Skip to content

rocm: make amdgpu_target sticky#34591

Merged
alalazo merged 3 commits intospack:developfrom
cgmb:sticky-amdgpu
Dec 21, 2022
Merged

rocm: make amdgpu_target sticky#34591
alalazo merged 3 commits intospack:developfrom
cgmb:sticky-amdgpu

Conversation

@cgmb
Copy link
Copy Markdown
Contributor

@cgmb cgmb commented Dec 17, 2022

The sticky property will prevent clingo from changing the amdgpu_target to work around conflicts. This is the same behaviour as was adopted for cuda_arch in 055c9d1.

The sticky property will prevent clingo from changing the amdgpu_target
to work around conflicts. This is the same behaviour as was adopted for
cuda_arch in 055c9d1.
@cgmb
Copy link
Copy Markdown
Contributor Author

cgmb commented Dec 17, 2022

This should prevent some of the weird choices of amdgpu_target variant values we see on the CI.

adamjstewart
adamjstewart previously approved these changes Dec 17, 2022
Copy link
Copy Markdown
Member

@adamjstewart adamjstewart left a comment

Choose a reason for hiding this comment

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

Why do so many packages define this variant? Can they subclass ROCmPackage or would that create circular dependencies?

@cgmb
Copy link
Copy Markdown
Contributor Author

cgmb commented Dec 17, 2022

Why do so many packages define this variant? Can they subclass ROCmPackage or would that create circular dependencies?

I was thinking the same thing, but I'm not sure how to do that nicely. The defaults for ROCmPackage don't make sense for libraries that are part of ROCm. Defaulting to ~rocm amdgpu_target=none seems more designed for something where GPU support is an optional add-on.

@adamjstewart
Copy link
Copy Markdown
Member

You could extend the base class but add conflicts("~rocm")

@cgmb
Copy link
Copy Markdown
Contributor Author

cgmb commented Dec 17, 2022

That sounds like a good idea. That might be a good next step to explore after #34586.

@alalazo alalazo merged commit e60e746 into spack:develop Dec 21, 2022
@cgmb cgmb deleted the sticky-amdgpu branch December 21, 2022 23:58
stephenmsachs pushed a commit to stephenmsachs/spack that referenced this pull request Jan 3, 2023
The sticky property will prevent clingo from changing the amdgpu_target
to work around conflicts. This is the same behaviour as was adopted for
cuda_arch in 055c9d1.
@haampie
Copy link
Copy Markdown
Member

haampie commented Jan 4, 2023

Confusingly CI passed here, even though it failed on develop.

RikkiButler20 pushed a commit to RikkiButler20/spack that referenced this pull request Jan 24, 2023
The sticky property will prevent clingo from changing the amdgpu_target
to work around conflicts. This is the same behaviour as was adopted for
cuda_arch in 055c9d1.
amd-toolchain-support pushed a commit to amd-toolchain-support/spack that referenced this pull request Feb 16, 2023
The sticky property will prevent clingo from changing the amdgpu_target
to work around conflicts. This is the same behaviour as was adopted for
cuda_arch in 055c9d1.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants