Skip to content

Conversation

@arun-thmn
Copy link
Contributor

@arun-thmn arun-thmn commented Sep 16, 2025

Bump are mostly related to below LLVM patches:
llvm/llvm-project#151150 - InitAll replaced by static library. So, added necessary header to our pipeline.
llvm/llvm-project#147961 - Remove or update lit tests related to matmul_transpose

For CI:

  • Removed v100 and clx partition + removed benchmarking related to CLX base and omp runs,
  • Updated the gcc src to system src + cuda src to 12.9,

@arun-thmn arun-thmn marked this pull request as ready for review September 16, 2025 14:00
@arun-thmn arun-thmn added the benchmark-full Benchmark all targets label Sep 16, 2025
@adam-smnk
Copy link
Contributor

Almost fixed, now the problem is that CUDA 13 doesn't support architectures older than sm_75 while MLIR serializes for sm_50 by default.

Instead of the latest, we could try sourcing CUDA 12.9 that still supports older architectures and should be compatible with gcc 14.

@arun-thmn arun-thmn changed the title Bump LLVM [CI] Bump LLVM + CI partition changes Sep 17, 2025
@arun-thmn
Copy link
Contributor Author

@adam-smnk Should be fine now. Please verify.

@arun-thmn arun-thmn merged commit 380c056 into libxsmm:main Sep 17, 2025
19 of 26 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

benchmark-full Benchmark all targets

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants