6_mfd_ind_obj_base_addr_state(ctx, slice_data_bo, MFX_FORMAT_AVC, gen6_mfd_context); if (j == decode_state->num_slice_params - 1) next_slice_group_param = NULL; else next_slice_group_param = (VASliceParameterBufferH264 *)decode_state->slice_params[j + 1]->buffer; for (i = 0; i < decode_state->slice_params[j]->num_elements; 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_SI) || (slice_param->slice_type == SLICE_TYPE_P) || (slice_param->slice_type == SLICE_TYPE_SP) || (slice_param->slice_type == SLICE_TYPE_B)); if (i < decode_state->slice_params[j]->num_elements - 1) next_slice_param = slice_param + 1; else next_slice_param = next_slice_group_param; gen6_mfd_avc_directmode_state(ctx, pic_param, slice_param, gen6_mfd_context); gen6_mfd_avc_slice_state(ctx, pic_param, slice_param, next_slice_param, gen6_mfd_context); gen6_mfd_avc_ref_idx_state(ctx, pic_param, slice_param, gen6_mfd_context); gen6_mfd_avc_weightoffset_state(ctx, pic_param, slice_param, gen6_mfd_context); gen6_mfd_avc_bsd_object(ctx, pic_param, slice_param, slice_data_bo, gen6_mfd_context); slice_param++; } } gen6_mfd_avc_phantom_slice(ctx, pic_param, gen6_mfd_context); intel_batchbuffer_end_atomic(batch); intel_batchbuffer_flush(batch); } static void gen6_mfd_mpeg2_decode_init(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { VAPictureParameterBufferMPEG2 *pic_param; struct i965_driver_data *i965 = i965_driver_data(ctx); struct object_surface *obj_surface; int i; dri_bo *bo; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferMPEG2 *)decode_state->pic_param->buffer; /* reference picture */ obj_surface = SURFACE(pic_param->forward_reference_picture); if (obj_surface && obj_surface->bo) gen6_mfd_context->reference_surface[0].surface_id = pic_param->forward_reference_picture; else gen6_mfd_context->reference_surface[0].surface_id = VA_INVALID_ID; obj_surface = SURFACE(pic_param->backward_reference_picture); if (obj_surface && obj_surface->bo) gen6_mfd_context->reference_surface[1].surface_id = pic_param->backward_reference_picture; else gen6_mfd_context->reference_surface[1].surface_id = gen6_mfd_context->reference_surface[0].surface_id; /* must do so !!! */ for (i = 2; i < ARRAY_ELEMS(gen6_mfd_context->reference_surface); i++) gen6_mfd_context->reference_surface[i].surface_id = gen6_mfd_context->reference_surface[i % 2].surface_id; /* Current decoded picture */ obj_surface = SURFACE(decode_state->current_render_target); assert(obj_surface); i965_check_alloc_surface_bo(ctx, obj_surface, 1, VA_FOURCC('N','V','1','2'), SUBSAMPLE_YUV420); dri_bo_unreference(gen6_mfd_context->pre_deblocking_output.bo); gen6_mfd_context->pre_deblocking_output.bo = obj_surface->bo; dri_bo_reference(gen6_mfd_context->pre_deblocking_output.bo); gen6_mfd_context->pre_deblocking_output.valid = 1; dri_bo_unreference(gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "bsd mpc row store", 11520, /* 1.5 * 120 * 64 */ 0x1000); assert(bo); gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo = bo; gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.valid = 1; gen6_mfd_context->post_deblocking_output.valid = 0; gen6_mfd_context->intra_row_store_scratch_buffer.valid = 0; gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.valid = 0; gen6_mfd_context->mpr_row_store_scratch_buffer.valid = 0; gen6_mfd_context->bitplane_read_buffer.valid = 0; } static void gen6_mfd_mpeg2_pic_state(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; VAPictureParameterBufferMPEG2 *pic_param; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferMPEG2 *)decode_state->pic_param->buffer; BEGIN_BCS_BATCH(batch, 4); OUT_BCS_BATCH(batch, MFX_MPEG2_PIC_STATE | (4 - 2)); OUT_BCS_BATCH(batch, (pic_param->f_code & 0xf) << 28 | /* f_code[1][1] */ ((pic_param->f_code >> 4) & 0xf) << 24 | /* f_code[1][0] */ ((pic_param->f_code >> 8) & 0xf) << 20 | /* f_code[0][1] */ ((pic_param->f_code >> 12) & 0xf) << 16 | /* f_code[0][0] */ pic_param->picture_coding_extension.bits.intra_dc_precision << 14 | pic_param->picture_coding_extension.bits.picture_structure << 12 | pic_param->picture_coding_extension.bits.top_field_first << 11 | pic_param->picture_coding_extension.bits.frame_pred_frame_dct << 10 | pic_param->picture_coding_extension.bits.concealment_motion_vectors << 9 | pic_param->picture_coding_extension.bits.q_scale_type << 8 | pic_param->picture_coding_extension.bits.intra_vlc_format << 7 | pic_param->picture_coding_extension.bits.alternate_scan << 6); OUT_BCS_BATCH(batch, pic_param->picture_coding_type << 9); OUT_BCS_BATCH(batch, (ALIGN(pic_param->vertical_size, 16) / 16) << 16 | (ALIGN(pic_param->horizontal_size, 16) / 16)); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_mpeg2_qm_state(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; VAIQMatrixBufferMPEG2 * const gen_iq_matrix = &gen6_mfd_context->iq_matrix.mpeg2; int i, j; /* Update internal QM state */ if (decode_state->iq_matrix && decode_state->iq_matrix->buffer) { VAIQMatrixBufferMPEG2 * const iq_matrix = (VAIQMatrixBufferMPEG2 *)decode_state->iq_matrix->buffer; gen_iq_matrix->load_intra_quantiser_matrix = iq_matrix->load_intra_quantiser_matrix; if (iq_matrix->load_intra_quantiser_matrix) { for (j = 0; j < 64; j++) gen_iq_matrix->intra_quantiser_matrix[zigzag_direct[j]] = iq_matrix->intra_quantiser_matrix[j]; } gen_iq_matrix->load_non_intra_quantiser_matrix = iq_matrix->load_non_intra_quantiser_matrix; if (iq_matrix->load_non_intra_quantiser_matrix) { for (j = 0; j < 64; j++) gen_iq_matrix->non_intra_quantiser_matrix[zigzag_direct[j]] = iq_matrix->non_intra_quantiser_matrix[j]; } } /* Commit QM state to HW */ for (i = 0; i < 2; i++) { unsigned char *qm = NULL; if (i == 0) { if (gen_iq_matrix->load_intra_quantiser_matrix) qm = gen_iq_matrix->intra_quantiser_matrix; } else { if (gen_iq_matrix->load_non_intra_quantiser_matrix) qm = gen_iq_matrix->non_intra_quantiser_matrix; } if (!qm) continue; BEGIN_BCS_BATCH(batch, 18); OUT_BCS_BATCH(batch, MFX_MPEG2_QM_STATE | (18 - 2)); OUT_BCS_BATCH(batch, i); intel_batchbuffer_data(batch, qm, 64); ADVANCE_BCS_BATCH(batch); } } static void gen6_mfd_mpeg2_bsd_object(VADriverContextP ctx, VAPictureParameterBufferMPEG2 *pic_param, VASliceParameterBufferMPEG2 *slice_param, VASliceParameterBufferMPEG2 *next_slice_param, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; unsigned int width_in_mbs = ALIGN(pic_param->horizontal_size, 16) / 16; int mb_count, vpos0, hpos0, vpos1, hpos1, is_field_pic = 0; if (pic_param->picture_coding_extension.bits.picture_structure == MPEG_TOP_FIELD || pic_param->picture_coding_extension.bits.picture_structure == MPEG_BOTTOM_FIELD) is_field_pic = 1; vpos0 = slice_param->slice_vertical_position / (1 + is_field_pic); hpos0 = slice_param->slice_horizontal_position; if (next_slice_param == NULL) { vpos1 = ALIGN(pic_param->vertical_size, 16) / 16 / (1 + is_field_pic); hpos1 = 0; } else { vpos1 = next_slice_param->slice_vertical_position / (1 + is_field_pic); hpos1 = next_slice_param->slice_horizontal_position; } mb_count = (vpos1 * width_in_mbs + hpos1) - (vpos0 * width_in_mbs + hpos0); BEGIN_BCS_BATCH(batch, 5); OUT_BCS_BATCH(batch, MFD_MPEG2_BSD_OBJECT | (5 - 2)); OUT_BCS_BATCH(batch, slice_param->slice_data_size - (slice_param->macroblock_offset >> 3)); OUT_BCS_BATCH(batch, slice_param->slice_data_offset + (slice_param->macroblock_offset >> 3)); OUT_BCS_BATCH(batch, hpos0 << 24 | vpos0 << 16 | mb_count << 8 | (next_slice_param == NULL) << 5 | (next_slice_param == NULL) << 3 | (slice_param->macroblock_offset & 0x7)); OUT_BCS_BATCH(batch, slice_param->quantiser_scale_code << 24); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_mpeg2_decode_picture(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; VAPictureParameterBufferMPEG2 *pic_param; VASliceParameterBufferMPEG2 *slice_param, *next_slice_param, *next_slice_group_param; dri_bo *slice_data_bo; int i, j; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferMPEG2 *)decode_state->pic_param->buffer; gen6_mfd_mpeg2_decode_init(ctx, decode_state, gen6_mfd_context); intel_batchbuffer_start_atomic_bcs(batch, 0x1000); intel_batchbuffer_emit_mi_flush(batch); gen6_mfd_pipe_mode_select(ctx, decode_state, MFX_FORMAT_MPEG2, gen6_mfd_context); gen6_mfd_surface_state(ctx, decode_state, MFX_FORMAT_MPEG2, gen6_mfd_context); gen6_mfd_pipe_buf_addr_state(ctx, decode_state, MFX_FORMAT_MPEG2, gen6_mfd_context); gen6_mfd_bsp_buf_base_addr_state(ctx, decode_state, MFX_FORMAT_MPEG2, gen6_mfd_context); gen6_mfd_mpeg2_pic_state(ctx, decode_state, gen6_mfd_context); gen6_mfd_mpeg2_qm_state(ctx, decode_state, gen6_mfd_context); for (j = 0; j < decode_state->num_slice_params; j++) { assert(decode_state->slice_params && decode_state->slice_params[j]->buffer); slice_param = (VASliceParameterBufferMPEG2 *)decode_state->slice_params[j]->buffer; slice_data_bo = decode_state->slice_datas[j]->bo; gen6_mfd_ind_obj_base_addr_state(ctx, slice_data_bo, MFX_FORMAT_MPEG2, gen6_mfd_context); if (j == decode_state->num_slice_params - 1) next_slice_group_param = NULL; else next_slice_group_param = (VASliceParameterBufferMPEG2 *)decode_state->slice_params[j + 1]->buffer; for (i = 0; i < decode_state->slice_params[j]->num_elements; i++) { assert(slice_param->slice_data_flag == VA_SLICE_DATA_FLAG_ALL); if (i < decode_state->slice_params[j]->num_elements - 1) next_slice_param = slice_param + 1; else next_slice_param = next_slice_group_param; gen6_mfd_mpeg2_bsd_object(ctx, pic_param, slice_param, next_slice_param, gen6_mfd_context); slice_param++; } } intel_batchbuffer_end_atomic(batch); intel_batchbuffer_flush(batch); } static const int va_to_gen6_vc1_pic_type[5] = { GEN6_VC1_I_PICTURE, GEN6_VC1_P_PICTURE, GEN6_VC1_B_PICTURE, GEN6_VC1_BI_PICTURE, GEN6_VC1_P_PICTURE, }; static const int va_to_gen6_vc1_mv[4] = { 1, /* 1-MV */ 2, /* 1-MV half-pel */ 3, /* 1-MV half-pef bilinear */ 0, /* Mixed MV */ }; static const int b_picture_scale_factor[21] = { 128, 85, 170, 64, 192, 51, 102, 153, 204, 43, 215, 37, 74, 111, 148, 185, 222, 32, 96, 160, 224, }; static const int va_to_gen6_vc1_condover[3] = { 0, 2, 3 }; static const int va_to_gen6_vc1_profile[4] = { GEN6_VC1_SIMPLE_PROFILE, GEN6_VC1_MAIN_PROFILE, GEN6_VC1_RESERVED_PROFILE, GEN6_VC1_ADVANCED_PROFILE }; static void gen6_mfd_free_vc1_surface(void **data) { struct gen6_vc1_surface *gen6_vc1_surface = *data; if (!gen6_vc1_surface) return; dri_bo_unreference(gen6_vc1_surface->dmv); free(gen6_vc1_surface); *data = NULL; } static void gen6_mfd_init_vc1_surface(VADriverContextP ctx, VAPictureParameterBufferVC1 *pic_param, struct object_surface *obj_surface) { struct i965_driver_data *i965 = i965_driver_data(ctx); struct gen6_vc1_surface *gen6_vc1_surface = obj_surface->private_data; obj_surface->free_private_data = gen6_mfd_free_vc1_surface; if (!gen6_vc1_surface) { gen6_vc1_surface = calloc(sizeof(struct gen6_vc1_surface), 1); assert((obj_surface->size & 0x3f) == 0); obj_surface->private_data = gen6_vc1_surface; } gen6_vc1_surface->picture_type = pic_param->picture_fields.bits.picture_type; if (gen6_vc1_surface->dmv == NULL) { gen6_vc1_surface->dmv = dri_bo_alloc(i965->intel.bufmgr, "direct mv w/r buffer", 557056, /* 64 * 128 * 64 */ 0x1000); } } static void gen6_mfd_vc1_decode_init(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { VAPictureParameterBufferVC1 *pic_param; struct i965_driver_data *i965 = i965_driver_data(ctx); struct object_surface *obj_surface; int i; dri_bo *bo; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferVC1 *)decode_state->pic_param->buffer; /* reference picture */ obj_surface = SURFACE(pic_param->forward_reference_picture); if (obj_surface && obj_surface->bo) gen6_mfd_context->reference_surface[0].surface_id = pic_param->forward_reference_picture; else gen6_mfd_context->reference_surface[0].surface_id = VA_INVALID_ID; obj_surface = SURFACE(pic_param->backward_reference_picture); if (obj_surface && obj_surface->bo) gen6_mfd_context->reference_surface[1].surface_id = pic_param->backward_reference_picture; else gen6_mfd_context->reference_surface[1].surface_id = pic_param->forward_reference_picture; /* must do so !!! */ for (i = 2; i < ARRAY_ELEMS(gen6_mfd_context->reference_surface); i++) gen6_mfd_context->reference_surface[i].surface_id = gen6_mfd_context->reference_surface[i % 2].surface_id; /* Current decoded picture */ obj_surface = SURFACE(decode_state->current_render_target); assert(obj_surface); i965_check_alloc_surface_bo(ctx, obj_surface, 1, VA_FOURCC('N','V','1','2'), SUBSAMPLE_YUV420); gen6_mfd_init_vc1_surface(ctx, pic_param, obj_surface); dri_bo_unreference(gen6_mfd_context->post_deblocking_output.bo); gen6_mfd_context->post_deblocking_output.bo = obj_surface->bo; dri_bo_reference(gen6_mfd_context->post_deblocking_output.bo); gen6_mfd_context->post_deblocking_output.valid = pic_param->entrypoint_fields.bits.loopfilter; dri_bo_unreference(gen6_mfd_context->pre_deblocking_output.bo); gen6_mfd_context->pre_deblocking_output.bo = obj_surface->bo; dri_bo_reference(gen6_mfd_context->pre_deblocking_output.bo); gen6_mfd_context->pre_deblocking_output.valid = !pic_param->entrypoint_fields.bits.loopfilter; dri_bo_unreference(gen6_mfd_context->intra_row_store_scratch_buffer.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "intra row store", 128 * 64, 0x1000); assert(bo); gen6_mfd_context->intra_row_store_scratch_buffer.bo = bo; gen6_mfd_context->intra_row_store_scratch_buffer.valid = 1; dri_bo_unreference(gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "deblocking filter row store", 46080, /* 6 * 120 * 64 */ 0x1000); assert(bo); gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.bo = bo; gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.valid = 1; dri_bo_unreference(gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo); bo = dri_bo_alloc(i965->intel.bufmgr, "bsd mpc row store", 11520, /* 1.5 * 120 * 64 */ 0x1000); assert(bo); gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo = bo; gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.valid = 1; gen6_mfd_context->mpr_row_store_scratch_buffer.valid = 0; gen6_mfd_context->bitplane_read_buffer.valid = !!pic_param->bitplane_present.value; dri_bo_unreference(gen6_mfd_context->bitplane_read_buffer.bo); if (gen6_mfd_context->bitplane_read_buffer.valid) { int width_in_mbs = ALIGN(pic_param->coded_width, 16) / 16; int height_in_mbs = ALIGN(pic_param->coded_height, 16) / 16; int bitplane_width = ALIGN(width_in_mbs, 2) / 2; int src_w, src_h; uint8_t *src = NULL, *dst = NULL; assert(decode_state->bit_plane->buffer); src = decode_state->bit_plane->buffer; bo = dri_bo_alloc(i965->intel.bufmgr, "VC-1 Bitplane", bitplane_width * bitplane_width, 0x1000); assert(bo); gen6_mfd_context->bitplane_read_buffer.bo = bo; dri_bo_map(bo, True); assert(bo->virtual); dst = bo->virtual; for (src_h = 0; src_h < height_in_mbs; src_h++) { for(src_w = 0; src_w < width_in_mbs; src_w++) { int src_index, dst_index; int src_shift; uint8_t src_value; src_index = (src_h * width_in_mbs + src_w) / 2; src_shift = !((src_h * width_in_mbs + src_w) & 1) * 4; src_value = ((src[src_index] >> src_shift) & 0xf); dst_index = src_w / 2; dst[dst_index] = ((dst[dst_index] >> 4) | (src_value << 4)); } if (src_w & 1) dst[src_w / 2] >>= 4; dst += bitplane_width; } dri_bo_unmap(bo); } else gen6_mfd_context->bitplane_read_buffer.bo = NULL; } static void gen6_mfd_vc1_pic_state(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; VAPictureParameterBufferVC1 *pic_param; struct i965_driver_data *i965 = i965_driver_data(ctx); struct object_surface *obj_surface; int alt_pquant_config = 0, alt_pquant_edge_mask = 0, alt_pq; int dquant, dquantfrm, dqprofile, dqdbedge, dqsbedge, dqbilevel; int unified_mv_mode; int ref_field_pic_polarity = 0; int scale_factor = 0; int trans_ac_y = 0; int dmv_surface_valid = 0; int brfd = 0; int fcm = 0; int picture_type; int profile; int overlap; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferVC1 *)decode_state->pic_param->buffer; profile = va_to_gen6_vc1_profile[pic_param->sequence_fields.bits.profile]; dquant = pic_param->pic_quantizer_fields.bits.dquant; dquantfrm = pic_param->pic_quantizer_fields.bits.dq_frame; dqprofile = pic_param->pic_quantizer_fields.bits.dq_profile; dqdbedge = pic_param->pic_quantizer_fields.bits.dq_db_edge; dqsbedge = pic_param->pic_quantizer_fields.bits.dq_sb_edge; dqbilevel = pic_param->pic_quantizer_fields.bits.dq_binary_level; alt_pq = pic_param->pic_quantizer_fields.bits.alt_pic_quantizer; if (dquant == 0) { alt_pquant_config = 0; alt_pquant_edge_mask = 0; } else if (dquant == 2) { alt_pquant_config = 1; alt_pquant_edge_mask = 0xf; } else { assert(dquant == 1); if (dquantfrm == 0) { alt_pquant_config = 0; alt_pquant_edge_mask = 0; alt_pq = 0; } else { assert(dquantfrm == 1); alt_pquant_config = 1; switch (dqprofile) { case 3: if (dqbilevel == 0) { alt_pquant_config = 2; alt_pquant_edge_mask = 0; } else { assert(dqbilevel == 1); alt_pquant_config = 3; alt_pquant_edge_mask = 0; } break; case 0: alt_pquant_edge_mask = 0xf; break; case 1: if (dqdbedge == 3) alt_pquant_edge_mask = 0x9; else alt_pquant_edge_mask = (0x3 << dqdbedge); break; case 2: alt_pquant_edge_mask = (0x1 << dqsbedge); break; default: assert(0); } } } if (pic_param->mv_fields.bits.mv_mode == VAMvModeIntensityCompensation) { assert(pic_param->mv_fields.bits.mv_mode2 < 4); unified_mv_mode = va_to_gen6_vc1_mv[pic_param->mv_fields.bits.mv_mode2]; } else { assert(pic_param->mv_fields.bits.mv_mode < 4); unified_mv_mode = va_to_gen6_vc1_mv[pic_param->mv_fields.bits.mv_mode]; } if (pic_param->sequence_fields.bits.interlace == 1 && pic_param->picture_fields.bits.frame_coding_mode != 0) { /* frame-interlace or field-interlace */ /* FIXME: calculate reference field picture polarity */ assert(0); ref_field_pic_polarity = 0; } if (pic_param->b_picture_fraction < 21) scale_factor = b_picture_scale_factor[pic_param->b_picture_fraction]; picture_type = va_to_gen6_vc1_pic_type[pic_param->picture_fields.bits.picture_type]; if (profile == GEN6_VC1_ADVANCED_PROFILE && picture_type == GEN6_VC1_I_PICTURE) picture_type = GEN6_VC1_BI_PICTURE; if (picture_type == GEN6_VC1_I_PICTURE || picture_type == GEN6_VC1_BI_PICTURE) /* I picture */ trans_ac_y = pic_param->transform_fields.bits.transform_ac_codingset_idx2; else trans_ac_y = pic_param->transform_fields.bits.transform_ac_codingset_idx1; if (picture_type == GEN6_VC1_B_PICTURE) { struct gen6_vc1_surface *gen6_vc1_surface = NULL; obj_surface = SURFACE(pic_param->backward_reference_picture); assert(obj_surface); gen6_vc1_surface = obj_surface->private_data; if (!gen6_vc1_surface || (va_to_gen6_vc1_pic_type[gen6_vc1_surface->picture_type] == GEN6_VC1_I_PICTURE || va_to_gen6_vc1_pic_type[gen6_vc1_surface->picture_type] == GEN6_VC1_BI_PICTURE)) dmv_surface_valid = 0; else dmv_surface_valid = 1; } assert(pic_param->picture_fields.bits.frame_coding_mode < 3); if (pic_param->picture_fields.bits.frame_coding_mode < 2) fcm = pic_param->picture_fields.bits.frame_coding_mode; else { if (pic_param->picture_fields.bits.top_field_first) fcm = 2; else fcm = 3; } if (pic_param->picture_fields.bits.picture_type == GEN6_VC1_B_PICTURE) { /* B picture */ brfd = pic_param->reference_fields.bits.reference_distance; brfd = (scale_factor * brfd) >> 8; brfd = pic_param->reference_fields.bits.reference_distance - brfd - 1; if (brfd < 0) brfd = 0; } overlap = pic_param->sequence_fields.bits.overlap; if (profile != GEN6_VC1_ADVANCED_PROFILE && pic_param->pic_quantizer_fields.bits.pic_quantizer_scale < 9) overlap = 0; assert(pic_param->conditional_overlap_flag < 3); assert(pic_param->mv_fields.bits.mv_table < 4); /* FIXME: interlace mode */ BEGIN_BCS_BATCH(batch, 6); OUT_BCS_BATCH(batch, MFX_VC1_PIC_STATE | (6 - 2)); OUT_BCS_BATCH(batch, (ALIGN(pic_param->coded_height, 16) / 16) << 16 | (ALIGN(pic_param->coded_width, 16) / 16)); OUT_BCS_BATCH(batch, pic_param->sequence_fields.bits.syncmarker << 31 | 1 << 29 | /* concealment */ alt_pq << 24 | pic_param->entrypoint_fields.bits.loopfilter << 23 | overlap << 22 | (pic_param->pic_quantizer_fields.bits.quantizer == 0) << 21 | /* implicit quantizer */ pic_param->pic_quantizer_fields.bits.pic_quantizer_scale << 16 | alt_pquant_edge_mask << 12 | alt_pquant_config << 10 | pic_param->pic_quantizer_fields.bits.half_qp << 9 | pic_param->pic_quantizer_fields.bits.pic_quantizer_type << 8 | va_to_gen6_vc1_condover[pic_param->conditional_overlap_flag] << 6 | !pic_param->picture_fields.bits.is_first_field << 5 | picture_type << 2 | fcm << 0); OUT_BCS_BATCH(batch, !!pic_param->bitplane_present.value << 23 | !pic_param->bitplane_present.flags.bp_forward_mb << 22 | !pic_param->bitplane_present.flags.bp_mv_type_mb << 21 | !pic_param->bitplane_present.flags.bp_skip_mb << 20 | !pic_param->bitplane_present.flags.bp_direct_mb << 19 | !pic_param->bitplane_present.flags.bp_overflags << 18 | !pic_param->bitplane_present.flags.bp_ac_pred << 17 | !pic_param->bitplane_present.flags.bp_field_tx << 16 | pic_param->mv_fields.bits.extended_dmv_range << 14 | pic_param->mv_fields.bits.extended_mv_range << 12 | pic_param->mv_fields.bits.four_mv_switch << 11 | pic_param->fast_uvmc_flag << 10 | unified_mv_mode << 8 | ref_field_pic_polarity << 6 | pic_param->reference_fields.bits.num_reference_pictures << 5 | pic_param->reference_fields.bits.reference_distance << 0); OUT_BCS_BATCH(batch, scale_factor << 24 | pic_param->mv_fields.bits.mv_table << 20 | pic_param->mv_fields.bits.four_mv_block_pattern_table << 18 | pic_param->mv_fields.bits.two_mv_block_pattern_table << 16 | pic_param->transform_fields.bits.frame_level_transform_type << 12 | pic_param->transform_fields.bits.mb_level_transform_type_flag << 11 | pic_param->mb_mode_table << 8 | trans_ac_y << 6 | pic_param->transform_fields.bits.transform_ac_codingset_idx1 << 4 | pic_param->transform_fields.bits.intra_transform_dc_table << 3 | pic_param->cbp_table << 0); OUT_BCS_BATCH(batch, dmv_surface_valid << 13 | brfd << 8 | ((ALIGN(pic_param->coded_width, 16) / 16 + 1) / 2 - 1)); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_vc1_pred_pipe_state(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; VAPictureParameterBufferVC1 *pic_param; int interpolation_mode = 0; int intensitycomp_single; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferVC1 *)decode_state->pic_param->buffer; if (pic_param->mv_fields.bits.mv_mode == VAMvMode1MvHalfPelBilinear || (pic_param->mv_fields.bits.mv_mode == VAMvModeIntensityCompensation && pic_param->mv_fields.bits.mv_mode2 == VAMvMode1MvHalfPelBilinear)) interpolation_mode = 2; /* Half-pel bilinear */ else if (pic_param->mv_fields.bits.mv_mode == VAMvMode1MvHalfPel || (pic_param->mv_fields.bits.mv_mode == VAMvModeIntensityCompensation && pic_param->mv_fields.bits.mv_mode2 == VAMvMode1MvHalfPel)) interpolation_mode = 0; /* Half-pel bicubic */ else interpolation_mode = 1; /* Quarter-pel bicubic */ assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferVC1 *)decode_state->pic_param->buffer; intensitycomp_single = (pic_param->mv_fields.bits.mv_mode == VAMvModeIntensityCompensation); BEGIN_BCS_BATCH(batch, 7); OUT_BCS_BATCH(batch, MFX_VC1_PRED_PIPE_STATE | (7 - 2)); OUT_BCS_BATCH(batch, 0 << 8 | /* FIXME: interlace mode */ pic_param->rounding_control << 4 | va_to_gen6_vc1_profile[pic_param->sequence_fields.bits.profile] << 2); OUT_BCS_BATCH(batch, pic_param->luma_shift << 16 | pic_param->luma_scale << 0); /* FIXME: Luma Scaling */ OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, 0); OUT_BCS_BATCH(batch, interpolation_mode << 19 | pic_param->fast_uvmc_flag << 18 | 0 << 17 | /* FIXME: scale up or down ??? */ pic_param->range_reduction_frame << 16 | 0 << 6 | /* FIXME: double ??? */ 0 << 4 | intensitycomp_single << 2 | intensitycomp_single << 0); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_vc1_directmode_state(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; VAPictureParameterBufferVC1 *pic_param; struct i965_driver_data *i965 = i965_driver_data(ctx); struct object_surface *obj_surface; dri_bo *dmv_read_buffer = NULL, *dmv_write_buffer = NULL; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferVC1 *)decode_state->pic_param->buffer; obj_surface = SURFACE(decode_state->current_render_target); if (obj_surface && obj_surface->private_data) { dmv_write_buffer = ((struct gen6_vc1_surface *)(obj_surface->private_data))->dmv; } obj_surface = SURFACE(pic_param->backward_reference_picture); if (obj_surface && obj_surface->private_data) { dmv_read_buffer = ((struct gen6_vc1_surface *)(obj_surface->private_data))->dmv; } BEGIN_BCS_BATCH(batch, 3); OUT_BCS_BATCH(batch, MFX_VC1_DIRECTMODE_STATE | (3 - 2)); if (dmv_write_buffer) OUT_BCS_RELOC(batch, dmv_write_buffer, I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION, 0); else OUT_BCS_BATCH(batch, 0); if (dmv_read_buffer) OUT_BCS_RELOC(batch, dmv_read_buffer, I915_GEM_DOMAIN_INSTRUCTION, 0, 0); else OUT_BCS_BATCH(batch, 0); ADVANCE_BCS_BATCH(batch); } static int gen6_mfd_vc1_get_macroblock_bit_offset(uint8_t *buf, int in_slice_data_bit_offset, int profile) { int out_slice_data_bit_offset; int slice_header_size = in_slice_data_bit_offset / 8; int i, j; if (profile != 3) out_slice_data_bit_offset = in_slice_data_bit_offset; else { for (i = 0, j = 0; i < slice_header_size; i++, j++) { if (!buf[j] && !buf[j + 1] && buf[j + 2] == 3 && buf[j + 3] < 4) { i++, j += 2; } } out_slice_data_bit_offset = 8 * j + in_slice_data_bit_offset % 8; } return out_slice_data_bit_offset; } static void gen6_mfd_vc1_bsd_object(VADriverContextP ctx, VAPictureParameterBufferVC1 *pic_param, VASliceParameterBufferVC1 *slice_param, VASliceParameterBufferVC1 *next_slice_param, dri_bo *slice_data_bo, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; int next_slice_start_vert_pos; int macroblock_offset; uint8_t *slice_data = NULL; dri_bo_map(slice_data_bo, 0); slice_data = (uint8_t *)(slice_data_bo->virtual + slice_param->slice_data_offset); macroblock_offset = gen6_mfd_vc1_get_macroblock_bit_offset(slice_data, slice_param->macroblock_offset, pic_param->sequence_fields.bits.profile); dri_bo_unmap(slice_data_bo); if (next_slice_param) next_slice_start_vert_pos = next_slice_param->slice_vertical_position; else next_slice_start_vert_pos = ALIGN(pic_param->coded_height, 16) / 16; BEGIN_BCS_BATCH(batch, 4); OUT_BCS_BATCH(batch, MFD_VC1_BSD_OBJECT | (4 - 2)); OUT_BCS_BATCH(batch, slice_param->slice_data_size - (macroblock_offset >> 3)); OUT_BCS_BATCH(batch, slice_param->slice_data_offset + (macroblock_offset >> 3)); OUT_BCS_BATCH(batch, slice_param->slice_vertical_position << 24 | next_slice_start_vert_pos << 16 | (macroblock_offset & 0x7)); ADVANCE_BCS_BATCH(batch); } static void gen6_mfd_vc1_decode_picture(VADriverContextP ctx, struct decode_state *decode_state, struct gen6_mfd_context *gen6_mfd_context) { struct intel_batchbuffer *batch = gen6_mfd_context->base.batch; VAPictureParameterBufferVC1 *pic_param; VASliceParameterBufferVC1 *slice_param, *next_slice_param, *next_slice_group_param; dri_bo *slice_data_bo; int i, j; assert(decode_state->pic_param && decode_state->pic_param->buffer); pic_param = (VAPictureParameterBufferVC1 *)decode_state->pic_param->buffer; gen6_mfd_vc1_decode_init(ctx, decode_state, gen6_mfd_context); intel_batchbuffer_start_atomic_bcs(batch, 