• Chuanbo Weng's avatar
    Remove unnecessary kercn limitation of 4. · 2d8c89c4
    Chuanbo Weng authored
    When accessing global memory by DWORD4, memory bandwidth
    can be fully utilized on Intel platform. This patch will
    make more image format(e.g. 8UC4) be processed in DWORD4
    by work-item. After applying this patch, 3 subcase of
    ./opencv_perf_core --gtest_filter=OCL_RepeatFixture_Repeat.Repeat/*
    can be speedup on HD4000 graphics card with Beignet:
    OCL_RepeatFixture_Repeat.Repeat/2, 64% improvement.
    OCL_RepeatFixture_Repeat.Repeat/6, 50% improvement.
    OCL_RepeatFixture_Repeat.Repeat/8, 56% improvement.
    Signed-off-by: 's avatarChuanbo Weng <chuanbo.weng@intel.com>
    2d8c89c4
Name
Last commit
Last update
..
androidcamera Loading commit data...
calib3d Loading commit data...
core Loading commit data...
cuda Loading commit data...
cudaarithm Loading commit data...
cudabgsegm Loading commit data...
cudacodec Loading commit data...
cudafeatures2d Loading commit data...
cudafilters Loading commit data...
cudaimgproc Loading commit data...
cudalegacy Loading commit data...
cudaoptflow Loading commit data...
cudastereo Loading commit data...
cudawarping Loading commit data...
cudev Loading commit data...
features2d Loading commit data...
flann Loading commit data...
highgui Loading commit data...
imgcodecs Loading commit data...
imgproc Loading commit data...
java Loading commit data...
ml Loading commit data...
objdetect Loading commit data...
photo Loading commit data...
python Loading commit data...
shape Loading commit data...
stitching Loading commit data...
superres Loading commit data...
ts Loading commit data...
video Loading commit data...
videoio Loading commit data...
videostab Loading commit data...
viz Loading commit data...
world Loading commit data...
CMakeLists.txt Loading commit data...
refman.rst.in Loading commit data...