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

fixed ocl::oclMat::setTo for 2-channel images

Browse Source
This commit is contained in:
Ilya Lavrenov
2013-09-20 14:22:18 +04:00
parent 94966b38c0
commit 362a67a695
5 changed files with 163 additions and 506 deletions
Download Patch File Download Diff File Expand all files Collapse all files

443
modules/ocl/src/matrix_operations.cpp
View File

@@ -445,200 +445,62 @@ void cv::ocl::oclMat::convertTo( oclMat &dst, int rtype, double alpha, double be
///////////////////////////////////////////////////////////////////////////
//////////////////////////////// setTo ////////////////////////////////////
///////////////////////////////////////////////////////////////////////////
oclMat &cv::ocl::oclMat::operator = (const Scalar &s)
{
setTo(s);
return *this;
}
static void set_to_withoutmask_run(const oclMat &dst, const Scalar &scalar, string kernelName)
{
vector<pair<size_t , const void *> > args;
size_t localThreads[3] = {16, 16, 1};
size_t globalThreads[3];
globalThreads[0] = (dst.cols + localThreads[0] - 1) / localThreads[0] * localThreads[0];
globalThreads[1] = (dst.rows + localThreads[1] - 1) / localThreads[1] * localThreads[1];
globalThreads[2] = 1;
size_t globalThreads[3] = { dst.cols, dst.rows, 1 };
int step_in_pixel = dst.step / dst.elemSize(), offset_in_pixel = dst.offset / dst.elemSize();
if(dst.type() == CV_8UC1)
{
if (dst.type() == CV_8UC1)
globalThreads[0] = ((dst.cols + 4) / 4 + localThreads[0] - 1) / localThreads[0] * localThreads[0];
}
char compile_option[32];
union sc
{
cl_uchar4 uval;
cl_char4 cval;
cl_ushort4 usval;
cl_short4 shval;
cl_int4 ival;
cl_float4 fval;
cl_double4 dval;
} val;
switch(dst.depth())
{
case CV_8U:
val.uval.s[0] = saturate_cast<uchar>(scalar.val[0]);
val.uval.s[1] = saturate_cast<uchar>(scalar.val[1]);
val.uval.s[2] = saturate_cast<uchar>(scalar.val[2]);
val.uval.s[3] = saturate_cast<uchar>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=uchar");
args.push_back( make_pair( sizeof(cl_uchar) , (void *)&val.uval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=uchar4");
args.push_back( make_pair( sizeof(cl_uchar4) , (void *)&val.uval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_8S:
val.cval.s[0] = saturate_cast<char>(scalar.val[0]);
val.cval.s[1] = saturate_cast<char>(scalar.val[1]);
val.cval.s[2] = saturate_cast<char>(scalar.val[2]);
val.cval.s[3] = saturate_cast<char>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=char");
args.push_back( make_pair( sizeof(cl_char) , (void *)&val.cval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=char4");
args.push_back( make_pair( sizeof(cl_char4) , (void *)&val.cval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_16U:
val.usval.s[0] = saturate_cast<ushort>(scalar.val[0]);
val.usval.s[1] = saturate_cast<ushort>(scalar.val[1]);
val.usval.s[2] = saturate_cast<ushort>(scalar.val[2]);
val.usval.s[3] = saturate_cast<ushort>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=ushort");
args.push_back( make_pair( sizeof(cl_ushort) , (void *)&val.usval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=ushort4");
args.push_back( make_pair( sizeof(cl_ushort4) , (void *)&val.usval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_16S:
val.shval.s[0] = saturate_cast<short>(scalar.val[0]);
val.shval.s[1] = saturate_cast<short>(scalar.val[1]);
val.shval.s[2] = saturate_cast<short>(scalar.val[2]);
val.shval.s[3] = saturate_cast<short>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=short");
args.push_back( make_pair( sizeof(cl_short) , (void *)&val.shval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=short4");
args.push_back( make_pair( sizeof(cl_short4) , (void *)&val.shval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_32S:
val.ival.s[0] = saturate_cast<int>(scalar.val[0]);
val.ival.s[1] = saturate_cast<int>(scalar.val[1]);
val.ival.s[2] = saturate_cast<int>(scalar.val[2]);
val.ival.s[3] = saturate_cast<int>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=int");
args.push_back( make_pair( sizeof(cl_int) , (void *)&val.ival.s[0] ));
break;
case 2:
sprintf(compile_option, "-D GENTYPE=int2");
cl_int2 i2val;
i2val.s[0] = val.ival.s[0];
i2val.s[1] = val.ival.s[1];
args.push_back( make_pair( sizeof(cl_int2) , (void *)&i2val ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=int4");
args.push_back( make_pair( sizeof(cl_int4) , (void *)&val.ival ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_32F:
val.fval.s[0] = scalar.val[0];
val.fval.s[1] = scalar.val[1];
val.fval.s[2] = scalar.val[2];
val.fval.s[3] = scalar.val[3];
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=float");
args.push_back( make_pair( sizeof(cl_float) , (void *)&val.fval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=float4");
args.push_back( make_pair( sizeof(cl_float4) , (void *)&val.fval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_64F:
val.dval.s[0] = scalar.val[0];
val.dval.s[1] = scalar.val[1];
val.dval.s[2] = scalar.val[2];
val.dval.s[3] = scalar.val[3];
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=double");
args.push_back( make_pair( sizeof(cl_double) , (void *)&val.dval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=double4");
args.push_back( make_pair( sizeof(cl_double4) , (void *)&val.dval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unknown depth");
}
const char * const typeMap[] = { "uchar", "char", "ushort", "short", "int", "float", "double" };
const char channelMap[] = { ' ', ' ', '2', '4', '4' };
std::string buildOptions = format("-D GENTYPE=%s%c", typeMap[dst.depth()], channelMap[dst.channels()]);
Mat mat(1, 1, dst.type(), scalar);
#ifdef CL_VERSION_1_2
//this enables backwards portability to
//run on OpenCL 1.1 platform if library binaries are compiled with OpenCL 1.2 support
if(Context::getContext()->supportsFeature(Context::CL_VER_1_2) &&
// this enables backwards portability to
// run on OpenCL 1.1 platform if library binaries are compiled with OpenCL 1.2 support
if (Context::getContext()->supportsFeature(Context::CL_VER_1_2) &&
dst.offset == 0 && dst.cols == dst.wholecols)
{
const int sizeofMap[][7] =
{
{ sizeof(cl_uchar) , sizeof(cl_char) , sizeof(cl_ushort) , sizeof(cl_short) , sizeof(cl_int) , sizeof(cl_float) , sizeof(cl_double) },
{ sizeof(cl_uchar2), sizeof(cl_char2), sizeof(cl_ushort2), sizeof(cl_short2), sizeof(cl_int2), sizeof(cl_float2), sizeof(cl_double2) },
{ 0 , 0 , 0 , 0 , 0 , 0 , 0 },
{ sizeof(cl_uchar4), sizeof(cl_char4), sizeof(cl_ushort4), sizeof(cl_short4), sizeof(cl_int4), sizeof(cl_float4), sizeof(cl_double4) },
};
int sizeofGeneric = sizeofMap[dst.oclchannels() - 1][dst.depth()];
clEnqueueFillBuffer((cl_command_queue)dst.clCxt->oclCommandQueue(),
(cl_mem)dst.data, args[0].second, args[0].first, 0, dst.step * dst.rows, 0, NULL, NULL);
(cl_mem)dst.data, (void*)mat.data, sizeofGeneric,
0, dst.step * dst.rows, 0, NULL, NULL);
}
else
#endif
{
oclMat m(mat);
args.push_back( make_pair( sizeof(cl_mem) , (void*)&m.data ));
args.push_back( make_pair( sizeof(cl_mem) , (void *)&dst.data ));
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 *)&step_in_pixel ));
args.push_back( make_pair( sizeof(cl_int) , (void *)&offset_in_pixel));
args.push_back( make_pair( sizeof(cl_int) , (void *)&offset_in_pixel ));
openCLExecuteKernel(dst.clCxt , &operator_setTo, kernelName, globalThreads,
localThreads, args, -1, -1, compile_option);
localThreads, args, -1, -1, buildOptions.c_str());
}
}
@@ -646,161 +508,16 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
{
CV_DbgAssert( dst.rows == mask.rows && dst.cols == mask.cols);
vector<pair<size_t , const void *> > args;
size_t localThreads[3] = {16, 16, 1};
size_t globalThreads[3];
globalThreads[0] = (dst.cols + localThreads[0] - 1) / localThreads[0] * localThreads[0];
globalThreads[1] = (dst.rows + localThreads[1] - 1) / localThreads[1] * localThreads[1];
globalThreads[2] = 1;
size_t localThreads[3] = { 16, 16, 1 };
size_t globalThreads[3] = { dst.cols, dst.rows, 1 };
int step_in_pixel = dst.step / dst.elemSize(), offset_in_pixel = dst.offset / dst.elemSize();
char compile_option[32];
union sc
{
cl_uchar4 uval;
cl_char4 cval;
cl_ushort4 usval;
cl_short4 shval;
cl_int4 ival;
cl_float4 fval;
cl_double4 dval;
} val;
switch(dst.depth())
{
case CV_8U:
val.uval.s[0] = saturate_cast<uchar>(scalar.val[0]);
val.uval.s[1] = saturate_cast<uchar>(scalar.val[1]);
val.uval.s[2] = saturate_cast<uchar>(scalar.val[2]);
val.uval.s[3] = saturate_cast<uchar>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=uchar");
args.push_back( make_pair( sizeof(cl_uchar) , (void *)&val.uval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=uchar4");
args.push_back( make_pair( sizeof(cl_uchar4) , (void *)&val.uval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_8S:
val.cval.s[0] = saturate_cast<char>(scalar.val[0]);
val.cval.s[1] = saturate_cast<char>(scalar.val[1]);
val.cval.s[2] = saturate_cast<char>(scalar.val[2]);
val.cval.s[3] = saturate_cast<char>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=char");
args.push_back( make_pair( sizeof(cl_char) , (void *)&val.cval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=char4");
args.push_back( make_pair( sizeof(cl_char4) , (void *)&val.cval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_16U:
val.usval.s[0] = saturate_cast<ushort>(scalar.val[0]);
val.usval.s[1] = saturate_cast<ushort>(scalar.val[1]);
val.usval.s[2] = saturate_cast<ushort>(scalar.val[2]);
val.usval.s[3] = saturate_cast<ushort>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=ushort");
args.push_back( make_pair( sizeof(cl_ushort) , (void *)&val.usval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=ushort4");
args.push_back( make_pair( sizeof(cl_ushort4) , (void *)&val.usval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_16S:
val.shval.s[0] = saturate_cast<short>(scalar.val[0]);
val.shval.s[1] = saturate_cast<short>(scalar.val[1]);
val.shval.s[2] = saturate_cast<short>(scalar.val[2]);
val.shval.s[3] = saturate_cast<short>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=short");
args.push_back( make_pair( sizeof(cl_short) , (void *)&val.shval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=short4");
args.push_back( make_pair( sizeof(cl_short4) , (void *)&val.shval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_32S:
val.ival.s[0] = saturate_cast<int>(scalar.val[0]);
val.ival.s[1] = saturate_cast<int>(scalar.val[1]);
val.ival.s[2] = saturate_cast<int>(scalar.val[2]);
val.ival.s[3] = saturate_cast<int>(scalar.val[3]);
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=int");
args.push_back( make_pair( sizeof(cl_int) , (void *)&val.ival.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=int4");
args.push_back( make_pair( sizeof(cl_int4) , (void *)&val.ival ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_32F:
val.fval.s[0] = scalar.val[0];
val.fval.s[1] = scalar.val[1];
val.fval.s[2] = scalar.val[2];
val.fval.s[3] = scalar.val[3];
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=float");
args.push_back( make_pair( sizeof(cl_float) , (void *)&val.fval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=float4");
args.push_back( make_pair( sizeof(cl_float4) , (void *)&val.fval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
case CV_64F:
val.dval.s[0] = scalar.val[0];
val.dval.s[1] = scalar.val[1];
val.dval.s[2] = scalar.val[2];
val.dval.s[3] = scalar.val[3];
switch(dst.oclchannels())
{
case 1:
sprintf(compile_option, "-D GENTYPE=double");
args.push_back( make_pair( sizeof(cl_double) , (void *)&val.dval.s[0] ));
break;
case 4:
sprintf(compile_option, "-D GENTYPE=double4");
args.push_back( make_pair( sizeof(cl_double4) , (void *)&val.dval ));
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unsupported channels");
}
break;
default:
CV_Error(CV_StsUnsupportedFormat, "unknown depth");
}
const char * const typeMap[] = { "uchar", "char", "ushort", "short", "int", "float", "double" };
const char channelMap[] = { ' ', ' ', '2', '4', '4' };
std::string buildOptions = format("-D GENTYPE=%s%c", typeMap[dst.depth()], channelMap[dst.channels()]);
oclMat m(Mat(1, 1, dst.type(), scalar));
args.push_back( make_pair( sizeof(cl_mem) , (void *)&m.data ));
args.push_back( make_pair( sizeof(cl_mem) , (void *)&dst.data ));
args.push_back( make_pair( sizeof(cl_int) , (void *)&dst.cols ));
args.push_back( make_pair( sizeof(cl_int) , (void *)&dst.rows ));
@@ -810,38 +527,21 @@ static void set_to_withmask_run(const oclMat &dst, const Scalar &scalar, const o
args.push_back( make_pair( sizeof(cl_int) , (void *)&mask.step ));
args.push_back( make_pair( sizeof(cl_int) , (void *)&mask.offset ));
openCLExecuteKernel(dst.clCxt , &operator_setToM, kernelName, globalThreads,
localThreads, args, -1, -1, compile_option);
localThreads, args, -1, -1, buildOptions.c_str());
}
oclMat &cv::ocl::oclMat::setTo(const Scalar &scalar, const oclMat &mask)
{
//cout << "cv::ocl::oclMat::setTo()" << endl;
CV_Assert(mask.type() == CV_8UC1);
CV_Assert( this->depth() >= 0 && this->depth() <= 6 );
CV_DbgAssert( !this->empty());
//cl_int status;
//cl_mem mem;
//mem = clCreateBuffer(this->clCxt->clContext,CL_MEM_READ_WRITE,
// sizeof(double)*4,NULL,&status);
//openCLVerifyCall(status);
//double* s = (double *)scalar.val;
//openCLSafeCall(clEnqueueWriteBuffer(this->clCxt->clCmdQueue,
// (cl_mem)mem,1,0,sizeof(double)*4,s,0,0,0));
if (mask.empty())
{
if(type() == CV_8UC1)
{
set_to_withoutmask_run(*this, scalar, "set_to_without_mask_C1_D0");
}
else
{
set_to_withoutmask_run(*this, scalar, "set_to_without_mask");
}
set_to_withoutmask_run(*this, scalar, type() == CV_8UC1 ?
"set_to_without_mask_C1_D0" : "set_to_without_mask");
}
else
{
set_to_withmask_run(*this, scalar, mask, "set_to_with_mask");
}
return *this;
}
@@ -849,91 +549,43 @@ oclMat &cv::ocl::oclMat::setTo(const Scalar &scalar, const oclMat &mask)
oclMat cv::ocl::oclMat::reshape(int new_cn, int new_rows) const
{
if( new_rows != 0 && new_rows != rows)
{
CV_Error( CV_StsBadFunc,
"oclMat's number of rows can not be changed for current version" );
}
CV_Error( CV_StsBadFunc, "oclMat's number of rows can not be changed for current version" );
oclMat hdr = *this;
int cn = oclchannels();
if (new_cn == 0)
new_cn = cn;
int total_width = cols * cn;
if ((new_cn > total_width || total_width % new_cn != 0) && new_rows == 0)
new_rows = rows * total_width / new_cn;
if (new_rows != 0 && new_rows != rows)
{
int total_size = total_width * rows;
if (!isContinuous())
CV_Error(CV_BadStep, "The matrix is not continuous, thus its number of rows can not be changed");
if ((unsigned)new_rows > (unsigned)total_size)
CV_Error(CV_StsOutOfRange, "Bad new number of rows");
total_width = total_size / new_rows;
if (total_width * new_rows != total_size)
CV_Error(CV_StsBadArg, "The total number of matrix elements is not divisible by the new number of rows");
hdr.rows = new_rows;
hdr.step = total_width * elemSize1();
}
int new_width = total_width / new_cn;
if (new_width * new_cn != total_width)
CV_Error(CV_BadNumChannels, "The total width is not divisible by the new number of channels");
hdr.cols = new_width;
hdr.wholecols = new_width;
hdr.flags = (hdr.flags & ~CV_MAT_CN_MASK) | ((new_cn - 1) << CV_CN_SHIFT);
return hdr;
}
@@ -953,11 +605,6 @@ void cv::ocl::oclMat::createEx(int _rows, int _cols, int _type, DevMemRW rw_type
clCxt = Context::getContext();
/* core logic */
_type &= TYPE_MASK;
//download_channels = CV_MAT_CN(_type);
//if(download_channels==3)
//{
// _type = CV_MAKE_TYPE((CV_MAT_DEPTH(_type)),4);
//}
if( rows == _rows && cols == _cols && type() == _type && data )
return;
if( data )
@@ -974,7 +621,6 @@ void cv::ocl::oclMat::createEx(int _rows, int _cols, int _type, DevMemRW rw_type
void *dev_ptr;
openCLMallocPitchEx(clCxt, &dev_ptr, &step, GPU_MATRIX_MALLOC_STEP(esz * cols), rows, rw_type, mem_type);
//openCLMallocPitch(clCxt,&dev_ptr, &step, esz * cols, rows);
if (esz * cols == step)
flags |= Mat::CONTINUOUS_FLAG;
@@ -992,7 +638,6 @@ void cv::ocl::oclMat::createEx(int _rows, int _cols, int _type, DevMemRW rw_type
void cv::ocl::oclMat::release()
{
//cout << "cv::ocl::oclMat::release()" << endl;
if( refcount && CV_XADD(refcount, -1) == 1 )
{
fastFree(refcount);

25
modules/ocl/src/opencl/operator_copyToM.cl
View File

@@ -34,6 +34,14 @@
//
//
#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
__kernel void copy_to_with_mask(
__global const GENTYPE* restrict srcMat,
__global GENTYPE* dstMat,
@@ -47,16 +55,17 @@ __kernel void copy_to_with_mask(
int maskStep,
int maskoffset)
{
int x=get_global_id(0);
int y=get_global_id(1);
x = x< cols ? x: cols-1;
y = y< rows ? y: rows-1;
int srcidx = mad24(y,srcStep_in_pixel,x+ srcoffset_in_pixel);
int dstidx = mad24(y,dstStep_in_pixel,x+ dstoffset_in_pixel);
int x=get_global_id(0);
int y=get_global_id(1);
if (x < cols && y < rows)
{
int maskidx = mad24(y,maskStep,x+ maskoffset);
uchar mask = maskMat[maskidx];
if (mask)
if ( maskMat[maskidx])
{
int srcidx = mad24(y,srcStep_in_pixel,x+ srcoffset_in_pixel);
int dstidx = mad24(y,dstStep_in_pixel,x+ dstoffset_in_pixel);
dstMat[dstidx] = srcMat[srcidx];
}
}
}

25
modules/ocl/src/opencl/operator_setTo.cl
View File

@@ -34,17 +34,22 @@
//
//
#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
__kernel void set_to_without_mask_C1_D0(uchar scalar,__global uchar * dstMat,
__kernel void set_to_without_mask_C1_D0(__global uchar * scalar,__global uchar * dstMat,
int cols,int rows,int dstStep_in_pixel,int offset_in_pixel)
{
int x=get_global_id(0)<<2;
int y=get_global_id(1);
//int addr_start = mad24(y,dstStep_in_pixel,offset_in_pixel);
//int addr_end = mad24(y,dstStep_in_pixel,cols+offset_in_pixel);
int idx = mad24(y,dstStep_in_pixel,x+ offset_in_pixel);
uchar4 out;
out.x = out.y = out.z = out.w = scalar;
out.x = out.y = out.z = out.w = scalar[0];
if ( (x+3 < cols) && (y < rows)&& ((offset_in_pixel&3) == 0))
{
@@ -77,14 +82,14 @@ __kernel void set_to_without_mask_C1_D0(uchar scalar,__global uchar * dstMat,
}
}
__kernel void set_to_without_mask(GENTYPE scalar,__global GENTYPE * dstMat,
int cols,int rows,int dstStep_in_pixel,int offset_in_pixel)
__kernel void set_to_without_mask(__global GENTYPE * scalar,__global GENTYPE * dstMat,
int cols, int rows, int dstStep_in_pixel, int offset_in_pixel)
{
int x=get_global_id(0);
int y=get_global_id(1);
int x = get_global_id(0);
int y = get_global_id(1);
if ( (x < cols) & (y < rows))
{
int idx = mad24(y,dstStep_in_pixel,x+ offset_in_pixel);
dstMat[idx] = scalar;
int idx = mad24(y, dstStep_in_pixel, x + offset_in_pixel);
dstMat[idx] = scalar[0];
}
}

33
modules/ocl/src/opencl/operator_setToM.cl
View File

@@ -33,8 +33,17 @@
// 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
__kernel void set_to_with_mask(
GENTYPE scalar,
__global GENTYPE * scalar,
__global GENTYPE * dstMat,
int cols,
int rows,
@@ -44,16 +53,16 @@ __kernel void set_to_with_mask(
int maskStep,
int maskoffset)
{
int x=get_global_id(0);
int y=get_global_id(1);
x = x< cols ? x: cols-1;
y = y< rows ? y: rows-1;
int dstidx = mad24(y,dstStep_in_pixel,x+ dstoffset_in_pixel);
int maskidx = mad24(y,maskStep,x+ maskoffset);
uchar mask = maskMat[maskidx];
if (mask)
{
dstMat[dstidx] = scalar;
}
int x = get_global_id(0);
int y = get_global_id(1);
if (x < cols && y < rows)
{
int maskidx = mad24(y,maskStep,x+ maskoffset);
if (maskMat[maskidx])
{
int dstidx = mad24(y,dstStep_in_pixel,x+ dstoffset_in_pixel);
dstMat[dstidx] = scalar[0];
}
}
}