Skip to content

Ci rocm stack#31450

Closed
kwryankrattiger wants to merge 2 commits intospack:developfrom
kwryankrattiger:ci_rocm_stack
Closed

Ci rocm stack#31450
kwryankrattiger wants to merge 2 commits intospack:developfrom
kwryankrattiger:ci_rocm_stack

Conversation

@kwryankrattiger
Copy link
Copy Markdown
Contributor

Add a pipeline that tests building the ROCm stack. There have been increasingly frequent issues keeping this stack stable as more packages depend on it, so it seems now is a good time to add this pipeline.

@srekolam @chuckatkins

@spackbot-app spackbot-app bot added dependencies gitlab Issues related to gitlab integration update-package labels Jul 5, 2022
srekolam
srekolam previously approved these changes Jul 5, 2022
Copy link
Copy Markdown
Contributor

@cgmb cgmb left a comment

Choose a reason for hiding this comment

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

You can get a pretty big reduction in your build time if you specify an amdgpu_target variant for the mathlibs, particularly rocblas, rocfft, rocsparse and rocsolver. The build time cost of each architecture is basically linear and the default amdgpu_target=auto builds for ~7 architectures. Though, perhaps setting it for rocrand has forced the rest to resolve to amdgpu_target=gfx1030 as well?

@kwryankrattiger
Copy link
Copy Markdown
Contributor Author

@cgmb I wasn't sure how to set the amdgpu_target since it is inconsistent across packages, but it looks like #30582 fixes some of that. There was another PR by @chuckatkins that does something similar to split out the features as well #29018.

I think now that the #30582 is merged I can put that in more easily.

@kwryankrattiger kwryankrattiger force-pushed the ci_rocm_stack branch 2 times, most recently from 61c806d to cc414e9 Compare July 7, 2022 16:02
@kwryankrattiger kwryankrattiger marked this pull request as ready for review July 7, 2022 19:26
@kwryankrattiger
Copy link
Copy Markdown
Contributor Author

@scottwittenburg or @srekolam if you guys could give this an approval to get it merged, I think it is ready now!

@spackbot-app spackbot-app bot added the core PR affects Spack core functionality label Jul 26, 2022
@kwryankrattiger kwryankrattiger force-pushed the ci_rocm_stack branch 2 times, most recently from a2f4f3a to 746f568 Compare August 2, 2022 02:34
srekolam
srekolam previously approved these changes Aug 18, 2022
@kwryankrattiger
Copy link
Copy Markdown
Contributor Author

@srekolam I just rebased, hopefully that fixes the coverage CI.

@cgmb
Copy link
Copy Markdown
Contributor

cgmb commented Oct 1, 2022

Is this change waiting on something? A CI run for ROCm would be very useful.

@kwryankrattiger
Copy link
Copy Markdown
Contributor Author

Just waiting on a clean CI. My current plan is to disable the failing packages, but I think there may just need to point at the correct amdgpu_targets.

@cgmb
Copy link
Copy Markdown
Contributor

cgmb commented Oct 5, 2022

The failures are because the build is passing -DAMDGPU_TARGETS:STRING=gfx1030;gfx701 but gfx701 is not a supported architecture. Where is that value coming from? The target list only specifies amdgpu_target=gfx1030.

@cgmb cgmb mentioned this pull request Dec 6, 2022
@kwryankrattiger
Copy link
Copy Markdown
Contributor Author

Closing this for now.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

core PR affects Spack core functionality dependencies gitlab Issues related to gitlab integration update-package

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants