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

align grid by 4

Browse Source
This commit is contained in:
marina.kolpakova 2012-11-24 01:55:03 +04:00
parent 8579666b03
commit be0c20b758
2 changed files with 11 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
modules/gpu/src/cuda/integral_image.cu
View File

@ -357,18 +357,25 @@ namespace cv { namespace gpu { namespace device
#endif #endif
} }
void shfl_integral_gpu(PtrStepSzb img, PtrStepSz<unsigned int> integral, cudaStream_t stream) void shfl_integral_gpu(const PtrStepSzb& img, PtrStepSz<unsigned int> integral, cudaStream_t stream)
{ {
{ {
// each thread handles 16 values, use 1 block/row // each thread handles 16 values, use 1 block/row
const int block = img.cols / 16; int block = img.cols / 16;
// save, becouse step is actually can't be less 512 bytes
int align = img.cols % 4;
if ( align != 0)
{
block += (4 - align);
}
// launch 1 block / row // launch 1 block / row
const int grid = img.rows; const int grid = img.rows;
cudaSafeCall( cudaFuncSetCacheConfig(shfl_integral_horizontal, cudaFuncCachePreferL1) ); cudaSafeCall( cudaFuncSetCacheConfig(shfl_integral_horizontal, cudaFuncCachePreferL1) );
shfl_integral_horizontal<<<grid, block, 0, stream>>>((PtrStepSz<uint4>) img, (PtrStepSz<uint4>) integral); shfl_integral_horizontal<<<grid, block, 0, stream>>>((const PtrStepSz<uint4>) img, (PtrStepSz<uint4>) integral);
cudaSafeCall( cudaGetLastError() ); cudaSafeCall( cudaGetLastError() );
} }

2
modules/gpu/src/imgproc.cpp
View File

@ -537,7 +537,7 @@ namespace cv { namespace gpu { namespace device
{ {
namespace imgproc namespace imgproc
{ {
void shfl_integral_gpu(PtrStepSzb img, PtrStepSz<unsigned int> integral, cudaStream_t stream); void shfl_integral_gpu(const PtrStepSzb& img, PtrStepSz<unsigned int> integral, cudaStream_t stream);
} }
}}} }}}