generic-library/opencv
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge pull request #817 from pengx17:2.4_ocl_bitwise_cleanup

Browse Source
This commit is contained in:
Andrey Kamaev 2013-04-17 15:45:37 +04:00 committed by OpenCV Buildbot
commit 28aefc4f5a
13 changed files with 246 additions and 5609 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

96
modules/ocl/src/arithm.cpp
View File

@ -62,8 +62,6 @@ namespace cv
namespace ocl
{
////////////////////////////////OpenCL kernel strings/////////////////////
extern const char *bitwise;
extern const char *bitwiseM;
extern const char *transpose_kernel;
extern const char *arithm_nonzero;
extern const char *arithm_sum;
@ -77,19 +75,11 @@ namespace cv
extern const char *arithm_add;
extern const char *arithm_add_scalar;
extern const char *arithm_add_scalar_mask;
extern const char *arithm_bitwise_binary;
extern const char *arithm_bitwise_binary_mask;
extern const char *arithm_bitwise_binary_scalar;
extern const char *arithm_bitwise_binary_scalar_mask;
extern const char *arithm_bitwise_not;
extern const char *arithm_bitwise_and;
extern const char *arithm_bitwise_and_mask;
extern const char *arithm_bitwise_and_scalar;
extern const char *arithm_bitwise_and_scalar_mask;
extern const char *arithm_bitwise_or;
extern const char *arithm_bitwise_or_mask;
extern const char *arithm_bitwise_or_scalar;
extern const char *arithm_bitwise_or_scalar_mask;
extern const char *arithm_bitwise_xor;
extern const char *arithm_bitwise_xor_mask;
extern const char *arithm_bitwise_xor_scalar;
extern const char *arithm_bitwise_xor_scalar_mask;
extern const char *arithm_compare_eq;
extern const char *arithm_compare_ne;
extern const char *arithm_mul;
@ -1646,7 +1636,8 @@ static void bitwise_run(const oclMat &src1, oclMat &dst, string kernelName, cons
template<typename T>
void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, string kernelName, const char **kernelString, void *_scalar)
void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, string kernelName,
const char **kernelString, void *_scalar, const char* _opt = NULL)
{
dst.create(src1.size(), src1.type());
CV_Assert(src1.cols == src2.cols && src2.cols == dst.cols &&
@ -1696,13 +1687,15 @@ void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, string ker
args.push_back( make_pair( sizeof(T), (void *)&scalar ));
}
openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth);
openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, -1, depth, _opt);
}
static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, string kernelName, const char **kernelString)
static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst,
string kernelName, const char **kernelString, const char* _opt = NULL)
{
bitwise_run<char>(src1, src2, dst, kernelName, kernelString, (void *)NULL);
bitwise_run<char>(src1, src2, dst, kernelName, kernelString, (void *)NULL, _opt);
}
static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask, string kernelName, const char **kernelString)
static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst,
const oclMat &mask, string kernelName, const char **kernelString, const char* _opt = NULL)
{
dst.create(src1.size(), src1.type());
CV_Assert(src1.cols == src2.cols && src2.cols == dst.cols &&
@ -1750,12 +1743,13 @@ static void bitwise_run(const oclMat &src1, const oclMat &src2, oclMat &dst, con
args.push_back( make_pair( sizeof(cl_int), (void *)&cols ));
args.push_back( make_pair( sizeof(cl_int), (void *)&dst_step1 ));
openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth);
openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, _opt);
}
template <typename WT , typename CL_WT>
void bitwise_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, string kernelName, const char **kernelString, int isMatSubScalar)
void bitwise_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst,
const oclMat &mask, string kernelName, const char **kernelString, int isMatSubScalar, const char* opt = NULL)
{
dst.create(src1.size(), src1.type());
@ -1817,14 +1811,16 @@ void bitwise_scalar_run(const oclMat &src1, const Scalar &src2, oclMat &dst, con
args.push_back( make_pair( sizeof(cl_int) , (void *)&isMatSubScalar));
}
openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth);
openCLExecuteKernel(clCxt, kernelString, kernelName, globalThreads, localThreads, args, channels, depth, opt);
}
typedef void (*BitwiseFuncS)(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, string kernelName, const char **kernelString, int isMatSubScalar);
typedef void (*BitwiseFuncS)(const oclMat &src1, const Scalar &src2, oclMat &dst,
const oclMat &mask, string kernelName, const char **kernelString, int isMatSubScalar, const char* opt);
static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, string kernelName, const char **kernelString, int isMatSubScalar)
static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst,
const oclMat &mask, string kernelName, const char **kernelString, int isMatSubScalar, const char* opt)
{
static BitwiseFuncS tab[8] =
{
@ -1852,11 +1848,12 @@ static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst,
BitwiseFuncS func = tab[src1.depth()];
if(func == 0)
cv::ocl::error("Unsupported arithmetic operation", __FILE__, __LINE__);
func(src1, src2, dst, mask, kernelName, kernelString, isMatSubScalar);
func(src1, src2, dst, mask, kernelName, kernelString, isMatSubScalar, opt);
}
static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask, string kernelName, const char **kernelString)
static void bitwise_scalar(const oclMat &src1, const Scalar &src2, oclMat &dst,
const oclMat &mask, string kernelName, const char **kernelString, const char * opt = NULL)
{
bitwise_scalar(src1, src2, dst, mask, kernelName, kernelString, 0);
bitwise_scalar(src1, src2, dst, mask, kernelName, kernelString, 0, opt);
}
void cv::ocl::bitwise_not(const oclMat &src, oclMat &dst)
@ -1879,12 +1876,13 @@ void cv::ocl::bitwise_or(const oclMat &src1, const oclMat &src2, oclMat &dst, co
cout << "Selected device do not support double" << endl;
return;
}
oclMat emptyMat;
string kernelName = mask.empty() ? "arithm_bitwise_or" : "arithm_bitwise_or_with_mask";
string kernelName = mask.empty() ? "arithm_bitwise_binary" : "arithm_bitwise_binary_with_mask";
static const char opt [] = "-D OP_BINARY=|";
if (mask.empty())
bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_or);
bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_binary, opt);
else
bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_or_mask);
bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_mask, opt);
}
@ -1895,11 +1893,12 @@ void cv::ocl::bitwise_or(const oclMat &src1, const Scalar &src2, oclMat &dst, co
cout << "Selected device do not support double" << endl;
return;
}
string kernelName = mask.data ? "arithm_s_bitwise_or_with_mask" : "arithm_s_bitwise_or";
static const char opt [] = "-D OP_BINARY=|";
string kernelName = mask.data ? "arithm_s_bitwise_binary_with_mask" : "arithm_s_bitwise_binary";
if (mask.data)
bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_or_scalar_mask);
bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar_mask, opt);
else
bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_or_scalar);
bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar, opt);
}
void cv::ocl::bitwise_and(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask)
@ -1912,12 +1911,13 @@ void cv::ocl::bitwise_and(const oclMat &src1, const oclMat &src2, oclMat &dst, c
}
oclMat emptyMat;
string kernelName = mask.empty() ? "arithm_bitwise_and" : "arithm_bitwise_and_with_mask";
string kernelName = mask.empty() ? "arithm_bitwise_binary" : "arithm_bitwise_binary_with_mask";
static const char opt [] = "-D OP_BINARY=&";
if (mask.empty())
bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_and);
bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_binary, opt);
else
bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_and_mask);
bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_mask, opt);
}
void cv::ocl::bitwise_and(const oclMat &src1, const Scalar &src2, oclMat &dst, const oclMat &mask)
@ -1927,11 +1927,12 @@ void cv::ocl::bitwise_and(const oclMat &src1, const Scalar &src2, oclMat &dst, c
cout << "Selected device do not support double" << endl;
return;
}
string kernelName = mask.data ? "arithm_s_bitwise_and_with_mask" : "arithm_s_bitwise_and";
static const char opt [] = "-D OP_BINARY=&";
string kernelName = mask.data ? "arithm_s_bitwise_binary_with_mask" : "arithm_s_bitwise_binary";
if (mask.data)
bitwise_scalar(src1, src2, dst, mask, kernelName, &arithm_bitwise_and_scalar_mask);
bitwise_scalar(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar_mask, opt);
else
bitwise_scalar(src1, src2, dst, mask, kernelName, &arithm_bitwise_and_scalar);
bitwise_scalar(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar, opt);
}
void cv::ocl::bitwise_xor(const oclMat &src1, const oclMat &src2, oclMat &dst, const oclMat &mask)
@ -1941,14 +1942,14 @@ void cv::ocl::bitwise_xor(const oclMat &src1, const oclMat &src2, oclMat &dst, c
cout << "Selected device do not support double" << endl;
return;
}
oclMat emptyMat;
string kernelName = mask.empty() ? "arithm_bitwise_xor" : "arithm_bitwise_xor_with_mask";
string kernelName = mask.empty() ? "arithm_bitwise_binary" : "arithm_bitwise_binary_with_mask";
static const char opt [] = "-D OP_BINARY=^";
if (mask.empty())
bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_xor);
bitwise_run(src1, src2, dst, kernelName, &arithm_bitwise_binary, opt);
else
bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_xor_mask);
bitwise_run(src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_mask, opt);
}
@ -1960,11 +1961,12 @@ void cv::ocl::bitwise_xor(const oclMat &src1, const Scalar &src2, oclMat &dst, c
cout << "Selected device do not support double" << endl;
return;
}
string kernelName = mask.data ? "arithm_s_bitwise_xor_with_mask" : "arithm_s_bitwise_xor";
string kernelName = mask.data ? "arithm_s_bitwise_binary_with_mask" : "arithm_s_bitwise_binary";
static const char opt [] = "-D OP_BINARY=^";
if (mask.data)
bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_xor_scalar_mask);
bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar_mask, opt);
else
bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_xor_scalar);
bitwise_scalar( src1, src2, dst, mask, kernelName, &arithm_bitwise_binary_scalar, opt);
}
oclMat cv::ocl::operator ~ (const oclMat &src)

