Skip to content

O
opencv
Commit 2681cef5 authored by Ilya Lavrenov's avatar Ilya Lavrenov
Browse files
Options

rewrote ocl::remap

parent 3fe74ad1
Hide whitespace changes
Inline Side-by-side
Showing with 248 additions and 947 deletions
modules/ocl/src/imgproc.cpp
View file @ 2681cef5
...@@ -183,111 +183,88 @@ namespace cv ...@@ -183,111 +183,88 @@ namespace cv
void remap( const oclMat &src, oclMat &dst, oclMat &map1, oclMat &map2, int interpolation, int borderType, const Scalar &borderValue ) void remap( const oclMat &src, oclMat &dst, oclMat &map1, oclMat &map2, int interpolation, int borderType, const Scalar &borderValue )
{ {
Context *clCxt = src.clCxt; Context *clCxt = src.clCxt;
bool supportsDouble = clCxt->supportsFeature(FEATURE_CL_DOUBLE);
if (!supportsDouble && src.depth() == CV_64F)
{
CV_Error(CV_OpenCLDoubleNotSupported, "Selected device does not support double");
return;
}
CV_Assert(interpolation == INTER_LINEAR || interpolation == INTER_NEAREST CV_Assert(interpolation == INTER_LINEAR || interpolation == INTER_NEAREST
|| interpolation == INTER_CUBIC || interpolation == INTER_LANCZOS4); || interpolation == INTER_CUBIC || interpolation == INTER_LANCZOS4);
CV_Assert((map1.type() == CV_16SC2 && !map2.data) || (map1.type() == CV_32FC2 && !map2.data) || (map1.type() == CV_32FC1 && map2.type() == CV_32FC1)); CV_Assert((map1.type() == CV_16SC2 && !map2.data) || (map1.type() == CV_32FC2 && !map2.data) ||
(map1.type() == CV_32FC1 && map2.type() == CV_32FC1));
CV_Assert(!map2.data || map2.size() == map1.size()); CV_Assert(!map2.data || map2.size() == map1.size());
CV_Assert(dst.size() == map1.size()); CV_Assert(borderType == BORDER_CONSTANT);
dst.create(map1.size(), src.type()); dst.create(map1.size(), src.type());
string kernelName; const char * const typeMap[] = { "uchar", "char", "ushort", "short", "int", "float", "double" };
const char * const channelMap[] = { "", "", "2", "4", "4" };
const char * const interMap[] = { "INTER_NEAREST", "INTER_LINEAR", "INTER_CUBIC", "INTER_LINEAR", "INTER_LANCZOS" };
const char * const borderMap[] = { "BORDER_CONSTANT", "BORDER_REPLICATE", "BORDER_REFLECT", "BORDER_WRAP",
"BORDER_REFLECT_101", "BORDER_TRANSPARENT" };
string kernelName = "remap";
if ( map1.type() == CV_32FC2 && !map2.data ) if ( map1.type() == CV_32FC2 && !map2.data )
{ kernelName += "_32FC2";
if (interpolation == INTER_LINEAR && borderType == BORDER_CONSTANT)
kernelName = "remapLNFConstant";
else if (interpolation == INTER_NEAREST && borderType == BORDER_CONSTANT)
kernelName = "remapNNFConstant";
}
else if (map1.type() == CV_16SC2 && !map2.data) else if (map1.type() == CV_16SC2 && !map2.data)
{ kernelName += "_16SC2";
if (interpolation == INTER_LINEAR && borderType == BORDER_CONSTANT)
kernelName = "remapLNSConstant";
else if (interpolation == INTER_NEAREST && borderType == BORDER_CONSTANT)
kernelName = "remapNNSConstant";
}
else if (map1.type() == CV_32FC1 && map2.type() == CV_32FC1) else if (map1.type() == CV_32FC1 && map2.type() == CV_32FC1)
{ kernelName += "_2_32FC1";
if (interpolation == INTER_LINEAR && borderType == BORDER_CONSTANT) else
kernelName = "remapLNF1Constant"; CV_Error(CV_StsBadArg, "Unsupported map types");
else if (interpolation == INTER_NEAREST && borderType == BORDER_CONSTANT)
kernelName = "remapNNF1Constant";
}
size_t blkSizeX = 16, blkSizeY = 16; int ocn = dst.oclchannels();
size_t glbSizeX; size_t localThreads[3] = { 16, 16, 1};
int cols = dst.cols; size_t globalThreads[3] = { dst.cols, dst.rows, 1};
if (src.type() == CV_8UC1)
{
cols = (dst.cols + dst.offset % 4 + 3) / 4;
glbSizeX = cols % blkSizeX == 0 ? cols : (cols / blkSizeX + 1) * blkSizeX;
} Mat scalar(1, 1, CV_MAKE_TYPE(dst.depth(), ocn), borderValue);
else if (src.type() == CV_32FC1 && interpolation == INTER_LINEAR) std::string buildOptions = format("-D %s -D %s -D T=%s%s", interMap[interpolation],
borderMap[borderType], typeMap[src.depth()], channelMap[ocn]);
if (interpolation != INTER_NEAREST)
{ {
cols = (dst.cols + (dst.offset >> 2) % 4 + 3) / 4; int wdepth = std::max(CV_32F, dst.depth());
glbSizeX = cols % blkSizeX == 0 ? cols : (cols / blkSizeX + 1) * blkSizeX; if (!supportsDouble)
wdepth = std::min(CV_32F, wdepth);
buildOptions += format(" -D WT=%s%s -D convertToT=convert_%s%s%s -D convertToWT=convert_%s%s"
" -D convertToWT2=convert_%s2 -D WT2=%s2",
typeMap[wdepth], channelMap[ocn],
typeMap[src.depth()], channelMap[ocn], src.depth() < CV_32F ? "_sat_rte" : "",
typeMap[wdepth], channelMap[ocn],
typeMap[wdepth], typeMap[wdepth]);
} }
else
glbSizeX = dst.cols % blkSizeX == 0 ? dst.cols : (dst.cols / blkSizeX + 1) * blkSizeX;
size_t glbSizeY = dst.rows % blkSizeY == 0 ? dst.rows : (dst.rows / blkSizeY + 1) * blkSizeY; int src_step = src.step / src.elemSize(), src_offset = src.offset / src.elemSize();
size_t globalThreads[3] = {glbSizeX, glbSizeY, 1}; int map1_step = map1.step / map1.elemSize(), map1_offset = map1.offset / map1.elemSize();
size_t localThreads[3] = {blkSizeX, blkSizeY, 1}; int map2_step = map2.step / map2.elemSize(), map2_offset = map2.offset / map2.elemSize();
int dst_step = dst.step / dst.elemSize(), dst_offset = dst.offset / dst.elemSize();
float borderFloat[4] = {(float)borderValue[0], (float)borderValue[1], (float)borderValue[2], (float)borderValue[3]};
vector< pair<size_t, const void *> > args; vector< pair<size_t, const void *> > args;
if (map1.channels() == 2) args.push_back( make_pair(sizeof(cl_mem), (void *)&src.data));
{ args.push_back( make_pair(sizeof(cl_mem), (void *)&dst.data));
args.push_back( make_pair(sizeof(cl_mem), (void *)&dst.data)); args.push_back( make_pair(sizeof(cl_mem), (void *)&map1.data));
args.push_back( make_pair(sizeof(cl_mem), (void *)&src.data)); if (!map2.empty())
args.push_back( make_pair(sizeof(cl_mem), (void *)&map1.data));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&map1.offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.step));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.step));
args.push_back( make_pair(sizeof(cl_int), (void *)&map1.step));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.cols));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.rows));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.cols));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.rows));
args.push_back( make_pair(sizeof(cl_int), (void *)&map1.cols));
args.push_back( make_pair(sizeof(cl_int), (void *)&map1.rows));
args.push_back( make_pair(sizeof(cl_int), (void *)&cols));
if (src.clCxt->supportsFeature(FEATURE_CL_DOUBLE))
args.push_back( make_pair(sizeof(cl_double4), (void *)&borderValue));
else
args.push_back( make_pair(sizeof(cl_float4), (void *)&borderFloat));
}
if (map1.channels() == 1)
{
args.push_back( make_pair(sizeof(cl_mem), (void *)&dst.data));
args.push_back( make_pair(sizeof(cl_mem), (void *)&src.data));
args.push_back( make_pair(sizeof(cl_mem), (void *)&map1.data));
args.push_back( make_pair(sizeof(cl_mem), (void *)&map2.data)); args.push_back( make_pair(sizeof(cl_mem), (void *)&map2.data));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.offset)); args.push_back( make_pair(sizeof(cl_int), (void *)&src_offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.offset)); args.push_back( make_pair(sizeof(cl_int), (void *)&dst_offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&map1.offset)); args.push_back( make_pair(sizeof(cl_int), (void *)&map1_offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.step)); if (!map2.empty())
args.push_back( make_pair(sizeof(cl_int), (void *)&src.step)); args.push_back( make_pair(sizeof(cl_int), (void *)&map2_offset));
args.push_back( make_pair(sizeof(cl_int), (void *)&map1.step)); args.push_back( make_pair(sizeof(cl_int), (void *)&src_step));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.cols)); args.push_back( make_pair(sizeof(cl_int), (void *)&dst_step));
args.push_back( make_pair(sizeof(cl_int), (void *)&src.rows)); args.push_back( make_pair(sizeof(cl_int), (void *)&map1_step));
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.cols)); if (!map2.empty())
args.push_back( make_pair(sizeof(cl_int), (void *)&dst.rows)); args.push_back( make_pair(sizeof(cl_int), (void *)&map2_step));
args.push_back( make_pair(sizeof(cl_int), (void *)&map1.cols)); args.push_back( make_pair(sizeof(cl_int), (void *)&src.cols));
args.push_back( make_pair(sizeof(cl_int), (void *)&map1.rows)); args.push_back( make_pair(sizeof(cl_int), (void *)&src.rows));
args.push_back( make_pair(sizeof(cl_int), (void *)&cols)); args.push_back( make_pair(sizeof(cl_int), (void *)&dst.cols));
if (src.clCxt->supportsFeature(FEATURE_CL_DOUBLE)) args.push_back( make_pair(sizeof(cl_int), (void *)&dst.rows));
args.push_back( make_pair(sizeof(cl_double4), (void *)&borderValue)); args.push_back( make_pair(scalar.elemSize(), (void *)scalar.data));
else
args.push_back( make_pair(sizeof(cl_float4), (void *)&borderFloat)); openCLExecuteKernel(clCxt, &imgproc_remap, kernelName, globalThreads, localThreads, args, -1, -1, buildOptions.c_str());
}
openCLExecuteKernel(clCxt, &imgproc_remap, kernelName, globalThreads, localThreads, args, src.oclchannels(), src.depth());
} }
//////////////////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////////////////
......
modules/ocl/src/opencl/imgproc_remap.cl
View file @ 2681cef5
/*M/////////////////////////////////////////////////////////////////////////////////////// /*M///////////////////////////////////////////////////////////////////////////////////////
// //
// IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING. // IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
...@@ -43,940 +42,265 @@ ...@@ -43,940 +42,265 @@
// the use of this software, even if advised of the possibility of such damage. // the use of this software, even if advised of the possibility of such damage.
// //
//M*/ //M*/
//#pragma OPENCL EXTENSION cl_amd_printf : enable
#if defined DOUBLE_SUPPORT #if defined (DOUBLE_SUPPORT)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64:enable #pragma OPENCL EXTENSION cl_khr_fp64:enable
typedef double4 F4 ; #elif defined (cl_amd_fp64)
#else #pragma OPENCL EXTENSION cl_amd_fp64:enable
typedef float4 F4; #endif
#endif #endif
#define NEED_EXTRAPOLATION(gx, gy) (gx >= src_cols || gy >= src_rows || gx < 0 || gy < 0)
///////////////////////////////////////////////////////// #ifdef INTER_NEAREST
///////////////////////using buffer//////////////////////
/////////////////////////////////////////////////////////
__kernel void remapNNSConstant_C1_D0(__global unsigned char* dst, __global unsigned char const * restrict src,
__global short * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows, int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
x = x << 2;
int gx = x - (dst_offset&3);
int4 Gx = (int4)(gx, gx+1, gx+2, gx+3);
uchar4 nval =convert_uchar4(nVal);
uchar4 val = (uchar4)(nval.s0);
int dstStart = (y * dst_step + x + dst_offset) - (dst_offset&3);
int map1Start = y * map1_step + (x << 2) + map1_offset - ((dst_offset & 3) << 2);
short8 map1_data;
map1_data = *((__global short8 *)((__global char*)map1 + map1Start));
int4 srcIdx = convert_int4(map1_data.odd) * src_step + convert_int4(map1_data.even) + src_offset;
uchar4 con = convert_uchar4(convert_int4(map1_data.even) >= (int4)(src_cols) || convert_int4(map1_data.odd) >= (int4)(src_rows) || convert_int4(map1_data.even) < (int4)(0) || convert_int4(map1_data.odd) < (int4)(0));
uchar4 src_data = val;
if (con.s0 == 0)
src_data.s0 = *(src + srcIdx.s0);
if (con.s1 == 0)
src_data.s1 = *(src + srcIdx.s1);
if (con.s2 == 0)
src_data.s2 = *(src + srcIdx.s2);
if (con.s3 == 0)
src_data.s3 = *(src + srcIdx.s3);
uchar4 dst_data;
__global uchar4* d = (__global uchar4 *)(dst + dstStart);
uchar4 dVal = *d;
int4 dcon = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
dst_data = (convert_uchar4(dcon) != convert_uchar4((int4)(0))) ? src_data : dVal;
*d = dst_data;
}
} __kernel void remap_2_32FC1(__global const T * restrict src, __global T * dst,
__global float * map1, __global float * map2,
__kernel void remapNNFConstant_C1_D0(__global unsigned char* dst, __global unsigned char const * restrict src, int src_offset, int dst_offset, int map1_offset, int map2_offset,
__global float * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step, int src_step, int dst_step, int map1_step, int map2_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows, int threadCols, F4 nVal) int src_cols, int src_rows, int dst_cols, int dst_rows, T scalar)
{ {
int x = get_global_id(0); int x = get_global_id(0);
int y = get_global_id(1); int y = get_global_id(1);
if(x < threadCols && y < dst_rows) if (x < dst_cols && y < dst_rows)
{ {
x = x << 2; int dstIdx = mad24(y, dst_step, x + dst_offset);
int gx = x - (dst_offset&3); int map1Idx = mad24(y, map1_step, x + map1_offset);
int4 Gx = (int4)(gx, gx+1, gx+2, gx+3); int map2Idx = mad24(y, map2_step, x + map2_offset);
uchar4 nval =convert_uchar4(nVal);
uchar val = nval.s0;
int dstStart = (y * dst_step + x + dst_offset) - (dst_offset&3);
int map1Start = y * map1_step + (x << 3) + map1_offset - ((dst_offset & 3) << 3);
float8 map1_data;
map1_data = *((__global float8 *)((__global char*)map1 + map1Start));
int8 map1_dataZ = convert_int8_sat_rte(map1_data);
int4 srcIdx = map1_dataZ.odd * src_step + map1_dataZ.even + src_offset;
uchar4 src_data = val; int gx = convert_int_sat_rte(map1[map1Idx]);
uchar4 con = convert_uchar4(map1_dataZ.even >= (int4)(src_cols) || map1_dataZ.odd >= (int4)(src_rows) || map1_dataZ.even < (int4)(0) || map1_dataZ.odd < (int4)(0)); int gy = convert_int_sat_rte(map2[map2Idx]);
if (con.s0 == 0) if (NEED_EXTRAPOLATION(gx, gy))
src_data.s0 = *(src + srcIdx.s0); {
if (con.s1 == 0) #ifdef BORDER_CONSTANT
src_data.s1 = *(src + srcIdx.s1); dst[dstIdx] = scalar;
if (con.s2 == 0) #else
src_data.s2 = *(src + srcIdx.s2); #error No extrapolation method
if (con.s3 == 0) #endif
src_data.s3 = *(src + srcIdx.s3); }
uchar4 dst_data; else
// dst_data = convert_uchar4(map1_dataZ.even >= (int4)(src_cols) || map1_dataZ.odd >= (int4)(src_rows)) ? (uchar4)(val) : src_data; {
__global uchar4* d = (__global uchar4 *)(dst + dstStart); int srcIdx = mad24(gy, src_step, gx + src_offset);
dst[dstIdx] = src[srcIdx];
uchar4 dVal = *d; }
int4 dcon = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
dst_data = (convert_uchar4(dcon) != convert_uchar4((int4)(0))) ? src_data : dVal;
*d = dst_data;
} }
} }
__kernel void remapNNF1Constant_C1_D0(__global unsigned char* dst, __global unsigned char const * restrict src, __kernel void remap_32FC2(__global const T * restrict src, __global T * dst, __global float2 * map1,
__global float * map1, __global float * map2, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step, int src_offset, int dst_offset, int map1_offset,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows, int threadCols, F4 nVal) int src_step, int dst_step, int map1_step,
int src_cols, int src_rows, int dst_cols, int dst_rows, T scalar)
{ {
int x = get_global_id(0); int x = get_global_id(0);
int y = get_global_id(1); int y = get_global_id(1);
if(x < threadCols && y < dst_rows) if (x < dst_cols && y < dst_rows)
{ {
x = x << 2; int dstIdx = mad24(y, dst_step, x + dst_offset);
int gx = x - (dst_offset&3); int map1Idx = mad24(y, map1_step, x + map1_offset);
int4 Gx = (int4)(gx, gx+1, gx+2, gx+3);
uchar4 nval =convert_uchar4(nVal);
uchar4 val = (uchar4)(nval.s0);
int dstStart = (y * dst_step + x + dst_offset) - (dst_offset&3);
int map1Start = y * map1_step + (x << 2) + map1_offset - ((dst_offset & 3) << 2);
float4 map1_data;
float4 map2_data;
map1_data = *((__global float4 *)((__global char*)map1 + map1Start));
map2_data = *((__global float4 *)((__global char*)map2 + map1Start));
float8 map_data = (float8)(map1_data.s0, map2_data.s0, map1_data.s1, map2_data.s1, map1_data.s2, map2_data.s2, map1_data.s3, map2_data.s3);
int8 map_dataZ = convert_int8_sat_rte(map_data);
int4 srcIdx = map_dataZ.odd * src_step + map_dataZ.even + src_offset;
uchar4 src_data = val;
uchar4 con = convert_uchar4(map_dataZ.even >= (int4)(src_cols) || map_dataZ.odd >= (int4)(src_rows)|| map_dataZ.even < (int4)(0) || map_dataZ.odd < (int4)(0));
