-
-
Notifications
You must be signed in to change notification settings - Fork 55.6k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CL_INVALID_WORK_GROUP_SIZE
on calling OpenCL kernels minmaxloc
, reduce
and some others
#11797
Comments
CL_INVALID_WORK_GROUP_SIZE
on calling OpenCL kernels minmaxloc
and reduce
CL_INVALID_WORK_GROUP_SIZE
on calling OpenCL kernels minmaxloc
, reduce
and some others
because of this device setting:
Tests don't fail, because OpenCV fallback to CPU code if OpenCL runtime reports some error (or OpenCL is not available). Perhaps OpenCV should query kernel-specific limits (but there is recursion because these kernels are built for specific group size) or limit max group size globally as a workaround. Another question is efficiency / performance of such large groups (it is platform specific, so there is no straightforward way) - we don't run OpenCL kernels on devices with groups more that 256 items. Could you try to override |
Now running OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('fft_multi_radix_rows', dims=2, globalsize=160x720x1, localsize=160x1x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('fft_multi_radix_rows', dims=2, globalsize=240x1080x1, localsize=240x1x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('fft_multi_radix_rows', dims=2, globalsize=256x2048x1, localsize=256x1x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('gemm', dims=2, globalsize=160x640x1, localsize=16x16x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('gemm', dims=2, globalsize=320x1280x1, localsize=16x16x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('ifft_multi_radix_cols', dims=2, globalsize=1025x256x1, localsize=1x256x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('ifft_multi_radix_cols', dims=2, globalsize=961x135x1, localsize=1x135x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('ifft_multi_radix_rows', dims=2, globalsize=160x720x1, localsize=160x1x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('ifft_multi_radix_rows', dims=2, globalsize=240x1080x1, localsize=240x1x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('ifft_multi_radix_rows', dims=2, globalsize=256x2048x1, localsize=256x1x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('meanStdDev', dims=1, globalsize=1024x1x1, localsize=256x1x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('reduce', dims=1, globalsize=1024x1x1, localsize=256x1x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('reduce_horz_opt', dims=2, globalsize=32x1080x1, localsize=32x8x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('reduce_horz_opt', dims=2, globalsize=32x2160x1, localsize=32x8x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('reduce_horz_opt', dims=2, globalsize=32x480x1, localsize=32x8x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('reduce_horz_opt', dims=2, globalsize=32x720x1, localsize=32x8x1) sync=false
OpenCL error CL_INVALID_WORK_GROUP_SIZE (-54) during call: clEnqueueNDRangeKernel('stage1_with_sobel', dims=2, globalsize=1920x1080x1, localsize=32x8x1) sync=false |
Hi, Have you managed to solve that problem? |
Hi Szymon. Unfortunately, not. However, issue #13414 seems to be related and is more active. |
Try to override group size in OpenCV through environment variable |
We are having the same issue with i.mx8mp SoC's. Is there any update on this issue ? Thanks in advance. |
System information (version)
Description
When running
opencv_perf_core
, some tests output one of the following lines up to a hundred times:The test result, however, is successful.
Are these errors expected?
Detailed description
A complete output:
Device info:
Cmake output:
The text was updated successfully, but these errors were encountered: