/*
 * Copyright (c) 2018 Samsung Electronics Co., Ltd. All Rights Reserved
 * Copyright (c) 2016, 2017 ARM Limited.
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *      http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
#include "helpers.h"

#ifdef SATURATE
#define CONVERT_OP_FLOAT_STR(x, type, round) (convert_##type##_sat##round(x))
#else /* SATURATE */
#define CONVERT_OP_FLOAT_STR(x, type, round) (convert_##type##round(x))
#endif /* SATURATE */
#define CONVERT_OP_FLOAT(x, type, round) CONVERT_OP_FLOAT_STR(x, type, round)

/** Performs a pixelwise division with float scale of either integer or float inputs.
 *
 * @attention The inputs and output data types need to be passed at compile time using -DDATA_TYPE_IN1, -DDATA_TYPE_IN2 and -DDATA_TYPE_OUT:
 * e.g. -DDATA_TYPE_IN1=uchar -DDATA_TYPE_IN2=ushort -DDATA_TYPE_OUT=short
 * @attention The data type of the intermediate result of the division should passed as well using -DDATA_TYPE_RES.
 * e.g. If one of inputs is S16 -DDATA_TYPE_RES=int should be passed else -DDATA_TYPE_RES=short.
 * @attention -DDATA_TYPE_FLOAT must be passed if floating point inputs are provided.
 *
 * @param[in]  in1_ptr                           Pointer to the source image. Supported data types: U8, S16, F16, F32
 * @param[in]  in1_stride_x                      Stride of the source image in X dimension (in bytes)
 * @param[in]  in1_step_x                        in1_stride_x * number of elements along X processed per workitem(in bytes)
 * @param[in]  in1_stride_y                      Stride of the source image in Y dimension (in bytes)
 * @param[in]  in1_step_y                        in1_stride_y * number of elements along Y processed per workitem(in bytes)
 * @param[in]  in1_stride_z                      Stride of the source image in Y dimension (in bytes)
 * @param[in]  in1_step_z                        in1_stride_z * number of elements along Y processed per workitem(in bytes)
 * @param[in]  in1_offset_first_element_in_bytes The offset of the first element in the source image
 * @param[in]  in2_ptr                           Pointer to the source image. Supported data types: U8, S16, F16, F32
 * @param[in]  in2_stride_x                      Stride of the source image in X dimension (in bytes)
 * @param[in]  in2_step_x                        in2_stride_x * number of elements along X processed per workitem(in bytes)
 * @param[in]  in2_stride_y                      Stride of the source image in Y dimension (in bytes)
 * @param[in]  in2_step_y                        in2_stride_y * number of elements along Y processed per workitem(in bytes)
 * @param[in]  in2_stride_z                      Stride of the source image in Y dimension (in bytes)
 * @param[in]  in2_step_z                        in2_stride_z * number of elements along Y processed per workitem(in bytes)
 * @param[in]  in2_offset_first_element_in_bytes The offset of the first element in the source image
 * @param[out] out_ptr                           Pointer to the destination image. Supported data types: U8, S16, F16, F32
 * @param[in]  out_stride_x                      Stride of the destination image in X dimension (in bytes)
 * @param[in]  out_step_x                        out_stride_x * number of elements along X processed per workitem(in bytes)
 * @param[in]  out_stride_y                      Stride of the destination image in Y dimension (in bytes)
 * @param[in]  out_step_y                        out_stride_y * number of elements along Y processed per workitem(in bytes)
 * @param[in]  out_stride_z                      Stride of the destination image in Y dimension (in bytes)
 * @param[in]  out_step_z                        out_stride_z * number of elements along Y processed per workitem(in bytes)
 * @param[in]  out_offset_first_element_in_bytes The offset of the first element in the destination image
 * @param[in]  scale                             Float scaling factor. Supported data types: F32
 */
__kernel void pixelwise_div_float(
    TENSOR3D_DECLARATION(in1),
    TENSOR3D_DECLARATION(in2),
    TENSOR3D_DECLARATION(out),
    const float scale)
{
    // Get pixels pointer
    Tensor3D in1 = CONVERT_TO_TENSOR3D_STRUCT(in1);
    Tensor3D in2 = CONVERT_TO_TENSOR3D_STRUCT(in2);
    Tensor3D out = CONVERT_TO_TENSOR3D_STRUCT(out);

    // Load data
    VEC_DATA_TYPE(DATA_TYPE_RES, 16)
    in1_data = CONVERT(vload16(0, (__global DATA_TYPE_IN1 *)in1.ptr), VEC_DATA_TYPE(DATA_TYPE_RES, 16));
    VEC_DATA_TYPE(DATA_TYPE_RES, 16)
    in2_data = CONVERT(vload16(0, (__global DATA_TYPE_IN2 *)in2.ptr), VEC_DATA_TYPE(DATA_TYPE_RES, 16));

    // Perform division
#ifdef DATA_TYPE_FLOAT
    VEC_DATA_TYPE(DATA_TYPE_OUT, 16)
    res = CONVERT(in1_data / in2_data * (DATA_TYPE_RES)scale, VEC_DATA_TYPE(DATA_TYPE_OUT, 16));
#else  /* DATA_TYPE_FLOAT */
    VEC_DATA_TYPE(DATA_TYPE_OUT, 16)
    res = CONVERT_OP_FLOAT(CONVERT_OP_FLOAT((convert_float16(in1_data / in2_data) * scale), VEC_DATA_TYPE(DATA_TYPE_RES, 16), ROUND), VEC_DATA_TYPE(DATA_TYPE_OUT, 16), ROUND);
#endif /* DATA_TYPE_FLOAT */

    // Store result
    vstore16(res, 0, (__global DATA_TYPE_OUT *)out.ptr);
}