0x1000); intel_batchbuffer_emit_mi_flush(batch); gen6_mfd_pipe_mode_select(ctx, decode_state, MFX_FORMAT_VC1, gen6_mfd_context); gen6_mfd_surface_state(ctx, decode_state, MFX_FORMAT_VC1, gen6_mfd_context); gen6_mfd_pipe_buf_addr_state(ctx, decode_state, MFX_FORMAT_VC1, gen6_mfd_context); gen6_mfd_bsp_buf_base_addr_state(ctx, decode_state, MFX_FORMAT_VC1, gen6_mfd_context); gen6_mfd_vc1_pic_state(ctx, decode_state, gen6_mfd_context); gen6_mfd_vc1_pred_pipe_state(ctx, decode_state, gen6_mfd_context); gen6_mfd_vc1_directmode_state(ctx, decode_state, gen6_mfd_context); for (j = 0; j < decode_state->num_slice_params; j++) { assert(decode_state->slice_params && decode_state->slice_params[j]->buffer); slice_param = (VASliceParameterBufferVC1 *)decode_state->slice_params[j]->buffer; slice_data_bo = decode_state->slice_datas[j]->bo; gen6_mfd_ind_obj_base_addr_state(ctx, slice_data_bo, MFX_FORMAT_VC1, gen6_mfd_context); if (j == decode_state->num_slice_params - 1) next_slice_group_param = NULL; else next_slice_group_param = (VASliceParameterBufferVC1 *)decode_state->slice_params[j + 1]->buffer; for (i = 0; i < decode_state->slice_params[j]->num_elements; i++) { assert(slice_param->slice_data_flag == VA_SLICE_DATA_FLAG_ALL); if (i < decode_state->slice_params[j]->num_elements - 1) next_slice_param = slice_param + 1; else next_slice_param = next_slice_group_param; gen6_mfd_vc1_bsd_object(ctx, pic_param, slice_param, next_slice_param, slice_data_bo, gen6_mfd_context); slice_param++; } } intel_batchbuffer_end_atomic(batch); intel_batchbuffer_flush(batch); } static void gen6_mfd_decode_picture(VADriverContextP ctx, VAProfile profile, union codec_state *codec_state, struct hw_context *hw_context) { struct gen6_mfd_context *gen6_mfd_context = (struct gen6_mfd_context *)hw_context; struct decode_state *decode_state = &codec_state->decode; assert(gen6_mfd_context); switch (profile) { case VAProfileMPEG2Simple: case VAProfileMPEG2Main: gen6_mfd_mpeg2_decode_picture(ctx, decode_state, gen6_mfd_context); break; case VAProfileH264Baseline: case VAProfileH264Main: case VAProfileH264High: gen6_mfd_avc_decode_picture(ctx, decode_state, gen6_mfd_context); break; case VAProfileVC1Simple: case VAProfileVC1Main: case VAProfileVC1Advanced: gen6_mfd_vc1_decode_picture(ctx, decode_state, gen6_mfd_context); break; default: assert(0); break; } } static void gen6_mfd_context_destroy(void *hw_context) { struct gen6_mfd_context *gen6_mfd_context = (struct gen6_mfd_context *)hw_context; dri_bo_unreference(gen6_mfd_context->post_deblocking_output.bo); gen6_mfd_context->post_deblocking_output.bo = NULL; dri_bo_unreference(gen6_mfd_context->pre_deblocking_output.bo); gen6_mfd_context->pre_deblocking_output.bo = NULL; dri_bo_unreference(gen6_mfd_context->intra_row_store_scratch_buffer.bo); gen6_mfd_context->intra_row_store_scratch_buffer.bo = NULL; dri_bo_unreference(gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.bo); gen6_mfd_context->deblocking_filter_row_store_scratch_buffer.bo = NULL; dri_bo_unreference(gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo); gen6_mfd_context->bsd_mpc_row_store_scratch_buffer.bo = NULL; dri_bo_unreference(gen6_mfd_context->mpr_row_store_scratch_buffer.bo); gen6_mfd_context->mpr_row_store_scratch_buffer.bo = NULL; dri_bo_unreference(gen6_mfd_context->bitplane_read_buffer.bo); gen6_mfd_context->bitplane_read_buffer.bo = NULL; intel_batchbuffer_free(gen6_mfd_context->base.batch); free(gen6_mfd_context); } struct hw_context * gen6_dec_hw_context_init(VADriverContextP ctx, VAProfile profile) { struct intel_driver_data *intel = intel_driver_data(ctx); struct gen6_mfd_context *gen6_mfd_context = calloc(1, sizeof(struct gen6_mfd_context)); int i; gen6_mfd_context->base.destroy = gen6_mfd_context_destroy; gen6_mfd_context->base.run = gen6_mfd_decode_picture; gen6_mfd_context->base.batch = intel_batchbuffer_new(intel, I915_EXEC_RENDER); for (i = 0; i < ARRAY_ELEMS(gen6_mfd_context->reference_surface); i++) { gen6_mfd_context->reference_surface[i].surface_id = VA_INVALID_ID; gen6_mfd_context->reference_surface[i].frame_store_id = -1; } return (struct hw_context *)gen6_mfd_context; }