]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
CUDA: remove -sm row, refactor cuBLAS (#24216)
authorJohannes Gäßler <redacted>
Mon, 6 Jul 2026 18:04:53 +0000 (20:04 +0200)
committerGitHub <redacted>
Mon, 6 Jul 2026 18:04:53 +0000 (20:04 +0200)
commit74976e1aefd8d64845ae5510674b36079c855c0c
treeb4416ad52f96da216157ad73d2baeda4b858d55d
parent9abce7473ad9e6870080caf08e2dcf8592e086b1
CUDA: remove -sm row, refactor cuBLAS (#24216)

* CUDA: remove -sm row, refactor cuBLAS

* fix CDNA + BF16 logic

* fix bad return

* fix src0 strides, contiguous requirements

* fix GGML_CUDA_FORCE_CUBLAS

* fix casts to BF16
docs/build.md
ggml/include/ggml-cuda.h
ggml/src/ggml-cuda/convert.cu
ggml/src/ggml-cuda/ggml-cuda.cu
ggml/src/ggml-cuda/mmvq.cu
src/llama-model.cpp