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

Merge pull request #587 from bitwangyaoyao:2.4_fixHog

Browse Source
This commit is contained in:
Andrey Kamaev 2013-02-28 18:06:27 +04:00 committed by OpenCV Buildbot
commit fba0e6fc5c
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
modules/ocl/src/kernels/objdetect_hog.cl
View File

@ -140,6 +140,10 @@ float reduce_smem(volatile __local float* smem, int size)
if (tid < 32)
{
if (size >= 64) smem[tid] = sum = sum + smem[tid + 32];
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 16)
{
if (size >= 32) smem[tid] = sum = sum + smem[tid + 16];
if (size >= 16) smem[tid] = sum = sum + smem[tid + 8];
if (size >= 8) smem[tid] = sum = sum + smem[tid + 4];
@ -224,6 +228,11 @@ __kernel void classify_hists_kernel(const int cblock_hist_size, const int cdescr
{
volatile __local float* smem = products;
smem[tid] = product = product + smem[tid + 32];
}
barrier(CLK_LOCAL_MEM_FENCE);
if (tid < 16)
{
volatile __local float* smem = products;
smem[tid] = product = product + smem[tid + 16];
smem[tid] = product = product + smem[tid + 8];
smem[tid] = product = product + smem[tid + 4];