if (con.s0 == 0)
src_data.s0 = *(src + srcIdx.s0);
if (con.s1 == 0)
src_data.s1 = *(src + srcIdx.s1);
if (con.s2 == 0)
src_data.s2 = *(src + srcIdx.s2);
if (con.s3 == 0)
src_data.s3 = *(src + srcIdx.s3);
uchar4 dst_data;
// dst_data = convert_uchar4(map_dataZ.even >= (int4)(src_cols) || map_dataZ.odd >= (int4)(src_rows)) ? (uchar4)(val) : src_data;
__global uchar4* d = (__global uchar4 *)(dst + dstStart);
uchar4 dVal = *d; int2 gxy = convert_int2_sat_rte(map1[map1Idx]);
int gx = gxy.x, gy = gxy.y;
int4 dcon = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows); if (NEED_EXTRAPOLATION(gx, gy))
{
dst_data = (convert_uchar4(dcon) != convert_uchar4((int4)(0))) ? src_data : dVal; #ifdef BORDER_CONSTANT
*d = dst_data; dst[dstIdx] = scalar;
#else
#error No extrapolation method
#endif
}
else
{
int srcIdx = mad24(gy, src_step, gx + src_offset);
dst[dstIdx] = src[srcIdx];
}
} }
} }
__kernel void remap_16SC2(__global const T * restrict src, __global T * dst, __global short2 * map1,
__kernel void remapNNSConstant_C4_D0(__global unsigned char* dst, __global unsigned char const * restrict src, int src_offset, int dst_offset, int map1_offset,
__global short * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step, int src_step, int dst_step, int map1_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows, int threadCols, F4 nVal) int src_cols, int src_rows, int dst_cols, int dst_rows, T scalar)
{ {
int x = get_global_id(0); int x = get_global_id(0);
int y = get_global_id(1); int y = get_global_id(1);
if(x < threadCols && y < dst_rows) if (x < dst_cols && y < dst_rows)
{ {
int dstIdx = y * dst_step + (x << 2) + dst_offset; int dstIdx = mad24(y, dst_step, x + dst_offset);
int mapIdx = y * map1_step + (x << 2) + map1_offset; int map1Idx = mad24(y, map1_step, x + map1_offset);
short2 map1_data = *((__global short2 *)((__global char*)map1 + mapIdx));
int srcIdx = map1_data.y * src_step + (map1_data.x << 2) + src_offset;
uchar4 nval = convert_uchar4(nVal);
uchar4 src_data;
if(map1_data.x >= src_cols || map1_data.y >= src_rows || map1_data.x <0 || map1_data.y < 0 )
src_data = nval;
else
src_data = *((__global uchar4 *)((__global uchar *)src + srcIdx));
*((__global uchar4 *)((__global uchar*)dst + dstIdx)) = src_data;
int2 gxy = convert_int2(map1[map1Idx]);
int gx = gxy.x, gy = gxy.y;
if (NEED_EXTRAPOLATION(gx, gy))
{
#ifdef BORDER_CONSTANT
dst[dstIdx] = scalar;
#else
#error No extrapolation method
#endif
}
else
{
int srcIdx = mad24(gy, src_step, gx + src_offset);
dst[dstIdx] = src[srcIdx];
}
} }
} }
__kernel void remapNNFConstant_C4_D0(__global unsigned char* dst, __global unsigned char const * restrict src, #elif INTER_LINEAR
__global float * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows, int threadCols, F4 nVal) __kernel void remap_2_32FC1(__global T const * restrict src, __global T * dst,
__global float * map1, __global float * map2,
int src_offset, int dst_offset, int map1_offset, int map2_offset,
int src_step, int dst_step, int map1_step, int map2_step,
int src_cols, int src_rows, int dst_cols, int dst_rows, T nVal)
{ {
int x = get_global_id(0); int x = get_global_id(0);
int y = get_global_id(1); int y = get_global_id(1);
if(x < threadCols && y < dst_rows) if (x < dst_cols && y < dst_rows)
{ {
int dstIdx = y * dst_step + (x << 2) + dst_offset; int dstIdx = mad24(y, dst_step, x + dst_offset);
int mapIdx = y * map1_step + (x << 3) + map1_offset; int map1Idx = mad24(y, map1_step, x + map1_offset);
float2 map1_data = *((__global float2 *)((__global char*)map1 + mapIdx)); int map2Idx = mad24(y, map2_step, x + map2_offset);
int2 map1_dataZ = convert_int2_sat_rte(map1_data);
int srcIdx = map1_dataZ.y * src_step + (map1_dataZ.x << 2) + src_offset;
uchar4 nval = convert_uchar4(nVal);
uchar4 src_data;
if(map1_dataZ.x >= src_cols || map1_dataZ.y >= src_rows || map1_dataZ.x < 0 || map1_dataZ.y < 0)
src_data = nval;
else
src_data = *((__global uchar4 *)((__global uchar *)src + srcIdx));
*((__global uchar4 *)((__global uchar*)dst + dstIdx)) = src_data;
} float2 map_data = (float2)(map1[map1Idx], map2[map2Idx]);
} int2 map_dataA = convert_int2_sat_rtn(map_data);
int2 map_dataB = (int2)(map_dataA.x + 1, map_dataA.y);
int2 map_dataC = (int2)(map_dataA.x, map_dataA.y + 1);
int2 map_dataD = (int2)(map_dataA.x + 1, map_dataA.y +1);
__kernel void remapNNF1Constant_C4_D0(__global unsigned char* dst, __global unsigned char const * restrict src, float2 _u = map_data - convert_float2(map_dataA);
__global float * map1, __global float * map2, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step, WT2 u = convertToWT2(convert_int2_rte(convertToWT2(_u) * (WT2)32)) / (WT2)32;
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows, int threadCols, F4 nVal) WT nval = convertToWT(nVal);
{ WT a = nval, b = nval, c = nval, d = nval;
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows) if (!NEED_EXTRAPOLATION(map_dataA.x, map_dataA.y))
{ a = convertToWT(src[mad24(map_dataA.y, src_step, map_dataA.x + src_offset)]);
int dstIdx = y * dst_step + (x << 2) + dst_offset;
int mapIdx = y * map1_step + (x << 2) + map1_offset;
float map1_data = *((__global float *)((__global char*)map1 + mapIdx));
float map2_data = *((__global float *)((__global char*)map2 + mapIdx));
int srcIdx = convert_int_sat_rte(map2_data) * src_step + (convert_int_sat_rte(map1_data) << 2) + src_offset;
uchar4 nval = convert_uchar4(nVal);
uchar4 src_data;
if(convert_int_sat_rte(map1_data) >= src_cols || convert_int_sat_rte(map2_data) >= src_rows || convert_int_sat_rte(map1_data) < 0 || convert_int_sat_rte(map2_data) < 0)
src_data = nval;
else else
src_data = *((__global uchar4 *)((__global uchar *)src + srcIdx)); {
*((__global uchar4 *)((__global uchar*)dst + dstIdx)) = src_data; #ifdef BORDER_CONSTANT
} #else
} #error No extrapolation method
#endif
__kernel void remapNNSConstant_C1_D5(__global float* dst, __global float const * restrict src, }
__global short * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows ,int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows) if (!NEED_EXTRAPOLATION(map_dataB.x, map_dataB.y))
{ b = convertToWT(src[mad24(map_dataB.y, src_step, map_dataB.x + src_offset)]);
int dstIdx = y * dst_step + (x << 2) + dst_offset;
int mapIdx = y * map1_step + (x << 2) + map1_offset;
short2 map1_data = *((__global short2 *)((__global char*)map1 + mapIdx));
int srcIdx = map1_data.y * src_step + (map1_data.x << 2) + src_offset;
float nval = convert_float(nVal.x);
float src_data;
if(map1_data.x >= src_cols || map1_data.y >= src_rows|| map1_data.x < 0 || map1_data.y < 0)
src_data = nval;
else else
src_data = *((__global float *)((__global uchar *)src + srcIdx)); {
*((__global float *)((__global uchar*)dst + dstIdx)) = src_data; #ifdef BORDER_CONSTANT
#else
#error No extrapolation method
} #endif
}
}
__kernel void remapNNFConstant_C1_D5(__global float* dst, __global float const * restrict src,
__global float * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows ,int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows) if (!NEED_EXTRAPOLATION(map_dataC.x, map_dataC.y))
{ c = convertToWT(src[mad24(map_dataC.y, src_step, map_dataC.x + src_offset)]);
int dstIdx = y * dst_step + (x << 2) + dst_offset;
int mapIdx = y * map1_step + (x << 3) + map1_offset;
float2 map1_data = *((__global float2 *)((__global char*)map1 + mapIdx));
int2 map1_dataZ = convert_int2_sat_rte(map1_data);
int srcIdx = map1_dataZ.y * src_step + (map1_dataZ.x << 2) + src_offset;
float nval = convert_float(nVal.x);
float src_data;
if(map1_dataZ.x >= src_cols || map1_dataZ.y >= src_rows || map1_dataZ.x < 0 || map1_dataZ.y < 0)
src_data = nval;
else else
src_data = *((__global float *)((__global uchar *)src + srcIdx)); {
*((__global float *)((__global uchar*)dst + dstIdx)) = src_data; #ifdef BORDER_CONSTANT
#else
#error No extrapolation method
} #endif
}
}
__kernel void remapNNF1Constant_C1_D5(__global float* dst, __global float const * restrict src,
__global float * map1, __global float * map2, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows ,int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows) if (!NEED_EXTRAPOLATION(map_dataD.x, map_dataD.y))
{ d = convertToWT(src[mad24(map_dataD.y, src_step, map_dataD.x + src_offset)]);
int dstIdx = y * dst_step + (x << 2) + dst_offset;
int mapIdx = y * map1_step + (x << 2) + map1_offset;
float map1_data = *((__global float *)((__global char*)map1 + mapIdx));
float map2_data = *((__global float *)((__global char*)map2 + mapIdx));
float2 map_data = (float2)(map1_data, map2_data);
int2 map1_dataZ = convert_int2_sat_rte(map_data);
int srcIdx = map1_dataZ.y * src_step + (map1_dataZ.x << 2) + src_offset;
float nval = convert_float(nVal.x);
float src_data;
if(map1_dataZ.x >= src_cols || map1_dataZ.y >= src_rows || map1_dataZ.x < 0 || map1_dataZ.y < 0)
src_data = nval;
else else
src_data = *((__global float *)((__global uchar *)src + srcIdx)); {
*((__global float *)((__global uchar*)dst + dstIdx)) = src_data; #ifdef BORDER_CONSTANT
#else
#error No extrapolation method
} #endif
}
}
__kernel void remapNNSConstant_C4_D5(__global float * dst, __global float const * restrict src,
__global short * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
int dstIdx = y * dst_step + (x << 4) + dst_offset ;
int mapIdx = y * map1_step + (x << 2) + map1_offset ;
short2 map1_data = *((__global short2 *)((__global char*)map1 + mapIdx));
int srcIdx = map1_data.y * src_step + (map1_data.x << 4) + src_offset;
float4 nval = convert_float4(nVal);
float4 src_data;
if (map1_data.x <0 || map1_data.x >= src_cols || map1_data.y <0 || map1_data.y >= src_rows)
src_data = nval;
else
src_data = *((__global float4 *)((__global uchar *)src + srcIdx));
*((__global float4 *)((__global uchar*)dst + dstIdx)) = src_data;
}
}
__kernel void remapNNFConstant_C4_D5(__global float * dst, __global float const * restrict src,
__global float * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
int dstIdx = y * dst_step + (x << 4) + dst_offset ;
int mapIdx = y * map1_step + (x << 3) + map1_offset ;
float2 map1_data = *((__global float2 *)((__global char*)map1 + mapIdx));
int2 map1_dataZ = convert_int2_sat_rte(map1_data);
int srcIdx = map1_dataZ.y * src_step + (map1_dataZ.x << 4) + src_offset;
float4 nval = convert_float4(nVal);
float4 src_data = nval;
if(map1_dataZ.x >= 0 && map1_dataZ.x < src_cols && map1_dataZ.y >=0 && map1_dataZ.y < src_rows)
src_data = *((__global float4 *)((__global uchar *)src + srcIdx));
*((__global float4 *)((__global uchar*)dst + dstIdx)) = src_data;
}
}
__kernel void remapNNF1Constant_C4_D5(__global float * dst, __global float const * restrict src,
__global float * map1, __global float * map2, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows) WT dst_data = a * (WT)(1.0 - u.x) * (WT)(1.0 - u.y) +
{ b * (WT)(u.x) * (WT)(1.0 - u.y) +
int dstIdx = y * dst_step + (x << 4) + dst_offset ; c * (WT)(1.0 - u.x) * (WT)(u.y) +
int mapIdx = y * map1_step + (x << 2) + map1_offset ; d * (WT)(u.x) * (WT)(u.y);
float map1_data = *((__global float *)((__global char*)map1 + mapIdx)); dst[dstIdx] = convertToT(dst_data);
float map2_data = *((__global float *)((__global char*)map2 + mapIdx));
float2 map_data = (float2)(map1_data, map2_data);
int2 map1_dataZ = convert_int2_sat_rte(map_data);
int srcIdx = map1_dataZ.y * src_step + (map1_dataZ.x << 4) + src_offset;
float4 nval = convert_float4(nVal);
float4 src_data = nval;
if(map1_dataZ.x >= 0 && map1_dataZ.x < src_cols && map1_dataZ.y >= 0 && map1_dataZ.y < src_rows)
src_data = *((__global float4 *)((__global uchar *)src + srcIdx));
*((__global float4 *)((__global uchar*)dst + dstIdx)) = src_data;
} }
} }
__kernel void remap_32FC2(__global T const * restrict src, __global T * dst,
__global float2 * map1,
__kernel void remapLNFConstant_C1_D0(__global unsigned char* dst, __global unsigned char const * restrict src, int src_offset, int dst_offset, int map1_offset,
__global float * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step, int src_step, int dst_step, int map1_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal) int src_cols, int src_rows, int dst_cols, int dst_rows, T nVal)
{ {
int x = get_global_id(0); int x = get_global_id(0);
int y = get_global_id(1); int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
x = x << 2;
int gx = x - (dst_offset&3);
int4 Gx = (int4)(gx, gx+1, gx+2, gx+3);
uchar4 nval =convert_uchar4(nVal);
uchar4 val = (uchar4)(nval.s0);
int dstStart = (y * dst_step + x + dst_offset) - (dst_offset&3);
int map1Start = y * map1_step + (x << 3) + map1_offset - ((dst_offset & 3) << 3);
float8 map1_data;
map1_data = *((__global float8 *)((__global char*)map1 + map1Start));
int8 map1_dataD = convert_int8(map1_data);
float8 temp = map1_data - convert_float8(map1_dataD);
float4 u = temp.even;
float4 v = temp.odd;
float4 ud = (float4)(1.0) - u;
float4 vd = (float4)(1.0) - v;
//float8 map1_dataU = map1_dataD + 1;
int4 map1_dataDx = map1_dataD.even;
int4 map1_dataDy = map1_dataD.odd;
int4 map1_dataDx1 = map1_dataDx + (int4)(1);
int4 map1_dataDy1 = map1_dataDy + (int4)(1);
uchar4 a = val, b = val, c = val, d =val;
if (map1_dataDx.s0 < src_cols && map1_dataDx.s0 >= 0 && map1_dataDy.s0 < src_rows && map1_dataDy.s0 >= 0)
a.s0 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s0 * src_step + map1_dataDx.s0 + src_offset));
if (map1_dataDx.s1 < src_cols && map1_dataDx.s1 >= 0 && map1_dataDy.s1 < src_rows && map1_dataDy.s1 >= 0)
a.s1 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s1 * src_step + map1_dataDx.s1 + src_offset));
if (map1_dataDx.s2 < src_cols && map1_dataDx.s2 >= 0 && map1_dataDy.s2 < src_rows && map1_dataDy.s2 >= 0)
a.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s2 * src_step + map1_dataDx.s2 + src_offset));
if (map1_dataDx.s3 < src_cols && map1_dataDx.s3 >= 0 && map1_dataDy.s3 < src_rows && map1_dataDy.s3 >= 0)
