Skip to content
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

replace use of std math in kernels with quda versions #1524

Merged
merged 3 commits into from
Dec 20, 2024
Merged

replace use of std math in kernels with quda versions #1524

merged 3 commits into from
Dec 20, 2024

Conversation

jcosborn
Copy link
Contributor

replace use of std math in kernels with quda versions
allow building without cufft
change some casting

@jcosborn
replace use of std math in kernels with quda versions
40df91b 
allow building without cufft
change some casting
@jcosborn jcosborn requested review from a team as code owners December 11, 2024 16:16
maddyscientist
maddyscientist reviewed Dec 12, 2024
View reviewed changes
Copy link
Member

@maddyscientist maddyscientist left a comment

Choose a reason for hiding this comment

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

Looks good James. I've left a few questions.

Perhaps I forget the context, but what is the goal to allow building without cuFFT on CUDA? Does the same need to be enabled for the HIP target?

include/targets/hip/shared_memory_helper.h Outdated Show resolved Hide resolved
include/targets/cuda/shared_memory_helper.h Outdated Show resolved Hide resolved
include/targets/generic/mrg32k3a.h Show resolved Hide resolved
include/targets/cuda/math_helper.cuh Outdated Show resolved Hide resolved
@jcosborn
Copy link
Contributor Author

Perhaps I forget the context, but what is the goal to allow building without cuFFT on CUDA? Does the same need to be enabled for the HIP target?

This was for testing with LLVM on Polaris. By default, it didn't find the libraries, and it seemed easier to just disable them than try to point it in the right direction. Disabling worked fine for blas, but wasn't fully implemented for fft.
I don't think there is a similar need for HIP at the moment. It could be attempted now for completeness, or just wait until there is a test case.

maddyscientist
maddyscientist approved these changes Dec 13, 2024
View reviewed changes
Copy link
Member

@maddyscientist maddyscientist left a comment

Choose a reason for hiding this comment

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

Approved, but one final question: I see you've restore the constexpr but you still have the __device__ __host__ qualifiers on the shared memory constructor. Were there needed for some compiler?

@jcosborn
Copy link
Contributor Author

I think you mean in the sharedMem accessor? I just reverted that too.
f9a551a

@maddyscientist
Copy link
Member

cscs-ci run

@maddyscientist maddyscientist merged commit b58f1ec into develop Dec 20, 2024
7 checks passed
@maddyscientist maddyscientist deleted the feature/sycl-merge branch December 20, 2024 20:36
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@maddyscientist maddyscientist maddyscientist approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@jcosborn @maddyscientist