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

some optimizations to ocl::pyrDown, PyrLK and Canny

Browse Source
This commit is contained in:
bitwangyaoyao
2012-09-24 20:28:35 +08:00
parent 494ae1562d
commit 09359982b1
9 changed files with 1095 additions and 464 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
modules/ocl/src/canny.cpp
View File

@@ -45,6 +45,7 @@
#include <iomanip>
#include "precomp.hpp"
#include "mcwutil.hpp"
using namespace cv;
using namespace cv::ocl;
@@ -237,7 +238,7 @@ void canny::calcSobelRowPass_gpu(const oclMat& src, oclMat& dx_buf, oclMat& dy_b
size_t globalThreads[3] = {cols, rows, 1};
size_t localThreads[3] = {16, 16, 1};
openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
}
void canny::calcMagnitude_gpu(const oclMat& dx_buf, const oclMat& dy_buf, oclMat& dx, oclMat& dy, oclMat& mag, int rows, int cols, bool L2Grad)
@@ -272,7 +273,7 @@ void canny::calcMagnitude_gpu(const oclMat& dx_buf, const oclMat& dy_buf, oclMat
{
strcat(build_options, "-D L2GRAD");
}
openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
}
void canny::calcMagnitude_gpu(const oclMat& dx, const oclMat& dy, oclMat& mag, int rows, int cols, bool L2Grad)
{
@@ -300,7 +301,7 @@ void canny::calcMagnitude_gpu(const oclMat& dx, const oclMat& dy, oclMat& mag, i
{
strcat(build_options, "-D L2GRAD");
}
openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, build_options);
}
void canny::calcMap_gpu(oclMat& dx, oclMat& dy, oclMat& mag, oclMat& map, int rows, int cols, float low_thresh, float high_thresh)
@@ -331,7 +332,7 @@ void canny::calcMap_gpu(oclMat& dx, oclMat& dy, oclMat& mag, oclMat& map, int ro
string kernelName = "calcMap";
size_t localThreads[3] = {16, 16, 1};
openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
}
void canny::edgesHysteresisLocal_gpu(oclMat& map, oclMat& st1, void * counter, int rows, int cols)
@@ -351,7 +352,7 @@ void canny::edgesHysteresisLocal_gpu(oclMat& map, oclMat& st1, void * counter, i
size_t globalThreads[3] = {cols, rows, 1};
size_t localThreads[3] = {16, 16, 1};
openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
}
void canny::edgesHysteresisGlobal_gpu(oclMat& map, oclMat& st1, oclMat& st2, void * counter, int rows, int cols)
@@ -381,7 +382,7 @@ void canny::edgesHysteresisGlobal_gpu(oclMat& map, oclMat& st1, oclMat& st2, voi
args.push_back( make_pair( sizeof(cl_int), (void *)&map.step));
args.push_back( make_pair( sizeof(cl_int), (void *)&map.offset));
openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1, DISABLE);
openCLSafeCall(clEnqueueReadBuffer(Context::getContext()->impl->clCmdQueue, (cl_mem)counter, 1, 0, sizeof(int), &count, 0, NULL, NULL));
std::swap(st1, st2);
}
@@ -406,7 +407,7 @@ void canny::getEdges_gpu(oclMat& map, oclMat& dst, int rows, int cols)
size_t globalThreads[3] = {cols, rows, 1};
size_t localThreads[3] = {16, 16, 1};
openCLExecuteKernel(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
openCLExecuteKernel2(clCxt, &imgproc_canny, kernelName, globalThreads, localThreads, args, -1, -1);
}
#endif // HAVE_OPENCL