803
modules/ocl/src/opencl/arithm_bitwise_and_mask.cl
View File

@ -1,803 +0,0 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other oclMaterials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable
#elif defined (cl_amd_fp64)
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_AND////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_and with mask**************************************/
__kernel void arithm_bitwise_and_with_mask_C1_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = vload4(0, src2 + src2_index);
uchar4 mask_data = vload4(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data & src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C1_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = vload4(0, src2 + src2_index);
uchar4 mask_data = vload4(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data & src2_data;
data.x = convert_char((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_char((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = convert_char((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = convert_char((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C1_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
ushort2 src2_data = vload2(0, (__global ushort *)((__global char *)src2 + src2_index));
uchar2 mask_data = vload2(0, mask + mask_index);
ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
ushort2 tmp_data = src1_data & src2_data;
data.x = convert_ushort((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_ushort((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
*((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C1_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
short2 src2_data = vload2(0, (__global short *)((__global char *)src2 + src2_index));
uchar2 mask_data = vload2(0, mask + mask_index);
short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
short2 tmp_data = src1_data & src2_data;
data.x = convert_short((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_short((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
*((__global short2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C1_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
int dst_data = *((__global int *)((__global char *)dst + dst_index));
int data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global int *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C1_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
char4 src_data2 = *((__global char4 *)((__global char *)src2 + src2_index));
char4 dst_data = *((__global char4 *)((__global char *)dst + dst_index));
char4 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global char4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C1_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
char8 dst_data = *((__global char8 *)((__global char *)dst + dst_index));
char8 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global char8 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C2_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = vload4(0, src2 + src2_index);
uchar2 mask_data = vload2(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data & src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C2_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = vload4(0, src2 + src2_index);
uchar2 mask_data = vload2(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data & src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C2_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index));
ushort2 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global ushort2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C2_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index));
short2 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global short2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C2_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index));
int2 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global int2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C2_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
char8 dst_data = *((__global char8 *)((__global char *)dst + dst_index));
char8 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global char8 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C2_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
char16 dst_data = *((__global char16 *)((__global char *)dst + dst_index));
char16 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C4_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
uchar4 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C4_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
char4 src_data1 = *((__global char4 *)(src1 + src1_index));
char4 src_data2 = *((__global char4 *)(src2 + src2_index));
char4 dst_data = *((__global char4 *)(dst + dst_index));
char4 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C4_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
ushort4 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global ushort4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C4_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C4_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index));
int4 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global int4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_and_with_mask_C4_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
char16 dst_data = *((__global char16 *)((__global char *)dst + dst_index));
char16 data = src_data1 & src_data2;
data = mask_data ? data : dst_data;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_and_with_mask_C4_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 5) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 5) + dst_offset);
uchar mask_data = *(mask + mask_index);
char8 src_data1_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0));
char8 src_data1_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8));
char8 src_data1_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
char8 src_data1_3 = *((__global char8 *)((__global char *)src1 + src1_index + 24));
char8 src_data2_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0));
char8 src_data2_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8));
char8 src_data2_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
char8 src_data2_3 = *((__global char8 *)((__global char *)src2 + src2_index + 24));
char8 dst_data_0 = *((__global char8 *)((__global char *)dst + dst_index + 0));
char8 dst_data_1 = *((__global char8 *)((__global char *)dst + dst_index + 8));
char8 dst_data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16));
char8 dst_data_3 = *((__global char8 *)((__global char *)dst + dst_index + 24));
char8 data_0 = src_data1_0 & src_data2_0;
char8 data_1 = src_data1_1 & src_data2_1;
char8 data_2 = src_data1_2 & src_data2_2;
char8 data_3 = src_data1_3 & src_data2_3;
data_0 = mask_data ? data_0 : dst_data_0;
data_1 = mask_data ? data_1 : dst_data_1;
data_2 = mask_data ? data_2 : dst_data_2;
data_3 = mask_data ? data_3 : dst_data_3;
*((__global char8 *)((__global char *)dst + dst_index + 0)) = data_0;
*((__global char8 *)((__global char *)dst + dst_index + 8)) = data_1;
*((__global char8 *)((__global char *)dst + dst_index + 16)) = data_2;
*((__global char8 *)((__global char *)dst + dst_index + 24)) = data_3;
}
}
#endif

1117
modules/ocl/src/opencl/arithm_bitwise_and_scalar_mask.cl
View File

File diff suppressed because it is too large Load Diff

43
modules/ocl/src/opencl/arithm_bitwise_and.cl → modules/ocl/src/opencl/arithm_bitwise_binary.cl
View File

@ -16,6 +16,7 @@
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
// Peng Xiao, pengxiao@outlook.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
@ -50,11 +51,17 @@
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_AND////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_and without mask**************************************/
__kernel void arithm_bitwise_and_D0 (__global uchar *src1, int src1_step, int src1_offset,
//bitwise_binary without mask for and, or, xor operators
/////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////bitwise_binary///////////////////////////////////////////
/////////////////////////////////////////////////////////////////////////////////////////////////////
#ifndef OP_BINARY
#define OP_BINARY &
#endif
__kernel void arithm_bitwise_binary_D0 (__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
@ -95,7 +102,7 @@ __kernel void arithm_bitwise_and_D0 (__global uchar *src1, int src1_step, int sr
}
uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data & src2_data;
uchar4 tmp_data = src1_data OP_BINARY src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
@ -107,7 +114,7 @@ __kernel void arithm_bitwise_and_D0 (__global uchar *src1, int src1_step, int sr
}
__kernel void arithm_bitwise_and_D1 (__global char *src1, int src1_step, int src1_offset,
__kernel void arithm_bitwise_binary_D1 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
@ -148,7 +155,7 @@ __kernel void arithm_bitwise_and_D1 (__global char *src1, int src1_step, int src
src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
}
char4 dst_data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data & src2_data;
char4 tmp_data = src1_data OP_BINARY src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
@ -160,7 +167,7 @@ __kernel void arithm_bitwise_and_D1 (__global char *src1, int src1_step, int src
}
__kernel void arithm_bitwise_and_D2 (__global ushort *src1, int src1_step, int src1_offset,
__kernel void arithm_bitwise_binary_D2 (__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
@ -202,7 +209,7 @@ __kernel void arithm_bitwise_and_D2 (__global ushort *src1, int src1_step, int s
src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
}
ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
ushort4 tmp_data = src1_data & src2_data;
ushort4 tmp_data = src1_data OP_BINARY src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
@ -215,7 +222,7 @@ __kernel void arithm_bitwise_and_D2 (__global ushort *src1, int src1_step, int s
__kernel void arithm_bitwise_and_D3 (__global short *src1, int src1_step, int src1_offset,
__kernel void arithm_bitwise_binary_D3 (__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
@ -257,7 +264,7 @@ __kernel void arithm_bitwise_and_D3 (__global short *src1, int src1_step, int sr
src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
}
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 tmp_data = src1_data & src2_data;
short4 tmp_data = src1_data OP_BINARY src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
@ -270,7 +277,7 @@ __kernel void arithm_bitwise_and_D3 (__global short *src1, int src1_step, int sr
__kernel void arithm_bitwise_and_D4 (__global int *src1, int src1_step, int src1_offset,
__kernel void arithm_bitwise_binary_D4 (__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
@ -286,13 +293,13 @@ __kernel void arithm_bitwise_and_D4 (__global int *src1, int src1_step, int src1
int data1 = *((__global int *)((__global char *)src1 + src1_index));
int data2 = *((__global int *)((__global char *)src2 + src2_index));
int tmp = data1 & data2;
int tmp = data1 OP_BINARY data2;
*((__global int *)((__global char *)dst + dst_index)) = tmp;
}
}
__kernel void arithm_bitwise_and_D5 (__global char *src1, int src1_step, int src1_offset,
__kernel void arithm_bitwise_binary_D5 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
@ -308,14 +315,14 @@ __kernel void arithm_bitwise_and_D5 (__global char *src1, int src1_step, int src
char4 data1 = *((__global char4 *)((__global char *)src1 + src1_index));
char4 data2 = *((__global char4 *)((__global char *)src2 + src2_index));
char4 tmp = data1 & data2;
char4 tmp = data1 OP_BINARY data2;
*((__global char4 *)((__global char *)dst + dst_index)) = tmp;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_and_D6 (__global char *src1, int src1_step, int src1_offset,
__kernel void arithm_bitwise_binary_D6 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
@ -332,7 +339,7 @@ __kernel void arithm_bitwise_and_D6 (__global char *src1, int src1_step, int src
char8 data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 data2 = *((__global char8 *)((__global char *)src2 + src2_index));
*((__global char8 *)((__global char *)dst + dst_index)) = data1 & data2;
*((__global char8 *)((__global char *)dst + dst_index)) = data1 OP_BINARY data2;
}
}
#endif

117
modules/ocl/src/opencl/arithm_bitwise_or_mask.cl → modules/ocl/src/opencl/arithm_bitwise_binary_mask.cl
View File

@ -16,6 +16,7 @@
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
// Peng Xiao, pengxiao@outlook.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
@ -49,11 +50,16 @@
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
#ifndef OP_BINARY
#define OP_BINARY &
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_OR////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_or with mask**************************************/
__kernel void arithm_bitwise_or_with_mask_C1_D0 (
////////////////////////////////////////////bitwise_binary////////////////////////////////////////////
//////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_binary with mask**************************************/
__kernel void arithm_bitwise_binary_with_mask_C1_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -85,7 +91,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D0 (
uchar4 mask_data = vload4(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data | src2_data;
uchar4 tmp_data = src1_data OP_BINARY src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
@ -98,7 +104,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D0 (
__kernel void arithm_bitwise_or_with_mask_C1_D1 (
__kernel void arithm_bitwise_binary_with_mask_C1_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -130,7 +136,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D1 (
uchar4 mask_data = vload4(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data | src2_data;
char4 tmp_data = src1_data OP_BINARY src2_data;
data.x = convert_char((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_char((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
@ -143,7 +149,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D1 (
__kernel void arithm_bitwise_or_with_mask_C1_D2 (
__kernel void arithm_bitwise_binary_with_mask_C1_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -161,7 +167,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D2 (
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
#define dst_align ((dst_offset / 2) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@ -175,7 +181,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D2 (
uchar2 mask_data = vload2(0, mask + mask_index);
ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
ushort2 tmp_data = src1_data | src2_data;
ushort2 tmp_data = src1_data OP_BINARY src2_data;
data.x = convert_ushort((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_ushort((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
@ -186,7 +192,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D2 (
__kernel void arithm_bitwise_or_with_mask_C1_D3 (
__kernel void arithm_bitwise_binary_with_mask_C1_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -204,7 +210,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D3 (
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
#define dst_align ((dst_offset / 2) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@ -218,7 +224,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D3 (
uchar2 mask_data = vload2(0, mask + mask_index);
short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
short2 tmp_data = src1_data | src2_data;
short2 tmp_data = src1_data OP_BINARY src2_data;
data.x = convert_short((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_short((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
@ -229,7 +235,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D3 (
__kernel void arithm_bitwise_or_with_mask_C1_D4 (
__kernel void arithm_bitwise_binary_with_mask_C1_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -253,7 +259,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D4 (
int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
int dst_data = *((__global int *)((__global char *)dst + dst_index));
int data = src_data1 | src_data2;
int data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global int *)((__global char *)dst + dst_index)) = data;
@ -262,7 +268,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D4 (
__kernel void arithm_bitwise_or_with_mask_C1_D5 (
__kernel void arithm_bitwise_binary_with_mask_C1_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -286,7 +292,7 @@ __kernel void arithm_bitwise_or_with_mask_C1_D5 (
char4 src_data2 = *((__global char4 *)((__global char *)src2 + src2_index));
char4 dst_data = *((__global char4 *)((__global char *)dst + dst_index));
char4 data = src_data1 | src_data2;
char4 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global char4 *)((__global char *)dst + dst_index)) = data;
@ -294,8 +300,8 @@ __kernel void arithm_bitwise_or_with_mask_C1_D5 (
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_or_with_mask_C1_D6 (
__kernel void arithm_bitwise_binary_with_mask_C1_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -319,17 +325,17 @@ __kernel void arithm_bitwise_or_with_mask_C1_D6 (
char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
char8 dst_data = *((__global char8 *)((__global char *)dst + dst_index));
char8 data = src_data1 | src_data2;
char8 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global char8 *)((__global char *)dst + dst_index)) = data;
}
}
#endif
__kernel void arithm_bitwise_or_with_mask_C2_D0 (
__kernel void arithm_bitwise_binary_with_mask_C2_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -347,7 +353,7 @@ __kernel void arithm_bitwise_or_with_mask_C2_D0 (
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
#define dst_align ((dst_offset / 2) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@ -361,7 +367,7 @@ __kernel void arithm_bitwise_or_with_mask_C2_D0 (
uchar2 mask_data = vload2(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data | src2_data;
uchar4 tmp_data = src1_data OP_BINARY src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
@ -371,7 +377,7 @@ __kernel void arithm_bitwise_or_with_mask_C2_D0 (
}
__kernel void arithm_bitwise_or_with_mask_C2_D1 (
__kernel void arithm_bitwise_binary_with_mask_C2_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -389,7 +395,7 @@ __kernel void arithm_bitwise_or_with_mask_C2_D1 (
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
#define dst_align ((dst_offset / 2) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@ -403,7 +409,7 @@ __kernel void arithm_bitwise_or_with_mask_C2_D1 (
uchar2 mask_data = vload2(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data | src2_data;
char4 tmp_data = src1_data OP_BINARY src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
@ -412,7 +418,7 @@ __kernel void arithm_bitwise_or_with_mask_C2_D1 (
}
}
__kernel void arithm_bitwise_or_with_mask_C2_D2 (
__kernel void arithm_bitwise_binary_with_mask_C2_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -436,13 +442,13 @@ __kernel void arithm_bitwise_or_with_mask_C2_D2 (
ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index));
ushort2 data = src_data1 | src_data2;
ushort2 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global ushort2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_or_with_mask_C2_D3 (
__kernel void arithm_bitwise_binary_with_mask_C2_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -466,13 +472,13 @@ __kernel void arithm_bitwise_or_with_mask_C2_D3 (
short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index));
short2 data = src_data1 | src_data2;
short2 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global short2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_or_with_mask_C2_D4 (
__kernel void arithm_bitwise_binary_with_mask_C2_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -496,13 +502,13 @@ __kernel void arithm_bitwise_or_with_mask_C2_D4 (
int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index));
int2 data = src_data1 | src_data2;
int2 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global int2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_or_with_mask_C2_D5 (
__kernel void arithm_bitwise_binary_with_mask_C2_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -526,14 +532,14 @@ __kernel void arithm_bitwise_or_with_mask_C2_D5 (
char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
char8 dst_data = *((__global char8 *)((__global char *)dst + dst_index));
char8 data = src_data1 | src_data2;
char8 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global char8 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_or_with_mask_C2_D6 (
__kernel void arithm_bitwise_binary_with_mask_C2_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -557,16 +563,15 @@ __kernel void arithm_bitwise_or_with_mask_C2_D6 (
char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
char16 dst_data = *((__global char16 *)((__global char *)dst + dst_index));
char16 data = src_data1 | src_data2;
char16 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
#endif
__kernel void arithm_bitwise_or_with_mask_C4_D0 (
__kernel void arithm_bitwise_binary_with_mask_C4_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -590,7 +595,7 @@ __kernel void arithm_bitwise_or_with_mask_C4_D0 (
uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
uchar4 data = src_data1 | src_data2;
uchar4 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global uchar4 *)(dst + dst_index)) = data;
@ -598,7 +603,7 @@ __kernel void arithm_bitwise_or_with_mask_C4_D0 (
}
__kernel void arithm_bitwise_or_with_mask_C4_D1 (
__kernel void arithm_bitwise_binary_with_mask_C4_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -622,14 +627,14 @@ __kernel void arithm_bitwise_or_with_mask_C4_D1 (
char4 src_data2 = *((__global char4 *)(src2 + src2_index));
char4 dst_data = *((__global char4 *)(dst + dst_index));
char4 data = src_data1 | src_data2;
char4 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_or_with_mask_C4_D2 (
__kernel void arithm_bitwise_binary_with_mask_C4_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -653,13 +658,13 @@ __kernel void arithm_bitwise_or_with_mask_C4_D2 (
ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
ushort4 data = src_data1 | src_data2;
ushort4 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global ushort4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_or_with_mask_C4_D3 (
__kernel void arithm_bitwise_binary_with_mask_C4_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -683,13 +688,13 @@ __kernel void arithm_bitwise_or_with_mask_C4_D3 (
short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 data = src_data1 | src_data2;
short4 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_or_with_mask_C4_D4 (
__kernel void arithm_bitwise_binary_with_mask_C4_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -713,13 +718,13 @@ __kernel void arithm_bitwise_or_with_mask_C4_D4 (
int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index));
int4 data = src_data1 | src_data2;
int4 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global int4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_or_with_mask_C4_D5 (
__kernel void arithm_bitwise_binary_with_mask_C4_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -743,14 +748,14 @@ __kernel void arithm_bitwise_or_with_mask_C4_D5 (
char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
char16 dst_data = *((__global char16 *)((__global char *)dst + dst_index));
char16 data = src_data1 | src_data2;
char16 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_or_with_mask_C4_D6 (
__kernel void arithm_bitwise_binary_with_mask_C4_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -785,10 +790,10 @@ __kernel void arithm_bitwise_or_with_mask_C4_D6 (
char8 dst_data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16));
char8 dst_data_3 = *((__global char8 *)((__global char *)dst + dst_index + 24));
char8 data_0 = src_data1_0 | src_data2_0;
char8 data_1 = src_data1_1 | src_data2_1;
char8 data_2 = src_data1_2 | src_data2_2;
char8 data_3 = src_data1_3 | src_data2_3;
char8 data_0 = src_data1_0 OP_BINARY src_data2_0;
char8 data_1 = src_data1_1 OP_BINARY src_data2_1;
char8 data_2 = src_data1_2 OP_BINARY src_data2_2;
char8 data_3 = src_data1_3 OP_BINARY src_data2_3;
data_0 = mask_data ? data_0 : dst_data_0;
data_1 = mask_data ? data_1 : dst_data_1;

101
modules/ocl/src/opencl/arithm_bitwise_and_scalar.cl → modules/ocl/src/opencl/arithm_bitwise_binary_scalar.cl
View File

@ -16,6 +16,7 @@
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
// Peng Xiao, pengxiao@outlook.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
@ -50,11 +51,15 @@
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_AND////////////////////////////////////////////////////
#ifndef OP_BINARY
#define OP_BINARY &
#endif
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************and with scalar without mask**************************************/
__kernel void arithm_s_bitwise_and_C1_D0 (
////////////////////////////////////////////bitwise_binary/////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/******************************bitwise binary with scalar without mask********************************/
__kernel void arithm_s_bitwise_binary_C1_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
@ -80,7 +85,7 @@ __kernel void arithm_s_bitwise_and_C1_D0 (
uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data & src2_data;
uchar4 tmp_data = src1_data OP_BINARY src2_data;
data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
@ -92,7 +97,7 @@ __kernel void arithm_s_bitwise_and_C1_D0 (
}
__kernel void arithm_s_bitwise_and_C1_D1 (
__kernel void arithm_s_bitwise_binary_C1_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
@ -118,7 +123,7 @@ __kernel void arithm_s_bitwise_and_C1_D1 (
char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data & src2_data;
char4 tmp_data = src1_data OP_BINARY src2_data;
data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
@ -129,7 +134,7 @@ __kernel void arithm_s_bitwise_and_C1_D1 (
}
}
__kernel void arithm_s_bitwise_and_C1_D2 (
__kernel void arithm_s_bitwise_binary_C1_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
@ -156,7 +161,7 @@ __kernel void arithm_s_bitwise_and_C1_D2 (
ushort2 src2_data = (ushort2)(src2.x, src2.x);
ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
ushort2 tmp_data = src1_data & src2_data;
ushort2 tmp_data = src1_data OP_BINARY src2_data;
data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
data.y = (dst_index + 2 < dst_end ) ? tmp_data.y : data.y;
@ -164,7 +169,7 @@ __kernel void arithm_s_bitwise_and_C1_D2 (
*((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C1_D3 (
__kernel void arithm_s_bitwise_binary_C1_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
@ -191,7 +196,7 @@ __kernel void arithm_s_bitwise_and_C1_D3 (
short2 src2_data = (short2)(src2.x, src2.x);
short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
short2 tmp_data = src1_data & src2_data;
short2 tmp_data = src1_data OP_BINARY src2_data;
data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
data.y = (dst_index + 2 < dst_end ) ? tmp_data.y : data.y;
@ -199,7 +204,7 @@ __kernel void arithm_s_bitwise_and_C1_D3 (
*((__global short2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C1_D4 (
__kernel void arithm_s_bitwise_binary_C1_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
@ -216,12 +221,12 @@ __kernel void arithm_s_bitwise_and_C1_D4 (
int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
int src_data2 = src2.x;
int data = src_data1 & src_data2;
int data = src_data1 OP_BINARY src_data2;
*((__global int *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C1_D5 (
__kernel void arithm_s_bitwise_binary_C1_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
@ -242,7 +247,7 @@ __kernel void arithm_s_bitwise_and_C1_D5 (
char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
char4 data = *((__global char4 *)((__global char *)dst + dst_index));
char4 tmp_data = src1_data & src2_data;
char4 tmp_data = src1_data OP_BINARY src2_data;
data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
@ -253,7 +258,7 @@ __kernel void arithm_s_bitwise_and_C1_D5 (
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_and_C1_D6 (
__kernel void arithm_s_bitwise_binary_C1_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
@ -270,13 +275,13 @@ __kernel void arithm_s_bitwise_and_C1_D6 (
short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
short4 tmp_data = src1_data & src2_data;
short4 tmp_data = src1_data OP_BINARY src2_data;
*((__global short4 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#endif
__kernel void arithm_s_bitwise_and_C2_D0 (
__kernel void arithm_s_bitwise_binary_C2_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
@ -303,7 +308,7 @@ __kernel void arithm_s_bitwise_and_C2_D0 (
uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data & src2_data;
uchar4 tmp_data = src1_data OP_BINARY src2_data;
data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
@ -314,7 +319,7 @@ __kernel void arithm_s_bitwise_and_C2_D0 (
}
__kernel void arithm_s_bitwise_and_C2_D1 (
__kernel void arithm_s_bitwise_binary_C2_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
@ -341,7 +346,7 @@ __kernel void arithm_s_bitwise_and_C2_D1 (
char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data & src2_data;
char4 tmp_data = src1_data OP_BINARY src2_data;
data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
data.zw = (dst_index + 2 < dst_end ) ? tmp_data.zw : data.zw;
@ -350,7 +355,7 @@ __kernel void arithm_s_bitwise_and_C2_D1 (
}
}
__kernel void arithm_s_bitwise_and_C2_D2 (
__kernel void arithm_s_bitwise_binary_C2_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
@ -367,12 +372,12 @@ __kernel void arithm_s_bitwise_and_C2_D2 (
ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
ushort2 src_data2 = (ushort2)(src2.x, src2.y);
ushort2 data = src_data1 & src_data2;
ushort2 data = src_data1 OP_BINARY src_data2;
*((__global ushort2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C2_D3 (
__kernel void arithm_s_bitwise_binary_C2_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
@ -389,12 +394,12 @@ __kernel void arithm_s_bitwise_and_C2_D3 (
short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
short2 src_data2 = (short2)(src2.x, src2.y);
short2 data = src_data1 & src_data2;
short2 data = src_data1 OP_BINARY src_data2;
*((__global short2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C2_D4 (
__kernel void arithm_s_bitwise_binary_C2_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
@ -411,11 +416,11 @@ __kernel void arithm_s_bitwise_and_C2_D4 (
int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
int2 src_data2 = (int2)(src2.x, src2.y);
int2 data = src_data1 & src_data2;
int2 data = src_data1 OP_BINARY src_data2;
*((__global int2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C2_D5 (
__kernel void arithm_s_bitwise_binary_C2_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
@ -432,13 +437,13 @@ __kernel void arithm_s_bitwise_and_C2_D5 (
char8 src1_data = *((__global char8 *)((__global char *)src1 + src1_index));
char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
char8 tmp_data = src1_data & src2_data;
char8 tmp_data = src1_data OP_BINARY src2_data;
*((__global char8 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_and_C2_D6 (
__kernel void arithm_s_bitwise_binary_C2_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
@ -455,14 +460,14 @@ __kernel void arithm_s_bitwise_and_C2_D6 (
short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
short8 tmp_data = src1_data & src2_data;
short8 tmp_data = src1_data OP_BINARY src2_data;
*((__global short8 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#endif
__kernel void arithm_s_bitwise_and_C4_D0 (
__kernel void arithm_s_bitwise_binary_C4_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
@ -478,14 +483,14 @@ __kernel void arithm_s_bitwise_and_C4_D0 (
uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
uchar4 data = src_data1 & src2;
uchar4 data = src_data1 OP_BINARY src2;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C4_D1 (
__kernel void arithm_s_bitwise_binary_C4_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
@ -501,13 +506,13 @@ __kernel void arithm_s_bitwise_and_C4_D1 (
char4 src_data1 = *((__global char4 *)(src1 + src1_index));
char4 data = src_data1 & src2;
char4 data = src_data1 OP_BINARY src2;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C4_D2 (
__kernel void arithm_s_bitwise_binary_C4_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
@ -523,12 +528,12 @@ __kernel void arithm_s_bitwise_and_C4_D2 (
ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
ushort4 data = src_data1 & src2;
ushort4 data = src_data1 OP_BINARY src2;
*((__global ushort4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C4_D3 (
__kernel void arithm_s_bitwise_binary_C4_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
@ -544,12 +549,12 @@ __kernel void arithm_s_bitwise_and_C4_D3 (
short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
short4 data = src_data1 & src2;
short4 data = src_data1 OP_BINARY src2;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C4_D4 (
__kernel void arithm_s_bitwise_binary_C4_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
@ -565,12 +570,12 @@ __kernel void arithm_s_bitwise_and_C4_D4 (
int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
int4 data = src_data1 & src2;
int4 data = src_data1 OP_BINARY src2;
*((__global int4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_and_C4_D5 (
__kernel void arithm_s_bitwise_binary_C4_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
@ -588,13 +593,13 @@ __kernel void arithm_s_bitwise_and_C4_D5 (
char16 src2_data = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
char16 tmp_data = src1_data & src2_data;
char16 tmp_data = src1_data OP_BINARY src2_data;
*((__global char16 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_and_C4_D6 (
__kernel void arithm_s_bitwise_binary_C4_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
@ -618,10 +623,10 @@ __kernel void arithm_s_bitwise_and_C4_D6 (
short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
short4 tmp_data_0 = src1_data_0 & src2_data_0;
short4 tmp_data_1 = src1_data_1 & src2_data_1;
short4 tmp_data_2 = src1_data_2 & src2_data_2;
short4 tmp_data_3 = src1_data_3 & src2_data_3;
short4 tmp_data_0 = src1_data_0 OP_BINARY src2_data_0;
short4 tmp_data_1 = src1_data_1 OP_BINARY src2_data_1;
short4 tmp_data_2 = src1_data_2 OP_BINARY src2_data_2;
short4 tmp_data_3 = src1_data_3 OP_BINARY src2_data_3;
*((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
*((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;

112
modules/ocl/src/opencl/arithm_bitwise_xor_scalar_mask.cl → modules/ocl/src/opencl/arithm_bitwise_binary_scalar_mask.cl
View File

@ -42,7 +42,6 @@
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable
@ -50,11 +49,16 @@
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
#ifndef OP_BINARY
#define OP_BINARY &
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_XOR////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_xor with scalar with mask**************************************/
__kernel void arithm_s_bitwise_xor_with_mask_C1_D0 (
////////////////////////////////////////////bitwise_binary////////////////////////////////////////////
//////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_binary with scalar with mask**************************************/
__kernel void arithm_s_bitwise_binary_with_mask_C1_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -84,7 +88,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D0 (
uchar4 mask_data = vload4(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data ^ src2_data;
uchar4 tmp_data = src1_data OP_BINARY src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
@ -96,7 +100,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D0 (
}
__kernel void arithm_s_bitwise_xor_with_mask_C1_D1 (
__kernel void arithm_s_bitwise_binary_with_mask_C1_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -126,7 +130,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D1 (
uchar4 mask_data = vload4(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data ^ src2_data;
char4 tmp_data = src1_data OP_BINARY src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
@ -137,7 +141,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D1 (
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C1_D2 (
__kernel void arithm_s_bitwise_binary_with_mask_C1_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -154,7 +158,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D2 (
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
#define dst_align ((dst_offset / 2) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@ -167,7 +171,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D2 (
uchar2 mask_data = vload2(0, mask + mask_index);
ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
ushort2 tmp_data = src1_data ^ src2_data;
ushort2 tmp_data = src1_data OP_BINARY src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.y : data.y;
@ -175,7 +179,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D2 (
*((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C1_D3 (
__kernel void arithm_s_bitwise_binary_with_mask_C1_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -192,7 +196,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D3 (
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
#define dst_align ((dst_offset / 2) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@ -205,7 +209,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D3 (
uchar2 mask_data = vload2(0, mask + mask_index);
short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
short2 tmp_data = src1_data ^ src2_data;
short2 tmp_data = src1_data OP_BINARY src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.y : data.y;
@ -213,7 +217,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D3 (
*((__global short2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C1_D4 (
__kernel void arithm_s_bitwise_binary_with_mask_C1_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -235,14 +239,14 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D4 (
int src_data2 = src2.x;
int dst_data = *((__global int *)((__global char *)dst + dst_index));
int data = src_data1 ^ src_data2;
int data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global int *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C1_D5 (
__kernel void arithm_s_bitwise_binary_with_mask_C1_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -264,7 +268,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D5 (
char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
char4 dst_data = *((__global char4 *)((__global char *)dst + dst_index));
char4 data = src1_data ^ src2_data;
char4 data = src1_data OP_BINARY src2_data;
data = mask_data ? data : dst_data;
*((__global char4 *)((__global char *)dst + dst_index)) = data;
@ -272,7 +276,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D5 (
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_xor_with_mask_C1_D6 (
__kernel void arithm_s_bitwise_binary_with_mask_C1_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -294,14 +298,14 @@ __kernel void arithm_s_bitwise_xor_with_mask_C1_D6 (
short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 data = src1_data ^ src2_data;
short4 data = src1_data OP_BINARY src2_data;
data = mask_data ? data : dst_data;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
#endif
__kernel void arithm_s_bitwise_xor_with_mask_C2_D0 (
__kernel void arithm_s_bitwise_binary_with_mask_C2_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -318,7 +322,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D0 (
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
#define dst_align ((dst_offset / 2) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@ -331,7 +335,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D0 (
uchar2 mask_data = vload2(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data ^ src2_data;
uchar4 tmp_data = src1_data OP_BINARY src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
@ -341,7 +345,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D0 (
}
__kernel void arithm_s_bitwise_xor_with_mask_C2_D1 (
__kernel void arithm_s_bitwise_binary_with_mask_C2_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -358,7 +362,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D1 (
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
#define dst_align ((dst_offset / 2) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
@ -371,7 +375,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D1 (
uchar2 mask_data = vload2(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data ^ src2_data;
char4 tmp_data = src1_data OP_BINARY src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
@ -380,7 +384,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D1 (
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C2_D2 (
__kernel void arithm_s_bitwise_binary_with_mask_C2_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -402,13 +406,13 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D2 (
ushort2 src_data2 = (ushort2)(src2.x, src2.y);
ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index));
ushort2 data = src_data1 ^ src_data2;
ushort2 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global ushort2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C2_D3 (
__kernel void arithm_s_bitwise_binary_with_mask_C2_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -430,13 +434,13 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D3 (
short2 src_data2 = (short2)(src2.x, src2.y);
short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index));
short2 data = src_data1 ^ src_data2;
short2 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global short2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C2_D4 (
__kernel void arithm_s_bitwise_binary_with_mask_C2_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -458,13 +462,13 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D4 (
int2 src_data2 = (int2)(src2.x, src2.y);
int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index));
int2 data = src_data1 ^ src_data2;
int2 data = src_data1 OP_BINARY src_data2;
data = mask_data ? data : dst_data;
*((__global int2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C2_D5 (
__kernel void arithm_s_bitwise_binary_with_mask_C2_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -486,7 +490,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D5 (
char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
char8 dst_data = *((__global char8 *)((__global char *)dst + dst_index));
char8 data = src1_data ^ src2_data;
char8 data = src1_data OP_BINARY src2_data;
data = mask_data ? data : dst_data;
@ -494,7 +498,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D5 (
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_xor_with_mask_C2_D6 (
__kernel void arithm_s_bitwise_binary_with_mask_C2_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -516,15 +520,14 @@ __kernel void arithm_s_bitwise_xor_with_mask_C2_D6 (
short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
short8 dst_data = *((__global short8 *)((__global char *)dst + dst_index));
short8 data = src1_data ^ src2_data;
short8 data = src1_data OP_BINARY src2_data;
data = mask_data ? data : dst_data;
*((__global short8 *)((__global char *)dst + dst_index)) = data;
}
}
#endif
__kernel void arithm_s_bitwise_xor_with_mask_C4_D0 (
__kernel void arithm_s_bitwise_binary_with_mask_C4_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -545,7 +548,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D0 (
uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
uchar4 data = src_data1 ^ src2;
uchar4 data = src_data1 OP_BINARY src2;
data = mask_data ? data : dst_data;
*((__global uchar4 *)(dst + dst_index)) = data;
@ -553,7 +556,7 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D0 (
}
__kernel void arithm_s_bitwise_xor_with_mask_C4_D1 (
__kernel void arithm_s_bitwise_binary_with_mask_C4_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -574,14 +577,14 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D1 (
char4 src_data1 = *((__global char4 *)(src1 + src1_index));
char4 dst_data = *((__global char4 *)(dst + dst_index));
char4 data = src_data1 ^ src2;
char4 data = src_data1 OP_BINARY src2;
data = mask_data ? data : dst_data;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C4_D2 (
__kernel void arithm_s_bitwise_binary_with_mask_C4_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -602,13 +605,13 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D2 (
ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
ushort4 data = src_data1 ^ src2;
ushort4 data = src_data1 OP_BINARY src2;
data = mask_data ? data : dst_data;
*((__global ushort4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C4_D3 (
__kernel void arithm_s_bitwise_binary_with_mask_C4_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -629,13 +632,13 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D3 (
short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 data = src_data1 ^ src2;
short4 data = src_data1 OP_BINARY src2;
data = mask_data ? data : dst_data;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C4_D4 (
__kernel void arithm_s_bitwise_binary_with_mask_C4_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -656,13 +659,13 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D4 (
int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index));
int4 data = src_data1 ^ src2;
int4 data = src_data1 OP_BINARY src2;
data = mask_data ? data : dst_data;
*((__global int4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_with_mask_C4_D5 (
__kernel void arithm_s_bitwise_binary_with_mask_C4_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -685,14 +688,14 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D5 (
src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
char16 dst_data = *((__global char16 *)((__global char *)dst + dst_index));
char16 data = src1_data ^ src2_data;
char16 data = src1_data OP_BINARY src2_data;
data = mask_data ? data : dst_data;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_xor_with_mask_C4_D6 (
__kernel void arithm_s_bitwise_binary_with_mask_C4_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
@ -725,10 +728,10 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D6 (
short4 dst_data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16));
short4 dst_data_3 = *((__global short4 *)((__global char *)dst + dst_index + 24));
short4 data_0 = src1_data_0 ^ src2_data_0;
short4 data_1 = src1_data_1 ^ src2_data_1;
short4 data_2 = src1_data_2 ^ src2_data_2;
short4 data_3 = src1_data_3 ^ src2_data_3;
short4 data_0 = src1_data_0 OP_BINARY src2_data_0;
short4 data_1 = src1_data_1 OP_BINARY src2_data_1;
short4 data_2 = src1_data_2 OP_BINARY src2_data_2;
short4 data_3 = src1_data_3 OP_BINARY src2_data_3;
data_0 = mask_data ? data_0 : dst_data_0;
data_1 = mask_data ? data_1 : dst_data_1;
@ -742,3 +745,4 @@ __kernel void arithm_s_bitwise_xor_with_mask_C4_D6 (
}
}
#endif

294
modules/ocl/src/opencl/arithm_bitwise_or.cl
View File

@ -1,294 +0,0 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other oclMaterials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable
#elif defined (cl_amd_fp64)
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_OR////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_or without mask**************************************/
__kernel void arithm_bitwise_or_D0 (__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
int src1_index_fix = src1_index < 0 ? 0 : src1_index;
int src2_index_fix = src2_index < 0 ? 0 : src2_index;
uchar4 src1_data = vload4(0, src1 + src1_index_fix);
uchar4 src2_data = vload4(0, src2 + src2_index_fix);
if(src1_index < 0)
{
uchar4 tmp;
tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
}
if(src2_index < 0)
{
uchar4 tmp;
tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
}
uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data | src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
*((__global uchar4 *)(dst + dst_index)) = dst_data;
}
}
__kernel void arithm_bitwise_or_D1 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = vload4(0, src2 + src2_index);
char4 dst_data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data | src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
*((__global char4 *)(dst + dst_index)) = dst_data;
}
}
__kernel void arithm_bitwise_or_D2 (__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 3)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index));
ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index));
ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
ushort4 tmp_data = src1_data | src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
*((__global ushort4 *)((__global char *)dst + dst_index)) = dst_data;
}
}
__kernel void arithm_bitwise_or_D3 (__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 3)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index));
short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index));
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 tmp_data = src1_data | src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
*((__global short4 *)((__global char *)dst + dst_index)) = dst_data;
}
}
__kernel void arithm_bitwise_or_D4 (__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
int data1 = *((__global int *)((__global char *)src1 + src1_index));
int data2 = *((__global int *)((__global char *)src2 + src2_index));
int tmp = data1 | data2;
*((__global int *)((__global char *)dst + dst_index)) = tmp;
}
}
__kernel void arithm_bitwise_or_D5 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
char4 data1 = *((__global char4 *)((__global char *)src1 + src1_index));
char4 data2 = *((__global char4 *)((__global char *)src2 + src2_index));
char4 tmp = data1 | data2;
*((__global char4 *)((__global char *)dst + dst_index)) = tmp;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_or_D6 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
char8 data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 data2 = *((__global char8 *)((__global char *)src2 + src2_index));
*((__global char8 *)((__global char *)dst + dst_index)) = data1 | data2;
}
}
#endif

636
modules/ocl/src/opencl/arithm_bitwise_or_scalar.cl
View File

@ -1,636 +0,0 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other oclMaterials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable
#elif defined (cl_amd_fp64)
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_OR////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************and with scalar without mask**************************************/
__kernel void arithm_s_bitwise_or_C1_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data | src2_data;
data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C1_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data | src2_data;
data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C1_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
ushort2 src2_data = (ushort2)(src2.x, src2.x);
ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
ushort2 tmp_data = src1_data | src2_data;
data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
data.y = (dst_index + 2 < dst_end ) ? tmp_data.y : data.y;
*((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C1_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
short2 src2_data = (short2)(src2.x, src2.x);
short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
short2 tmp_data = src1_data | src2_data;
data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
data.y = (dst_index + 2 < dst_end ) ? tmp_data.y : data.y;
*((__global short2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C1_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
int src_data2 = src2.x;
int data = src_data1 | src_data2;
*((__global int *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C1_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
char4 src_data2 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
char4 data = src_data1 | src_data2;
*((__global char4 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_or_C1_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
short4 tmp_data = src1_data | src2_data;
*((__global short4 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#endif
__kernel void arithm_s_bitwise_or_C2_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data | src2_data;
data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
data.zw = (dst_index + 2 < dst_end ) ? tmp_data.zw : data.zw;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C2_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data | src2_data;
data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
data.zw = (dst_index + 2 < dst_end ) ? tmp_data.zw : data.zw;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C2_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
ushort2 src_data2 = (ushort2)(src2.x, src2.y);
ushort2 data = src_data1 | src_data2;
*((__global ushort2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C2_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
short2 src_data2 = (short2)(src2.x, src2.y);
short2 data = src_data1 | src_data2;
*((__global short2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C2_D4 (__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
int2 src_data2 = (int2)(src2.x, src2.y);
int2 data = src_data1 | src_data2;
*((__global int2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C2_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 src_data2 = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
char8 data = src_data1 | src_data2;
*((__global char8 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_or_C2_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
short8 tmp_data = src1_data & src2_data;
*((__global short8 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#endif
__kernel void arithm_s_bitwise_or_C4_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
uchar4 data = src_data1 | src2;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C4_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
char4 src_data1 = *((__global char4 *)(src1 + src1_index));
char4 data = src_data1 | src2;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C4_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
ushort4 data = src_data1 | src2;
*((__global ushort4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C4_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
short4 data = src_data1 | src2;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C4_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
int4 data = src_data1 | src2;
*((__global int4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_C4_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
char16 src_data2 = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
char16 data = src_data1 | src_data2;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_or_C4_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 5) + dst_offset);
short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
short4 tmp_data_0 = src1_data_0 | src2_data_0;
short4 tmp_data_1 = src1_data_1 | src2_data_1;
short4 tmp_data_2 = src1_data_2 | src2_data_2;
short4 tmp_data_3 = src1_data_3 | src2_data_3;
*((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
*((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
*((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
*((__global short4 *)((__global char *)dst + dst_index + 24))= tmp_data_3;
}
}
#endif

760
modules/ocl/src/opencl/arithm_bitwise_or_scalar_mask.cl
View File

@ -1,760 +0,0 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other GpuMaterials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable
#elif defined (cl_amd_fp64)
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_OR////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_or with scalar with mask**************************************/
__kernel void arithm_s_bitwise_or_with_mask_C1_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
uchar4 mask_data = vload4(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data | src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C1_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
uchar4 mask_data = vload4(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data | src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C1_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
ushort2 src2_data = (ushort2)(src2.x, src2.x);
uchar2 mask_data = vload2(0, mask + mask_index);
ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
ushort2 tmp_data = src1_data | src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.y : data.y;
*((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C1_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
short2 src2_data = (short2)(src2.x, src2.x);
uchar2 mask_data = vload2(0, mask + mask_index);
short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
short2 tmp_data = src1_data | src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.y : data.y;
*((__global short2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C1_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
int src_data2 = src2.x;
int dst_data = *((__global int *)((__global char *)dst + dst_index));
int data = src_data1 | src_data2;
data = mask_data ? data : dst_data;
*((__global int *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C1_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
char4 src_data2 = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
char4 dst_data = *((__global char4 *)((__global char *)dst + dst_index));
char4 data = src_data1 | src_data2;
data = mask_data ? data : dst_data;
*((__global char4 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_or_with_mask_C1_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 data = src1_data | src2_data;
data = mask_data ? data : dst_data;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
#endif
__kernel void arithm_s_bitwise_or_with_mask_C2_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
uchar2 mask_data = vload2(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data | src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C2_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
uchar2 mask_data = vload2(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data | src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C2_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
ushort2 src_data2 = (ushort2)(src2.x, src2.y);
ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index));
ushort2 data = src_data1 | src_data2;
data = mask_data ? data : dst_data;
*((__global ushort2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C2_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
short2 src_data2 = (short2)(src2.x, src2.y);
short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index));
short2 data = src_data1 | src_data2;
data = mask_data ? data : dst_data;
*((__global short2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C2_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
int2 src_data2 = (int2)(src2.x, src2.y);
int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index));
int2 data = src_data1 | src_data2;
data = mask_data ? data : dst_data;
*((__global int2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C2_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 src_data2 = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
char8 dst_data = *((__global char8 *)((__global char *)dst + dst_index));
char8 data = src_data1 | src_data2;
data = mask_data ? data : dst_data;
*((__global char8 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_or_with_mask_C2_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
short8 dst_data = *((__global short8 *)((__global char *)dst + dst_index));
short8 data = src1_data | src2_data;
data = mask_data ? data : dst_data;
*((__global short8 *)((__global char *)dst + dst_index)) = data;
}
}
#endif
__kernel void arithm_s_bitwise_or_with_mask_C4_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
uchar4 data = src_data1 | src2;
data = mask_data ? data : dst_data;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C4_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
char4 src_data1 = *((__global char4 *)(src1 + src1_index));
char4 dst_data = *((__global char4 *)(dst + dst_index));
char4 data = src_data1 | src2;
data = mask_data ? data : dst_data;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C4_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
ushort4 data = src_data1 | src2;
data = mask_data ? data : dst_data;
*((__global ushort4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C4_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 data = src_data1 | src2;
data = mask_data ? data : dst_data;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C4_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index));
int4 data = src_data1 | src2;
data = mask_data ? data : dst_data;
*((__global int4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_or_with_mask_C4_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
char16 src_data2 = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
char16 dst_data = *((__global char16 *)((__global char *)dst + dst_index));
char16 data = src_data1 | src_data2;
data = mask_data ? data : dst_data;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_or_with_mask_C4_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
__global uchar *mask, int mask_step, int mask_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 5) + dst_offset);
uchar mask_data = *(mask + mask_index);
short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
short4 dst_data_0 = *((__global short4 *)((__global char *)dst + dst_index + 0));
short4 dst_data_1 = *((__global short4 *)((__global char *)dst + dst_index + 8));
short4 dst_data_2 = *((__global short4 *)((__global char *)dst + dst_index + 16));
short4 dst_data_3 = *((__global short4 *)((__global char *)dst + dst_index + 24));
short4 data_0 = src1_data_0 | src2_data_0;
short4 data_1 = src1_data_1 | src2_data_1;
short4 data_2 = src1_data_2 | src2_data_2;
short4 data_3 = src1_data_3 | src2_data_3;
data_0 = mask_data ? data_0 : dst_data_0;
data_1 = mask_data ? data_1 : dst_data_1;
data_2 = mask_data ? data_2 : dst_data_2;
data_3 = mask_data ? data_3 : dst_data_3;
*((__global short4 *)((__global char *)dst + dst_index + 0)) = data_0;
*((__global short4 *)((__global char *)dst + dst_index + 8)) = data_1;
*((__global short4 *)((__global char *)dst + dst_index + 16)) = data_2;
*((__global short4 *)((__global char *)dst + dst_index + 24)) = data_3;
}
}
#endif

340
modules/ocl/src/opencl/arithm_bitwise_xor.cl
View File

@ -1,340 +0,0 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other GpuMaterials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable
#elif defined (cl_amd_fp64)
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_XOR////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_xor without mask**************************************/
__kernel void arithm_bitwise_xor_D0 (__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
int src1_index_fix = src1_index < 0 ? 0 : src1_index;
int src2_index_fix = src2_index < 0 ? 0 : src2_index;
uchar4 src1_data = vload4(0, src1 + src1_index_fix);
uchar4 src2_data = vload4(0, src2 + src2_index_fix);
if(src1_index < 0)
{
uchar4 tmp;
tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
}
if(src2_index < 0)
{
uchar4 tmp;
tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
}
uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data ^ src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
*((__global uchar4 *)(dst + dst_index)) = dst_data;
}
}
__kernel void arithm_bitwise_xor_D1 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
int src1_index_fix = src1_index < 0 ? 0 : src1_index;
int src2_index_fix = src2_index < 0 ? 0 : src2_index;
char4 src1_data = vload4(0, src1 + src1_index_fix);
char4 src2_data = vload4(0, src2 + src2_index_fix);
if(src1_index < 0)
{
char4 tmp;
tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
}
if(src2_index < 0)
{
char4 tmp;
tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
}
char4 dst_data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data ^ src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : dst_data.y;
dst_data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : dst_data.z;
dst_data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : dst_data.w;
*((__global char4 *)(dst + dst_index)) = dst_data;
}
}
__kernel void arithm_bitwise_xor_D2 (__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 3)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
int src1_index_fix = src1_index < 0 ? 0 : src1_index;
int src2_index_fix = src2_index < 0 ? 0 : src2_index;
ushort4 src1_data = vload4(0, (__global ushort *)((__global char *)src1 + src1_index_fix));
ushort4 src2_data = vload4(0, (__global ushort *)((__global char *)src2 + src2_index_fix));
if(src1_index < 0)
{
ushort4 tmp;
tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
}
if(src2_index < 0)
{
ushort4 tmp;
tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
}
ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
ushort4 tmp_data = src1_data ^ src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
*((__global ushort4 *)((__global char *)dst + dst_index)) = dst_data;
}
}
__kernel void arithm_bitwise_xor_D3 (__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 3)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffff8);
int src1_index_fix = src1_index < 0 ? 0 : src1_index;
int src2_index_fix = src2_index < 0 ? 0 : src2_index;
short4 src1_data = vload4(0, (__global short *)((__global char *)src1 + src1_index_fix));
short4 src2_data = vload4(0, (__global short *)((__global char *)src2 + src2_index_fix));
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
if(src1_index < 0)
{
short4 tmp;
tmp.xyzw = (src1_index == -2) ? src1_data.zwxy:src1_data.yzwx;
src1_data.xyzw = (src1_index == -1) ? src1_data.wxyz:tmp.xyzw;
}
if(src2_index < 0)
{
short4 tmp;
tmp.xyzw = (src2_index == -2) ? src2_data.zwxy:src2_data.yzwx;
src2_data.xyzw = (src2_index == -1) ? src2_data.wxyz:tmp.xyzw;
}
short4 tmp_data = src1_data ^ src2_data;
dst_data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : dst_data.x;
dst_data.y = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : dst_data.y;
dst_data.z = ((dst_index + 4 >= dst_start) && (dst_index + 4 < dst_end)) ? tmp_data.z : dst_data.z;
dst_data.w = ((dst_index + 6 >= dst_start) && (dst_index + 6 < dst_end)) ? tmp_data.w : dst_data.w;
*((__global short4 *)((__global char *)dst + dst_index)) = dst_data;
}
}
__kernel void arithm_bitwise_xor_D4 (__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
int data1 = *((__global int *)((__global char *)src1 + src1_index));
int data2 = *((__global int *)((__global char *)src2 + src2_index));
int tmp = data1 ^ data2;
*((__global int *)((__global char *)dst + dst_index)) = tmp;
}
}
__kernel void arithm_bitwise_xor_D5 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
char4 data1 = *((__global char4 *)((__global char *)src1 + src1_index));
char4 data2 = *((__global char4 *)((__global char *)src2 + src2_index));
char4 tmp = data1 ^ data2;
*((__global char4 *)((__global char *)dst + dst_index)) = tmp;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_xor_D6 (__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
char8 data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 data2 = *((__global char8 *)((__global char *)src2 + src2_index));
*((__global char8 *)((__global char *)dst + dst_index)) = data1 ^ data2;
}
}
#endif