a.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s3 * src_step + map1_dataDx.s3 + src_offset));
if (map1_dataDx1.s0 < src_cols && map1_dataDx1.s0 >= 0 && map1_dataDy.s0 < src_rows && map1_dataDy.s0 >= 0)
b.s0 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s0 * src_step + map1_dataDx1.s0 + src_offset));
if (map1_dataDx1.s1 < src_cols && map1_dataDx1.s1 >= 0 && map1_dataDy.s1 < src_rows && map1_dataDy.s1 >= 0)
b.s1 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s1 * src_step + map1_dataDx1.s1 + src_offset));
if (map1_dataDx1.s2 < src_cols && map1_dataDx1.s2 >= 0 && map1_dataDy.s2 < src_rows && map1_dataDy.s2 >= 0)
b.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s2 * src_step + map1_dataDx1.s2 + src_offset));
if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy.s3 < src_rows && map1_dataDy.s3 >= 0)
b.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s3 * src_step + map1_dataDx1.s3 + src_offset));
if (map1_dataDx.s0 < src_cols && map1_dataDx.s0 >= 0 && map1_dataDy1.s0 < src_rows && map1_dataDy1.s0 >= 0)
c.s0 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s0 * src_step + map1_dataDx.s0 + src_offset));
if (map1_dataDx.s1 < src_cols && map1_dataDx.s1 >= 0 && map1_dataDy1.s1 < src_rows && map1_dataDy1.s1 >= 0)
c.s1 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s1 * src_step + map1_dataDx.s1 + src_offset));
if (map1_dataDx.s2 < src_cols && map1_dataDx.s2 >= 0 && map1_dataDy1.s2 < src_rows && map1_dataDy1.s2 >= 0)
c.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s2 * src_step + map1_dataDx.s2 + src_offset));
if (map1_dataDx.s3 < src_cols && map1_dataDx.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
c.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s3 * src_step + map1_dataDx.s3 + src_offset));
if (map1_dataDx1.s0 < src_cols && map1_dataDx1.s0 >= 0 && map1_dataDy1.s0 < src_rows && map1_dataDy1.s0 >= 0)
d.s0 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s0 * src_step + map1_dataDx1.s0 + src_offset));
if (map1_dataDx1.s1 < src_cols && map1_dataDx1.s1 >= 0 && map1_dataDy1.s1 < src_rows && map1_dataDy1.s1 >= 0)
d.s1 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s1 * src_step + map1_dataDx1.s1 + src_offset));
if (map1_dataDx1.s2 < src_cols && map1_dataDx1.s2 >= 0 && map1_dataDy1.s2 < src_rows && map1_dataDy1.s2 >= 0)
d.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s2 * src_step + map1_dataDx1.s2 + src_offset));
if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
d.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s3 * src_step + map1_dataDx1.s3 + src_offset));
uchar4 dst_data = convert_uchar4_sat_rte((convert_float4(a))* ud * vd +(convert_float4(b))* u * vd + (convert_float4(c))* ud * v + (convert_float4(d)) * u * v );
__global uchar4* D = (__global uchar4 *)(dst + dstStart);
uchar4 dVal = *D;
int4 con = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
dst_data = (convert_uchar4(con) != (uchar4)(0)) ? dst_data : dVal;
*D = dst_data;
}
}
__kernel void remapLNF1Constant_C1_D0(__global unsigned char* dst, __global unsigned char const * restrict src, if (x < dst_cols && y < dst_rows)
__global float * map1, __global float * map2, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{ {
x = x << 2; int dstIdx = mad24(y, dst_step, x + dst_offset);
int gx = x - (dst_offset&3); int map1Idx = mad24(y, map1_step, x + map1_offset);
int4 Gx = (int4)(gx, gx+1, gx+2, gx+3);
uchar4 nval =convert_uchar4(nVal);
uchar4 val = (uchar4)(nval.s0);
int dstStart = (y * dst_step + x + dst_offset) - (dst_offset&3);
int map1Start = y * map1_step + (x << 2) + map1_offset - ((dst_offset & 3) << 2);
float4 map1_data;
float4 map2_data;
map1_data = *((__global float4 *)((__global char*)map1 + map1Start));
map2_data = *((__global float4 *)((__global char*)map2 + map1Start));
float8 map_data = (float8)(map1_data.s0, map2_data.s0, map1_data.s1, map2_data.s1, map1_data.s2, map2_data.s2, map1_data.s3, map2_data.s3);
int8 map1_dataD = convert_int8(map_data);
float8 temp = map_data - convert_float8(map1_dataD);
float4 u = temp.even;
float4 v = temp.odd;
float4 ud = (float4)(1.0) - u;
float4 vd = (float4)(1.0) - v;
//float8 map1_dataU = map1_dataD + 1;
int4 map1_dataDx = map1_dataD.even;
int4 map1_dataDy = map1_dataD.odd;
int4 map1_dataDx1 = map1_dataDx + (int4)(1);
int4 map1_dataDy1 = map1_dataDy + (int4)(1);
uchar4 a = val, b = val, c = val, d =val;
if (map1_dataDx.s0 < src_cols && map1_dataDx.s0 >= 0 && map1_dataDy.s0 < src_rows && map1_dataDy.s0 >= 0)
a.s0 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s0 * src_step + map1_dataDx.s0 + src_offset));
if (map1_dataDx.s1 < src_cols && map1_dataDx.s1 >= 0 && map1_dataDy.s1 < src_rows && map1_dataDy.s1 >= 0)
a.s1 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s1 * src_step + map1_dataDx.s1 + src_offset));
if (map1_dataDx.s2 < src_cols && map1_dataDx.s2 >= 0 && map1_dataDy.s2 < src_rows && map1_dataDy.s2 >= 0)
a.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s2 * src_step + map1_dataDx.s2 + src_offset));
if (map1_dataDx.s3 < src_cols && map1_dataDx.s3 >= 0 && map1_dataDy.s3 < src_rows && map1_dataDy.s3 >= 0)
a.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s3 * src_step + map1_dataDx.s3 + src_offset));
if (map1_dataDx1.s0 < src_cols && map1_dataDx1.s0 >= 0 && map1_dataDy.s0 < src_rows && map1_dataDy.s0 >= 0)
b.s0 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s0 * src_step + map1_dataDx1.s0 + src_offset));
if (map1_dataDx1.s1 < src_cols && map1_dataDx1.s1 >= 0 && map1_dataDy.s1 < src_rows && map1_dataDy.s1 >= 0)
b.s1 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s1 * src_step + map1_dataDx1.s1 + src_offset));
if (map1_dataDx1.s2 < src_cols && map1_dataDx1.s2 >= 0 && map1_dataDy.s2 < src_rows && map1_dataDy.s2 >= 0)
b.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s2 * src_step + map1_dataDx1.s2 + src_offset));
if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy.s3 < src_rows && map1_dataDy.s3 >= 0)
b.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy.s3 * src_step + map1_dataDx1.s3 + src_offset));
if (map1_dataDx.s0 < src_cols && map1_dataDx.s0 >= 0 && map1_dataDy1.s0 < src_rows && map1_dataDy1.s0 >= 0)
c.s0 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s0 * src_step + map1_dataDx.s0 + src_offset));
if (map1_dataDx.s1 < src_cols && map1_dataDx.s1 >= 0 && map1_dataDy1.s1 < src_rows && map1_dataDy1.s1 >= 0)
c.s1 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s1 * src_step + map1_dataDx.s1 + src_offset));
if (map1_dataDx.s2 < src_cols && map1_dataDx.s2 >= 0 && map1_dataDy1.s2 < src_rows && map1_dataDy1.s2 >= 0)
c.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s2 * src_step + map1_dataDx.s2 + src_offset));
if (map1_dataDx.s3 < src_cols && map1_dataDx.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
c.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s3 * src_step + map1_dataDx.s3 + src_offset));
if (map1_dataDx1.s0 < src_cols && map1_dataDx1.s0 >= 0 && map1_dataDy1.s0 < src_rows && map1_dataDy1.s0 >= 0)
d.s0 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s0 * src_step + map1_dataDx1.s0 + src_offset));
if (map1_dataDx1.s1 < src_cols && map1_dataDx1.s1 >= 0 && map1_dataDy1.s1 < src_rows && map1_dataDy1.s1 >= 0)
d.s1 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s1 * src_step + map1_dataDx1.s1 + src_offset));
if (map1_dataDx1.s2 < src_cols && map1_dataDx1.s2 >= 0 && map1_dataDy1.s2 < src_rows && map1_dataDy1.s2 >= 0)
d.s2 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s2 * src_step + map1_dataDx1.s2 + src_offset));
if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
d.s3 = *((__global uchar*)((__global uchar *)src + map1_dataDy1.s3 * src_step + map1_dataDx1.s3 + src_offset));
uchar4 dst_data = convert_uchar4_sat_rte((convert_float4(a))* ud * vd +(convert_float4(b))* u * vd + (convert_float4(c))* ud * v + (convert_float4(d)) * u * v );
__global uchar4* D = (__global uchar4 *)(dst + dstStart);
uchar4 dVal = *D;
int4 con = (Gx >= 0 && Gx < dst_cols && y >= 0 && y < dst_rows);
dst_data = (convert_uchar4(con) != (uchar4)(0)) ? dst_data : dVal;
*D = dst_data;
}
}
float2 map_data = map1[map1Idx];
__kernel void remapLNFConstant_C4_D0(__global unsigned char* dst, __global unsigned char const * restrict src, int2 map_dataA = convert_int2_sat_rtn(map_data);
__global float * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
int dstIdx = y * dst_step + (x << 2) + dst_offset;
int mapIdx = y * map1_step + (x << 3) + map1_offset;
float2 map_data = *((__global float2 *)((__global char*)map1 + mapIdx));
int2 map_dataA = convert_int2(map_data);
float2 u = map_data - convert_float2(map_dataA);
int2 map_dataB = (int2)(map_dataA.x + 1, map_dataA.y); int2 map_dataB = (int2)(map_dataA.x + 1, map_dataA.y);
int2 map_dataC = (int2)(map_dataA.x, map_dataA.y + 1); int2 map_dataC = (int2)(map_dataA.x, map_dataA.y + 1);
int2 map_dataD = (int2)(map_dataA.x + 1, map_dataA.y +1); int2 map_dataD = (int2)(map_dataA.x + 1, map_dataA.y +1);
uchar4 nval = convert_uchar4(nVal);
uchar4 a, b, c , d;
if(map_dataA.x < 0 || map_dataA.x >= src_cols || map_dataA.y >= src_rows || map_dataA.y < 0)
a = nval;
else
a = *((__global uchar4 *)((__global uchar *)src + map_dataA.y * src_step + (map_dataA.x<<2) + src_offset ));
if(map_dataB.x < 0 || map_dataB.x >= src_cols || map_dataB.y >= src_rows || map_dataB.y < 0)
b = nval;
else
b = *((__global uchar4 *)((__global uchar *)src + map_dataB.y * src_step + (map_dataB.x<<2) + src_offset ));
if(map_dataC.x < 0 || map_dataC.x >= src_cols || map_dataC.y >= src_rows || map_dataC.y < 0) float2 _u = map_data - convert_float2(map_dataA);
c = nval; WT2 u = convertToWT2(convert_int2_rte(convertToWT2(_u) * (WT2)32)) / (WT2)32;
else WT nval = convertToWT(nVal);
c = *((__global uchar4 *)((__global uchar *)src + map_dataC.y * src_step + (map_dataC.x<<2) + src_offset )); WT a = nval, b = nval, c = nval, d = nval;
if(map_dataD.x < 0 || map_dataD.x >= src_cols || map_dataD.y >= src_rows || map_dataD.y < 0) if (!NEED_EXTRAPOLATION(map_dataA.x, map_dataA.y))
d = nval; a = convertToWT(src[mad24(map_dataA.y, src_step, map_dataA.x + src_offset)]);
else else
d = *((__global uchar4 *)((__global uchar *)src + map_dataD.y * src_step + (map_dataD.x<<2) + src_offset )); {
float4 dst_data = convert_float4(a)*((float4)(1.0-u.x)*((float4)(1.0-u.y))) + convert_float4(b)*((float4)(u.x))*((float4)(1.0-u.y)) + convert_float4(c)*((float4)(1.0-u.x))*((float4)(u.y)) + convert_float4(d)*((float4)(u.x))*((float4)(u.y)); #ifdef BORDER_CONSTANT
*((__global uchar4 *)((__global uchar*)dst + dstIdx)) = convert_uchar4_sat_rte(dst_data); #else
#error No extrapolation method
#endif
} }
}
__kernel void remapLNF1Constant_C4_D0(__global unsigned char* dst, __global unsigned char const * restrict src,
__global float * map1, __global float * map2, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0); if (!NEED_EXTRAPOLATION(map_dataB.x, map_dataB.y))
int y = get_global_id(1); b = convertToWT(src[mad24(map_dataB.y, src_step, map_dataB.x + src_offset)]);
if(x < threadCols && y < dst_rows)
{
int dstIdx = y * dst_step + (x << 2) + dst_offset;
int mapIdx = y * map1_step + (x << 2) + map1_offset;
float map1_data = *((__global float *)((__global char*)map1 + mapIdx));
float map2_data = *((__global float *)((__global char*)map2 + mapIdx));
float2 map_data = (float2)(map1_data, map2_data);
int2 map_dataA = convert_int2(map_data);
float2 u = map_data - convert_float2(map_dataA);
int2 map_dataB = (int2)(map_dataA.x + 1, map_dataA.y);
int2 map_dataC = (int2)(map_dataA.x, map_dataA.y + 1);
int2 map_dataD = (int2)(map_dataA.x + 1, map_dataA.y +1);
uchar4 nval = convert_uchar4(nVal);
uchar4 a, b, c , d;
if(map_dataA.x < 0 || map_dataA.x >= src_cols || map_dataA.y >= src_rows || map_dataA.y < 0)
a = nval;
else else
a = *((__global uchar4 *)((__global uchar *)src + map_dataA.y * src_step + (map_dataA.x<<2) + src_offset )); {
if(map_dataB.x < 0 || map_dataB.x >= src_cols || map_dataB.y >= src_rows || map_dataB.y < 0) #ifdef BORDER_CONSTANT
b = nval; #else
else #error No extrapolation method
b = *((__global uchar4 *)((__global uchar *)src + map_dataB.y * src_step + (map_dataB.x<<2) + src_offset )); #endif
}
if(map_dataC.x < 0 || map_dataC.x >= src_cols || map_dataC.y >= src_rows || map_dataC.y < 0) if (!NEED_EXTRAPOLATION(map_dataC.x, map_dataC.y))
c = nval; c = convertToWT(src[mad24(map_dataC.y, src_step, map_dataC.x + src_offset)]);
else else
c = *((__global uchar4 *)((__global uchar *)src + map_dataC.y * src_step + (map_dataC.x<<2) + src_offset )); {
#ifdef BORDER_CONSTANT
#else
#error No extrapolation method
#endif
}
if(map_dataD.x < 0 || map_dataD.x >= src_cols || map_dataD.y >= src_rows || map_dataD.y < 0) if (!NEED_EXTRAPOLATION(map_dataD.x, map_dataD.y))
d = nval; d = convertToWT(src[mad24(map_dataD.y, src_step, map_dataD.x + src_offset)]);
else else
d = *((__global uchar4 *)((__global uchar *)src + map_dataD.y * src_step + (map_dataD.x<<2) + src_offset )); {
float4 dst_data = convert_float4(a)*((float4)(1.0-u.x)*((float4)(1.0-u.y))) + convert_float4(b)*((float4)(u.x))*((float4)(1.0-u.y)) + convert_float4(c)*((float4)(1.0-u.x))*((float4)(u.y)) + convert_float4(d)*((float4)(u.x))*((float4)(u.y)); #ifdef BORDER_CONSTANT
*((__global uchar4 *)((__global uchar*)dst + dstIdx)) = convert_uchar4_sat_rte(dst_data); #else
#error No extrapolation method
#endif
}
}
}
__kernel void remapLNFConstant_C1_D5(__global float* dst, __global float const * restrict src,
__global float * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
x = x << 4;
int gx = x - (dst_offset&15);
int4 Gx = (int4)(gx, gx+4, gx+8, gx+12);
float4 nval =convert_float4(nVal);
float4 val = (float4)(nval.s0);
int dstStart = (y * dst_step + x + dst_offset) - (dst_offset&15);
int map1Start = y * map1_step + (x << 1) + map1_offset - ((dst_offset & 15) << 1);
float8 map1_data;
map1_data = *((__global float8 *)((__global char*)map1 + map1Start));
int8 map1_dataD = convert_int8(map1_data);
float8 temp = map1_data - convert_float8(map1_dataD);
float4 u = temp.even;
float4 v = temp.odd;
float4 ud = (float4)(1.0) - u;
float4 vd = (float4)(1.0) - v;
//float8 map1_dataU = map1_dataD + 1;
int4 map1_dataDx = map1_dataD.even;
int4 map1_dataDy = map1_dataD.odd;
int4 map1_dataDx1 = map1_dataDx + (int4)(1);
int4 map1_dataDy1 = map1_dataDy + (int4)(1);
float4 a = val, b = val, c = val, d = val;
if (map1_dataDx.s0 < src_cols && map1_dataDx.s0 >= 0 && map1_dataDy.s0 < src_rows && map1_dataDy.s0 >= 0)
a.s0 = *((__global float*)((__global uchar *)src + map1_dataDy.s0 * src_step + (map1_dataDx.s0 << 2) + src_offset));
if (map1_dataDx.s1 < src_cols && map1_dataDx.s1 >= 0 && map1_dataDy.s1 < src_rows && map1_dataDy.s1 >= 0)
a.s1 = *((__global float*)((__global uchar *)src + map1_dataDy.s1 * src_step + (map1_dataDx.s1 << 2) + src_offset));
if (map1_dataDx.s2 < src_cols && map1_dataDx.s2 >= 0 && map1_dataDy.s2 < src_rows && map1_dataDy.s2 >= 0)
a.s2 = *((__global float*)((__global uchar *)src + map1_dataDy.s2 * src_step + (map1_dataDx.s2 << 2) + src_offset));
if (map1_dataDx.s3 < src_cols && map1_dataDx.s3 >= 0 && map1_dataDy.s3 < src_rows && map1_dataDy.s3 >= 0)
a.s3 = *((__global float*)((__global uchar *)src + map1_dataDy.s3 * src_step + (map1_dataDx.s3 << 2) + src_offset));
if (map1_dataDx1.s0 < src_cols && map1_dataDx1.s0 >= 0 && map1_dataDy.s0 < src_rows && map1_dataDy.s0 >= 0)
b.s0 = *((__global float*)((__global uchar *)src + map1_dataDy.s0 * src_step + (map1_dataDx1.s0 << 2) + src_offset));
if (map1_dataDx1.s1 < src_cols && map1_dataDx1.s1 >= 0 && map1_dataDy.s1 < src_rows && map1_dataDy.s1 >= 0)
b.s1 = *((__global float*)((__global uchar *)src + map1_dataDy.s1 * src_step + (map1_dataDx1.s1 << 2) + src_offset));
if (map1_dataDx1.s2 < src_cols && map1_dataDx1.s2 >= 0 && map1_dataDy.s2 < src_rows && map1_dataDy.s2 >= 0)
b.s2 = *((__global float*)((__global uchar *)src + map1_dataDy.s2 * src_step + (map1_dataDx1.s2 << 2) + src_offset));
if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy.s3 < src_rows && map1_dataDy.s3 >= 0)
b.s3 = *((__global float*)((__global uchar *)src + map1_dataDy.s3 * src_step + (map1_dataDx1.s3 << 2) + src_offset));
if (map1_dataDx.s0 < src_cols && map1_dataDx.s0 >= 0 && map1_dataDy1.s0 < src_rows && map1_dataDy1.s0 >= 0)
c.s0 = *((__global float*)((__global uchar *)src + map1_dataDy1.s0 * src_step + (map1_dataDx.s0 << 2) + src_offset));
if (map1_dataDx.s1 < src_cols && map1_dataDx.s1 >= 0 && map1_dataDy1.s1 < src_rows && map1_dataDy1.s1 >= 0)
c.s1 = *((__global float*)((__global uchar *)src + map1_dataDy1.s1 * src_step + (map1_dataDx.s1 << 2) + src_offset));
if (map1_dataDx.s2 < src_cols && map1_dataDx.s2 >= 0 && map1_dataDy1.s2 < src_rows && map1_dataDy1.s2 >= 0)
c.s2 = *((__global float*)((__global uchar *)src + map1_dataDy1.s2 * src_step + (map1_dataDx.s2 << 2) + src_offset));
if (map1_dataDx.s3 < src_cols && map1_dataDx.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
c.s3 = *((__global float*)((__global uchar *)src + map1_dataDy1.s3 * src_step + (map1_dataDx.s3 << 2) + src_offset));
if (map1_dataDx1.s0 < src_cols && map1_dataDx1.s0 >= 0 && map1_dataDy1.s0 < src_rows && map1_dataDy1.s0 >= 0)
d.s0 = *((__global float*)((__global uchar *)src + map1_dataDy1.s0 * src_step + (map1_dataDx1.s0 << 2) + src_offset));
if (map1_dataDx1.s1 < src_cols && map1_dataDx1.s1 >= 0 && map1_dataDy1.s1 < src_rows && map1_dataDy1.s1 >= 0)
d.s1 = *((__global float*)((__global uchar *)src + map1_dataDy1.s1 * src_step + (map1_dataDx1.s1 << 2) + src_offset));
if (map1_dataDx1.s2 < src_cols && map1_dataDx1.s2 >= 0 && map1_dataDy1.s2 < src_rows && map1_dataDy1.s2 >= 0)
d.s2 = *((__global float*)((__global uchar *)src + map1_dataDy1.s2 * src_step + (map1_dataDx1.s2 << 2) + src_offset));
if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
d.s3 = *((__global float*)((__global uchar *)src + map1_dataDy1.s3 * src_step + (map1_dataDx1.s3 << 2) + src_offset));
float4 dst_data = a * ud * vd + b * u * vd + c * ud * v + d * u * v ;
__global float4* D = (__global float4 *)((__global char*)dst + dstStart);
float4 dVal = *D;
int4 con = (Gx >= 0 && Gx < (dst_cols << 2) && y >= 0 && y < dst_rows);
dst_data = (convert_float4(con) != (float4)(0)) ? dst_data : dVal;
*D = dst_data;
}
}
__kernel void remapLNF1Constant_C1_D5(__global float* dst, __global float const * restrict src,
__global float * map1, __global float * map2, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
x = x << 4;
int gx = x - (dst_offset&15);
int4 Gx = (int4)(gx, gx+4, gx+8, gx+12);
float4 nval =convert_float4(nVal);
float4 val = (float4)(nval.s0);
int dstStart = y * dst_step + x + dst_offset - (dst_offset & 15);
int map1Start = y * map1_step + x + map1_offset - (dst_offset & 15);
float4 map1_data;
float4 map2_data;
map1_data = *((__global float4 *)((__global char*)map1 + map1Start));
map2_data = *((__global float4 *)((__global char*)map2 + map1Start));
float8 map_data = (float8)(map1_data.s0, map2_data.s0, map1_data.s1, map2_data.s1, map1_data.s2, map2_data.s2, map1_data.s3, map2_data.s3);
int8 map1_dataD = convert_int8(map_data);
float8 temp = map_data - convert_float8(map1_dataD);
float4 u = temp.even;
float4 v = temp.odd;
float4 ud = (float4)(1.0) - u;
float4 vd = (float4)(1.0) - v;
//float8 map1_dataU = map1_dataD + 1;
int4 map1_dataDx = map1_dataD.even;
int4 map1_dataDy = map1_dataD.odd;
int4 map1_dataDx1 = map1_dataDx + (int4)(1);
int4 map1_dataDy1 = map1_dataDy + (int4)(1);
float4 a = val, b = val, c = val, d = val;
if (map1_dataDx.s0 < src_cols && map1_dataDx.s0 >= 0 && map1_dataDy.s0 < src_rows && map1_dataDy.s0 >= 0)
a.s0 = *((__global float*)((__global uchar *)src + map1_dataDy.s0 * src_step + (map1_dataDx.s0 << 2) + src_offset));
if (map1_dataDx.s1 < src_cols && map1_dataDx.s1 >= 0 && map1_dataDy.s1 < src_rows && map1_dataDy.s1 >= 0)
a.s1 = *((__global float*)((__global uchar *)src + map1_dataDy.s1 * src_step + (map1_dataDx.s1 << 2) + src_offset));
if (map1_dataDx.s2 < src_cols && map1_dataDx.s2 >= 0 && map1_dataDy.s2 < src_rows && map1_dataDy.s2 >= 0)
a.s2 = *((__global float*)((__global uchar *)src + map1_dataDy.s2 * src_step + (map1_dataDx.s2 << 2) + src_offset));
if (map1_dataDx.s3 < src_cols && map1_dataDx.s3 >= 0 && map1_dataDy.s3 < src_rows && map1_dataDy.s3 >= 0)
a.s3 = *((__global float*)((__global uchar *)src + map1_dataDy.s3 * src_step + (map1_dataDx.s3 << 2) + src_offset));
if (map1_dataDx1.s0 < src_cols && map1_dataDx1.s0 >= 0 && map1_dataDy.s0 < src_rows && map1_dataDy.s0 >= 0)
b.s0 = *((__global float*)((__global uchar *)src + map1_dataDy.s0 * src_step + (map1_dataDx1.s0 << 2) + src_offset));
if (map1_dataDx1.s1 < src_cols && map1_dataDx1.s1 >= 0 && map1_dataDy.s1 < src_rows && map1_dataDy.s1 >= 0)
b.s1 = *((__global float*)((__global uchar *)src + map1_dataDy.s1 * src_step + (map1_dataDx1.s1 << 2) + src_offset));
if (map1_dataDx1.s2 < src_cols && map1_dataDx1.s2 >= 0 && map1_dataDy.s2 < src_rows && map1_dataDy.s2 >= 0)
b.s2 = *((__global float*)((__global uchar *)src + map1_dataDy.s2 * src_step + (map1_dataDx1.s2 << 2) + src_offset));
if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy.s3 < src_rows && map1_dataDy.s3 >= 0)
b.s3 = *((__global float*)((__global uchar *)src + map1_dataDy.s3 * src_step + (map1_dataDx1.s3 << 2) + src_offset));
if (map1_dataDx.s0 < src_cols && map1_dataDx.s0 >= 0 && map1_dataDy1.s0 < src_rows && map1_dataDy1.s0 >= 0)
c.s0 = *((__global float*)((__global uchar *)src + map1_dataDy1.s0 * src_step + (map1_dataDx.s0 << 2) + src_offset));
if (map1_dataDx.s1 < src_cols && map1_dataDx.s1 >= 0 && map1_dataDy1.s1 < src_rows && map1_dataDy1.s1 >= 0)
c.s1 = *((__global float*)((__global uchar *)src + map1_dataDy1.s1 * src_step + (map1_dataDx.s1 << 2) + src_offset));
if (map1_dataDx.s2 < src_cols && map1_dataDx.s2 >= 0 && map1_dataDy1.s2 < src_rows && map1_dataDy1.s2 >= 0)
c.s2 = *((__global float*)((__global uchar *)src + map1_dataDy1.s2 * src_step + (map1_dataDx.s2 << 2) + src_offset));
if (map1_dataDx.s3 < src_cols && map1_dataDx.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
c.s3 = *((__global float*)((__global uchar *)src + map1_dataDy1.s3 * src_step + (map1_dataDx.s3 << 2) + src_offset));
if (map1_dataDx1.s0 < src_cols && map1_dataDx1.s0 >= 0 && map1_dataDy1.s0 < src_rows && map1_dataDy1.s0 >= 0)
d.s0 = *((__global float*)((__global uchar *)src + map1_dataDy1.s0 * src_step + (map1_dataDx1.s0 << 2) + src_offset));
if (map1_dataDx1.s1 < src_cols && map1_dataDx1.s1 >= 0 && map1_dataDy1.s1 < src_rows && map1_dataDy1.s1 >= 0)
d.s1 = *((__global float*)((__global uchar *)src + map1_dataDy1.s1 * src_step + (map1_dataDx1.s1 << 2) + src_offset));
if (map1_dataDx1.s2 < src_cols && map1_dataDx1.s2 >= 0 && map1_dataDy1.s2 < src_rows && map1_dataDy1.s2 >= 0)
d.s2 = *((__global float*)((__global uchar *)src + map1_dataDy1.s2 * src_step + (map1_dataDx1.s2 << 2) + src_offset));
if (map1_dataDx1.s3 < src_cols && map1_dataDx1.s3 >= 0 && map1_dataDy1.s3 < src_rows && map1_dataDy1.s3 >= 0)
d.s3 = *((__global float*)((__global uchar *)src + map1_dataDy1.s3 * src_step + (map1_dataDx1.s3 << 2) + src_offset));
float4 dst_data = a * ud * vd + b * u * vd + c * ud * v + d * u * v ;
__global float4* D = (__global float4 *)((__global char*)dst + dstStart);
float4 dVal = *D;
int4 con = (Gx >= 0 && Gx < (dst_cols << 2) && y >= 0 && y < dst_rows);
dst_data = (convert_float4(con) != (float4)(0)) ? dst_data : dVal;
*D = dst_data;
}
}
__kernel void remapLNFConstant_C4_D5(__global float * dst, __global float const * restrict src,
__global float * map1, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
int dstIdx = y * dst_step + (x << 4) + dst_offset ;
int mapIdx = y * map1_step + (x << 3) + map1_offset ;
float2 map_data = *((__global float2 *)((__global char*)map1 + mapIdx));
int2 map_dataA = convert_int2(map_data);
float2 u = map_data - convert_float2(map_dataA);
int2 map_dataB = (int2)(map_dataA.x + 1, map_dataA.y);
int2 map_dataC = (int2)(map_dataA.x, map_dataA.y + 1);
int2 map_dataD = (int2)(map_dataA.x + 1, map_dataA.y +1);
float4 nval = convert_float4(nVal);
float4 a, b, c , d;
if(map_dataA.x < 0 || map_dataA.x >= src_cols || map_dataA.y >= src_rows || map_dataA.y < 0)
a = nval;
else
a = *((__global float4 *)((__global uchar *)src + map_dataA.y * src_step + (map_dataA.x<<4) + src_offset ));
if(map_dataB.x < 0 || map_dataB.x >= src_cols || map_dataB.y >= src_rows || map_dataB.y < 0)
b = nval;
else
b = *((__global float4 *)((__global uchar *)src + map_dataB.y * src_step + (map_dataB.x<<4) + src_offset ));
if(map_dataC.x < 0 || map_dataC.x >= src_cols || map_dataC.y >= src_rows || map_dataC.y < 0)
c = nval;
else
c = *((__global float4 *)((__global uchar *)src + map_dataC.y * src_step + (map_dataC.x<<4) + src_offset ));
if(map_dataD.x < 0 || map_dataD.x >= src_cols || map_dataD.y >= src_rows || map_dataD.y < 0)
d = nval;
else
d = *((__global float4 *)((__global uchar *)src + map_dataD.y * src_step + (map_dataD.x<<4) + src_offset ));
float4 dst_data = a * ((float4)(1.0-u.x)) * ((float4)(1.0-u.y)) + b *((float4)(u.x)) * ((float4)(1.0-u.y)) + c * ((float4)(1.0-u.x)) *((float4)(u.y)) + d *((float4)(u.x)) *((float4)(u.y));
*((__global float4 *)((__global uchar*)dst + dstIdx)) = dst_data ;
WT dst_data = a * (WT)(1.0 - u.x) * (WT)(1.0 - u.y) +
b * (WT)(u.x) * (WT)(1.0 - u.y) +
c * (WT)(1.0 - u.x) * (WT)(u.y) +
d * (WT)(u.x) * (WT)(u.y);
dst[dstIdx] = convertToT(dst_data);
} }
} }
__kernel void remapLNF1Constant_C4_D5(__global float * dst, __global float const * restrict src, #endif
__global float * map1, __global float * map2, int dst_offset, int src_offset, int map1_offset, int dst_step, int src_step,
int map1_step, int src_cols, int src_rows, int dst_cols, int dst_rows, int map1_cols, int map1_rows , int threadCols, F4 nVal)
{
int x = get_global_id(0);
int y = get_global_id(1);
if(x < threadCols && y < dst_rows)
{
int dstIdx = y * dst_step + (x << 4) + dst_offset ;
int mapIdx = y * map1_step + (x << 2) + map1_offset ;
float map1_data = *((__global float *)((__global char*)map1 + mapIdx));
float map2_data = *((__global float *)((__global char*)map2 + mapIdx));
float2 map_data = (float2)(map1_data, map2_data);
int2 map_dataA = convert_int2(map_data);
float2 u = map_data - convert_float2(map_dataA);
int2 map_dataB = (int2)(map_dataA.x + 1, map_dataA.y);
int2 map_dataC = (int2)(map_dataA.x, map_dataA.y + 1);
int2 map_dataD = (int2)(map_dataA.x + 1, map_dataA.y +1);
float4 nval = convert_float4(nVal);
float4 a, b, c , d;
if(map_dataA.x < 0 || map_dataA.x >= src_cols || map_dataA.y >= src_rows || map_dataA.y < 0)
a = nval;
else
a = *((__global float4 *)((__global uchar *)src + map_dataA.y * src_step + (map_dataA.x<<4) + src_offset ));
if(map_dataB.x < 0 || map_dataB.x >= src_cols || map_dataB.y >= src_rows || map_dataB.y < 0)
b = nval;
else
b = *((__global float4 *)((__global uchar *)src + map_dataB.y * src_step + (map_dataB.x<<4) + src_offset ));
if(map_dataC.x < 0 || map_dataC.x >= src_cols || map_dataC.y >= src_rows || map_dataC.y < 0)
c = nval;
else
c = *((__global float4 *)((__global uchar *)src + map_dataC.y * src_step + (map_dataC.x<<4) + src_offset ));
if(map_dataD.x < 0 || map_dataD.x >= src_cols || map_dataD.y >= src_rows || map_dataD.y < 0)
d = nval;
else
d = *((__global float4 *)((__global uchar *)src + map_dataD.y * src_step + (map_dataD.x<<4) + src_offset ));
float4 dst_data = a * ((float4)(1.0-u.x)) * ((float4)(1.0-u.y)) + b *((float4)(u.x)) * ((float4)(1.0-u.y)) + c * ((float4)(1.0-u.x)) *((float4)(u.y)) + d *((float4)(u.x)) *((float4)(u.y));
*((__global float4 *)((__global uchar*)dst + dstIdx)) = dst_data ;
}
}
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment