ci: bundle all RDNA3/3.5/4 archs#49
Open
Geramy wants to merge 1 commit into
Open
Conversation
CI release built gfx1151-only, so other AMD GPUs had no native device code. Set HIP_BUILD_ARCHS to the full RDNA3 (gfx1100-1103), RDNA3.5 (gfx1150-1152), and RDNA4 (gfx1200-1201) set for CMAKE_HIP_ARCHITECTURES, so one release binary bundles them all (HIP selects the matching ISA at runtime). ROCM_ARCH stays gfx1151 for the runner's package install; the rest are cross-compiled by the base toolchain.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Bundle all RDNA3 / RDNA3.5 / RDNA4 archs in the CI release
The CI release was building gfx1151-only (
ROCM_ARCH), so other AMD GPUs (e.g. gfx1152 Radeon 860M, gfx1201 R9700) shipped with no native device code and fell back to broken/garbage paths.This sets
HIP_BUILD_ARCHSforCMAKE_HIP_ARCHITECTURESto the full set:One release binary bundles them all (HIP fatbin auto-selects the matching ISA at runtime).
ROCM_ARCHstaysgfx1151for the runner's ROCm package install; the other archs are cross-compiled by the base toolchain.Pairs with the matching NripeshN/mlx
rocm-supportchanges (native-WMMA allowlist incl. gfx1152, WMMA prefill default-on, legacy graph-build path disabled).