llama.cpp
ROCm/HIP: fixed BLAS workspace instead of rocBLAS allocating one per call
#31838

Merged

ROCm/HIP: fixed BLAS workspace instead of rocBLAS allocating one per call #31838

JohannesGaessler merged 2 commits into ggml-org:master from fox5611:rocm-cublas-workspace

ROCm/HIP: fixed BLAS workspace instead of rocBLAS allocating one per …

d1458d99

fox5611 requested a review from

IMbackK 5 days ago

fox5611 requested a review 5 days ago

github-actions added ggml

github-actions added CUDA

IMbackK approved these changes on 2026-11-07

JohannesGaessler approved these changes on 2026-11-10

ROCm/HIP: corrected by only setting the workspace on ROCm 7.0 or newer

ce72c056

JohannesGaessler approved these changes on 2026-11-11

JohannesGaessler merged 43243363 into master 1 day ago

Reviewers

JohannesGaessler

IMbackK

Assignees

No one assigned

Labels

ggml CUDA

Milestone

No milestone