Skip to content

__shfl_up_sync with mask for CUDA >= 9 #13658

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.

Sign up for GitHub

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

Jump to bottom
Merged
merged 1 commit into from
Jan 21, 2019
Merged

__shfl_up_sync with mask for CUDA >= 9 #13658

merged 1 commit into from
Jan 21, 2019

Conversation

nglee
Copy link
Contributor

@nglee nglee commented Jan 19, 2019 •
edited by alalek
Loading

For warp shuffle functions introduced from CUDA 9, giving full mask may cause the program to hang on some devices. This PR tries to fix this for __shfl_up_sync function. Without this PR, following tests hang with RTX 2080 Ti.

./bin/opencv_test_cudev --gtest_filter=*Integral*
./bin/opencv_test_cudaarithm --gtest_filter=*Integral*
./bin/opencv_test_cudaimgproc --gtest_filter=*CLAHE*

Resolves #13014

force_builders=Custom
docker_image:Custom=ubuntu-cuda:16.04
buildworker:Custom=linux-1

@nglee nglee force-pushed the dev_CudaShflUpCompat branch 2 times, most recently from e483c48 to 5cdf377 Compare January 21, 2019 01:27
@nglee
Copy link
Contributor Author

nglee commented Jan 21, 2019 •
edited
Loading

Following tests were performed:

opencv_test_cudev --gtest_filter=*BlockScan*
opencv_test_cudev --gtest_filter=*Integral*
opencv_test_cudaarithm --gtest_filter=*Integral*
opencv_test_cudaimgproc --gtest_filter=*CLAHE*

Verified on following test environments:

CUDA 10.0 on 2080 Ti
CUDA 10.0 on 1080
CUDA 8.0 on 1080

I'll squash commits into one.

@nglee
__shfl_up_sync with mask for CUDA >= 9
970293a 
* __shfl_up_sync with proper mask value for CUDA >= 9

* BlockScanInclusive for CUDA >= 9

* compatible_shfl_up for use in integral.hpp

* Use CLAHE in cudev

* Add tests for BlockScan
@nglee nglee force-pushed the dev_CudaShflUpCompat branch from 46374a1 to 970293a Compare January 21, 2019 15:36
alalek
alalek approved these changes Jan 21, 2019
View reviewed changes
Copy link
Member

@alalek alalek left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Well done! Thank you 👍

I will merge these changes into master branch in few days.

@opencv-pushbot opencv-pushbot merged commit 970293a into opencv:3.4 Jan 21, 2019
opencv-pushbot pushed a commit that referenced this pull request Jan 21, 2019
@alalek
Merge pull request #13658 from nglee:dev_CudaShflUpCompat
dcdbaef
@nglee nglee deleted the dev_CudaShflUpCompat branch January 22, 2019 01:02
@nglee nglee mentioned this pull request Jan 22, 2019
Closed
@alalek alalek mentioned this pull request Jan 22, 2019
Merged
@nglee nglee mentioned this pull request Feb 1, 2019
Closed
This was referenced Feb 12, 2019
Closed
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@alalek alalek alalek approved these changes

Assignees

@alalek alalek

Labels
bug category: gpu/cuda (contrib) OpenCV 4.0+: moved to opencv_contrib feature
Projects
None yet
Milestone
3.4.6
Development

Successfully merging this pull request may close these issues.
3 participants
@nglee @alalek @opencv-pushbot