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
This commit is contained in:
Johannes Gäßler
2026-07-06 20:04:53 +02:00
committed by GitHub
parent 9abce7473a
commit 74976e1aef
6 changed files with 275 additions and 1166 deletions
+2
View File
@@ -953,6 +953,8 @@ static buft_list_t make_gpu_buft_list(ggml_backend_dev_t dev, llama_split_mode s
if (buft != nullptr) {
buft_list.emplace_back(dev, buft);
}
} else {
throw std::runtime_error(format("device %s does not support split buffers", ggml_backend_dev_name(dev)));
}
}