Add AMDGCN option similar to cuda-compute-capabilities
#4860
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Thyre
commented
Apr 25, 2025
Thyre
force-pushed
the
support-passing-amdgcn
branch
from
958ad0a to
bff1bfb
Compare
Thyre
commented
Apr 25, 2025
Signed-off-by: Jan Andre Reuter <jan@zyten.de>
Thyre
force-pushed
the
support-passing-amdgcn
branch
from
bff1bfb to
0e7aaf3
Compare
Thyre
changed the title
cuda-compute-capabilitiescuda-compute-capabilities
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
This PR aims to implement a similar option to
cuda-compute-capabilities(and related options) for AMD GPUs.The option can then replace the manual handling done in some EasyBlocks, e.g. Clang & LLVM, allowing to enable (some) GPU builds without the need to alter the EasyConfig.
Most of the handling was copied from CUDA, while some options were skipped as they don't make much sense, e.g.
cuda_cc_space_sep_no_period.The used regex should support all GPU architectures starting from
gfx600, including the more recent generic targets.Actual compiler support then needs to be present in the compiler consuming these architectures. Both GCC and LLVM accept the same naming, i.e.
gfx[...], including generic targets.Missing features compared to CUDA
cuda_cache_diroption is missing. I haven't found something similar for HIP yet, but may simply have missed itgfx90agfx?More to be determined.
Known issues
gfx10--genericto pass, even though it is not allowed.Resolves #4829