803
modules/ocl/src/opencl/arithm_bitwise_xor_mask.cl
View File

@ -1,803 +0,0 @@
/*M///////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other oclMaterials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//M*/
#if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable
#elif defined (cl_amd_fp64)
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_XOR////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************bitwise_xor with mask**************************************/
__kernel void arithm_bitwise_xor_with_mask_C1_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = vload4(0, src2 + src2_index);
uchar4 mask_data = vload4(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data ^ src2_data;
data.x = ((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C1_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int src2_index = mad24(y, src2_step, x + src2_offset - dst_align);
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = vload4(0, src2 + src2_index);
uchar4 mask_data = vload4(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data ^ src2_data;
data.x = convert_char((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_char((mask_data.y) && (dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = convert_char((mask_data.z) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = convert_char((mask_data.w) && (dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C1_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
ushort2 src2_data = vload2(0, (__global ushort *)((__global char *)src2 + src2_index));
uchar2 mask_data = vload2(0, mask + mask_index);
ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
ushort2 tmp_data = src1_data ^ src2_data;
data.x = convert_ushort((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_ushort((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
*((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C1_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
short2 src2_data = vload2(0, (__global short *)((__global char *)src2 + src2_index));
uchar2 mask_data = vload2(0, mask + mask_index);
short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
short2 tmp_data = src1_data ^ src2_data;
data.x = convert_short((mask_data.x) && (dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = convert_short((mask_data.y) && (dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.y : data.y;
*((__global short2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C1_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
int src_data2 = *((__global int *)((__global char *)src2 + src2_index));
int dst_data = *((__global int *)((__global char *)dst + dst_index));
int data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global int *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C1_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
char4 src_data1 = *((__global char4 *)((__global char *)src1 + src1_index));
char4 src_data2 = *((__global char4 *)((__global char *)src2 + src2_index));
char4 dst_data = *((__global char4 *)((__global char *)dst + dst_index));
char4 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global char4 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_xor_with_mask_C1_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
char8 dst_data = *((__global char8 *)((__global char *)dst + dst_index));
char8 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global char8 *)((__global char *)dst + dst_index)) = data;
}
}
#endif
__kernel void arithm_bitwise_xor_with_mask_C2_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = vload4(0, src2 + src2_index);
uchar2 mask_data = vload2(0, mask + mask_index);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data ^ src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C2_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int src2_index = mad24(y, src2_step, (x << 1) + src2_offset - (dst_align << 1));
int mask_index = mad24(y, mask_step, x + mask_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = vload4(0, src2 + src2_index);
uchar2 mask_data = vload2(0, mask + mask_index);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data ^ src2_data;
data.xy = ((mask_data.x) && (dst_index + 0 >= dst_start)) ? tmp_data.xy : data.xy;
data.zw = ((mask_data.y) && (dst_index + 2 < dst_end )) ? tmp_data.zw : data.zw;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C2_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
ushort2 src_data2 = *((__global ushort2 *)((__global char *)src2 + src2_index));
ushort2 dst_data = *((__global ushort2 *)((__global char *)dst + dst_index));
ushort2 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global ushort2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C2_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
short2 src_data2 = *((__global short2 *)((__global char *)src2 + src2_index));
short2 dst_data = *((__global short2 *)((__global char *)dst + dst_index));
short2 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global short2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C2_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
int2 src_data2 = *((__global int2 *)((__global char *)src2 + src2_index));
int2 dst_data = *((__global int2 *)((__global char *)dst + dst_index));
int2 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global int2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C2_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
char8 src_data1 = *((__global char8 *)((__global char *)src1 + src1_index));
char8 src_data2 = *((__global char8 *)((__global char *)src2 + src2_index));
char8 dst_data = *((__global char8 *)((__global char *)dst + dst_index));
char8 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global char8 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_xor_with_mask_C2_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
char16 dst_data = *((__global char16 *)((__global char *)dst + dst_index));
char16 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
#endif
__kernel void arithm_bitwise_xor_with_mask_C4_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global uchar *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
uchar4 src_data2 = *((__global uchar4 *)(src2 + src2_index));
uchar4 dst_data = *((__global uchar4 *)(dst + dst_index));
uchar4 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C4_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 2) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar mask_data = *(mask + mask_index);
char4 src_data1 = *((__global char4 *)(src1 + src1_index));
char4 src_data2 = *((__global char4 *)(src2 + src2_index));
char4 dst_data = *((__global char4 *)(dst + dst_index));
char4 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C4_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global ushort *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
ushort4 src_data2 = *((__global ushort4 *)((__global char *)src2 + src2_index));
ushort4 dst_data = *((__global ushort4 *)((__global char *)dst + dst_index));
ushort4 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global ushort4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C4_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global short *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 3) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
uchar mask_data = *(mask + mask_index);
short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
short4 src_data2 = *((__global short4 *)((__global char *)src2 + src2_index));
short4 dst_data = *((__global short4 *)((__global char *)dst + dst_index));
short4 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C4_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global int *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
int4 src_data2 = *((__global int4 *)((__global char *)src2 + src2_index));
int4 dst_data = *((__global int4 *)((__global char *)dst + dst_index));
int4 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global int4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_bitwise_xor_with_mask_C4_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 4) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
uchar mask_data = *(mask + mask_index);
char16 src_data1 = *((__global char16 *)((__global char *)src1 + src1_index));
char16 src_data2 = *((__global char16 *)((__global char *)src2 + src2_index));
char16 dst_data = *((__global char16 *)((__global char *)dst + dst_index));
char16 data = src_data1 ^ src_data2;
data = mask_data ? data : dst_data;
*((__global char16 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_bitwise_xor_with_mask_C4_D6 (
__global char *src1, int src1_step, int src1_offset,
__global char *src2, int src2_step, int src2_offset,
__global uchar *mask, int mask_step, int mask_offset,
__global char *dst, int dst_step, int dst_offset,
int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
int src2_index = mad24(y, src2_step, (x << 5) + src2_offset);
int mask_index = mad24(y, mask_step, x + mask_offset);
int dst_index = mad24(y, dst_step, (x << 5) + dst_offset);
uchar mask_data = *(mask + mask_index);
char8 src_data1_0 = *((__global char8 *)((__global char *)src1 + src1_index + 0));
char8 src_data1_1 = *((__global char8 *)((__global char *)src1 + src1_index + 8));
char8 src_data1_2 = *((__global char8 *)((__global char *)src1 + src1_index + 16));
char8 src_data1_3 = *((__global char8 *)((__global char *)src1 + src1_index + 24));
char8 src_data2_0 = *((__global char8 *)((__global char *)src2 + src2_index + 0));
char8 src_data2_1 = *((__global char8 *)((__global char *)src2 + src2_index + 8));
char8 src_data2_2 = *((__global char8 *)((__global char *)src2 + src2_index + 16));
char8 src_data2_3 = *((__global char8 *)((__global char *)src2 + src2_index + 24));
char8 dst_data_0 = *((__global char8 *)((__global char *)dst + dst_index + 0));
char8 dst_data_1 = *((__global char8 *)((__global char *)dst + dst_index + 8));
char8 dst_data_2 = *((__global char8 *)((__global char *)dst + dst_index + 16));
char8 dst_data_3 = *((__global char8 *)((__global char *)dst + dst_index + 24));
char8 data_0 = src_data1_0 ^ src_data2_0;
char8 data_1 = src_data1_1 ^ src_data2_1;
char8 data_2 = src_data1_2 ^ src_data2_2;
char8 data_3 = src_data1_3 ^ src_data2_3;
data_0 = mask_data ? data_0 : dst_data_0;
data_1 = mask_data ? data_1 : dst_data_1;
data_2 = mask_data ? data_2 : dst_data_2;
data_3 = mask_data ? data_3 : dst_data_3;
*((__global char8 *)((__global char *)dst + dst_index + 0)) = data_0;
*((__global char8 *)((__global char *)dst + dst_index + 8)) = data_1;
*((__global char8 *)((__global char *)dst + dst_index + 16)) = data_2;
*((__global char8 *)((__global char *)dst + dst_index + 24)) = data_3;
}
}
#endif

633
modules/ocl/src/opencl/arithm_bitwise_xor_scalar.cl
View File

@ -1,633 +0,0 @@
////////////////////////////////////////////////////////////////////////////////////////
//
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
//
// By downloading, copying, installing or using the software you agree to this license.
// If you do not agree to this license, do not download, install,
// copy or use the software.
//
//
// License Agreement
// For Open Source Computer Vision Library
//
// Copyright (C) 2010-2012, Institute Of Software Chinese Academy Of Science, all rights reserved.
// Copyright (C) 2010-2012, Advanced Micro Devices, Inc., all rights reserved.
// Third party copyrights are property of their respective owners.
//
// @Authors
// Jiang Liyuan, jlyuan001.good@163.com
//
// Redistribution and use in source and binary forms, with or without modification,
// are permitted provided that the following conditions are met:
//
// * Redistribution's of source code must retain the above copyright notice,
// this list of conditions and the following disclaimer.
//
// * Redistribution's in binary form must reproduce the above copyright notice,
// this list of conditions and the following disclaimer in the documentation
// and/or other oclMaterials provided with the distribution.
//
// * The name of the copyright holders may not be used to endorse or promote products
// derived from this software without specific prior written permission.
//
// This software is provided by the copyright holders and contributors as is and
// any express or implied warranties, including, but not limited to, the implied
// warranties of merchantability and fitness for a particular purpose are disclaimed.
// In no event shall the Intel Corporation or contributors be liable for any direct,
// indirect, incidental, special, exemplary, or consequential damages
// (including, but not limited to, procurement of substitute goods or services;
// loss of use, data, or profits; or business interruption) however caused
// and on any theory of liability, whether in contract, strict liability,
// or tort (including negligence or otherwise) arising in any way out of
// the use of this software, even if advised of the possibility of such damage.
//
//
#if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable
#elif defined (cl_amd_fp64)
#pragma OPENCL EXTENSION cl_amd_fp64:enable
#endif
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////////
////////////////////////////////////////////BITWISE_XOR////////////////////////////////////////////////////
///////////////////////////////////////////////////////////////////////////////////////////////////////
/**************************************xor with scalar without mask**************************************/
__kernel void arithm_s_bitwise_xor_C1_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = (uchar4)(src2.x, src2.x, src2.x, src2.x);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data ^ src2_data;
data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C1_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 2;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align (dst_offset & 3)
int src1_index = mad24(y, src1_step, x + src1_offset - dst_align);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + x & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = (char4)(src2.x, src2.x, src2.x, src2.x);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data ^ src2_data;
data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C1_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
ushort2 src1_data = vload2(0, (__global ushort *)((__global char *)src1 + src1_index));
ushort2 src2_data = (ushort2)(src2.x, src2.x);
ushort2 data = *((__global ushort2 *)((__global uchar *)dst + dst_index));
ushort2 tmp_data = src1_data ^ src2_data;
data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
data.y = (dst_index + 2 < dst_end ) ? tmp_data.y : data.y;
*((__global ushort2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C1_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
short2 src1_data = vload2(0, (__global short *)((__global char *)src1 + src1_index));
short2 src2_data = (short2)(src2.x, src2.x);
short2 data = *((__global short2 *)((__global uchar *)dst + dst_index));
short2 tmp_data = src1_data ^ src2_data;
data.x = (dst_index + 0 >= dst_start) ? tmp_data.x : data.x;
data.y = (dst_index + 2 < dst_end ) ? tmp_data.y : data.y;
*((__global short2 *)((__global uchar *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C1_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
int src_data1 = *((__global int *)((__global char *)src1 + src1_index));
int src_data2 = src2.x;
int data = src_data1 ^ src_data2;
*((__global int *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C1_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
char4 src1_data = *((__global char4 *)((__global char *)src1 + src1_index));
char4 src2_data = (char4)(src2.s0, src2.s1, src2.s2, src2.s3);
char4 data = *((__global char4 *)((__global char *)dst + dst_index));
char4 tmp_data = src1_data ^ src2_data;
data.x = ((dst_index + 0 >= dst_start) && (dst_index + 0 < dst_end)) ? tmp_data.x : data.x;
data.y = ((dst_index + 1 >= dst_start) && (dst_index + 1 < dst_end)) ? tmp_data.y : data.y;
data.z = ((dst_index + 2 >= dst_start) && (dst_index + 2 < dst_end)) ? tmp_data.z : data.z;
data.w = ((dst_index + 3 >= dst_start) && (dst_index + 3 < dst_end)) ? tmp_data.w : data.w;
*((__global char4 *)((__global char *)dst + dst_index)) = data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_xor_C1_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
short4 src1_data = *((__global short4 *)((__global char *)src1 + src1_index));
short4 src2_data = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
short4 tmp_data = src1_data ^ src2_data;
*((__global short4 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#endif
__kernel void arithm_s_bitwise_xor_C2_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
uchar4 src1_data = vload4(0, src1 + src1_index);
uchar4 src2_data = (uchar4)(src2.x, src2.y, src2.x, src2.y);
uchar4 data = *((__global uchar4 *)(dst + dst_index));
uchar4 tmp_data = src1_data ^ src2_data;
data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
data.zw = (dst_index + 2 < dst_end ) ? tmp_data.zw : data.zw;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C2_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
x = x << 1;
#ifdef dst_align
#undef dst_align
#endif
#define dst_align ((dst_offset >> 1) & 1)
int src1_index = mad24(y, src1_step, (x << 1) + src1_offset - (dst_align << 1));
int dst_start = mad24(y, dst_step, dst_offset);
int dst_end = mad24(y, dst_step, dst_offset + dst_step1);
int dst_index = mad24(y, dst_step, dst_offset + (x << 1) & (int)0xfffffffc);
char4 src1_data = vload4(0, src1 + src1_index);
char4 src2_data = (char4)(src2.x, src2.y, src2.x, src2.y);
char4 data = *((__global char4 *)(dst + dst_index));
char4 tmp_data = src1_data ^ src2_data;
data.xy = (dst_index + 0 >= dst_start) ? tmp_data.xy : data.xy;
data.zw = (dst_index + 2 < dst_end ) ? tmp_data.zw : data.zw;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C2_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
ushort2 src_data1 = *((__global ushort2 *)((__global char *)src1 + src1_index));
ushort2 src_data2 = (ushort2)(src2.x, src2.y);
ushort2 data = src_data1 ^ src_data2;
*((__global ushort2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C2_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
short2 src_data1 = *((__global short2 *)((__global char *)src1 + src1_index));
short2 src_data2 = (short2)(src2.x, src2.y);
short2 data = src_data1 ^ src_data2;
*((__global short2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C2_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
int2 src_data1 = *((__global int2 *)((__global char *)src1 + src1_index));
int2 src_data2 = (int2)(src2.x, src2.y);
int2 data = src_data1 ^ src_data2;
*((__global int2 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C2_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
char8 src1_data = *((__global char8 *)((__global char *)src1 + src1_index));
char8 src2_data = (char8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
char8 tmp_data = src1_data ^ src2_data;
*((__global char8 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_xor_C2_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
short8 src1_data = *((__global short8 *)((__global char *)src1 + src1_index));
short8 src2_data = (short8)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7);
short8 tmp_data = src1_data ^ src2_data;
*((__global short8 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#endif
__kernel void arithm_s_bitwise_xor_C4_D0 (
__global uchar *src1, int src1_step, int src1_offset,
__global uchar *dst, int dst_step, int dst_offset,
uchar4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
uchar4 src_data1 = *((__global uchar4 *)(src1 + src1_index));
uchar4 data = src_data1 ^ src2;
*((__global uchar4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C4_D1 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 2) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 2) + dst_offset);
char4 src_data1 = *((__global char4 *)(src1 + src1_index));
char4 data = src_data1 ^ src2;
*((__global char4 *)(dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C4_D2 (
__global ushort *src1, int src1_step, int src1_offset,
__global ushort *dst, int dst_step, int dst_offset,
ushort4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
ushort4 src_data1 = *((__global ushort4 *)((__global char *)src1 + src1_index));
ushort4 data = src_data1 ^ src2;
*((__global ushort4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C4_D3 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 3) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 3) + dst_offset);
short4 src_data1 = *((__global short4 *)((__global char *)src1 + src1_index));
short4 data = src_data1 ^ src2;
*((__global short4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C4_D4 (
__global int *src1, int src1_step, int src1_offset,
__global int *dst, int dst_step, int dst_offset,
int4 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
int4 src_data1 = *((__global int4 *)((__global char *)src1 + src1_index));
int4 data = src_data1 ^ src2;
*((__global int4 *)((__global char *)dst + dst_index)) = data;
}
}
__kernel void arithm_s_bitwise_xor_C4_D5 (
__global char *src1, int src1_step, int src1_offset,
__global char *dst, int dst_step, int dst_offset,
char16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 4) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 4) + dst_offset);
char16 src1_data = *((__global char16 *)((__global char *)src1 + src1_index));
char16 src2_data = (char16)(src2.s0, src2.s1, src2.s2, src2.s3, src2.s4, src2.s5, src2.s6, src2.s7,
src2.s8, src2.s9, src2.sa, src2.sb, src2.sc, src2.sd, src2.se, src2.sf);
char16 tmp_data = src1_data ^ src2_data;
*((__global char16 *)((__global char *)dst + dst_index)) = tmp_data;
}
}
#if defined (DOUBLE_SUPPORT)
__kernel void arithm_s_bitwise_xor_C4_D6 (
__global short *src1, int src1_step, int src1_offset,
__global short *dst, int dst_step, int dst_offset,
short16 src2, int rows, int cols, int dst_step1)
{
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int src1_index = mad24(y, src1_step, (x << 5) + src1_offset);
int dst_index = mad24(y, dst_step, (x << 5) + dst_offset);
short4 src1_data_0 = *((__global short4 *)((__global char *)src1 + src1_index + 0));
short4 src1_data_1 = *((__global short4 *)((__global char *)src1 + src1_index + 8));
short4 src1_data_2 = *((__global short4 *)((__global char *)src1 + src1_index + 16));
short4 src1_data_3 = *((__global short4 *)((__global char *)src1 + src1_index + 24));
short4 src2_data_0 = (short4)(src2.s0, src2.s1, src2.s2, src2.s3);
short4 src2_data_1 = (short4)(src2.s4, src2.s5, src2.s6, src2.s7);
short4 src2_data_2 = (short4)(src2.s8, src2.s9, src2.sa, src2.sb);
short4 src2_data_3 = (short4)(src2.sc, src2.sd, src2.se, src2.sf);
short4 tmp_data_0 = src1_data_0 ^ src2_data_0;
short4 tmp_data_1 = src1_data_1 ^ src2_data_1;
short4 tmp_data_2 = src1_data_2 ^ src2_data_2;
short4 tmp_data_3 = src1_data_3 ^ src2_data_3;
*((__global short4 *)((__global char *)dst + dst_index + 0 ))= tmp_data_0;
*((__global short4 *)((__global char *)dst + dst_index + 8 ))= tmp_data_1;
*((__global short4 *)((__global char *)dst + dst_index + 16))= tmp_data_2;
*((__global short4 *)((__global char *)dst + dst_index + 24))= tmp_data_3;
}
}
#endif