[ROCm] CK-based GEMM (#131004)

- composable_kernel as a third_party submodule
- "ck" as a `torch.backends.cuda.preferred_linalg_library()`
- reference CK gemm implementations for float, bfloat16, and half types

Pull Request resolved: https://github.com/pytorch/pytorch/pull/131004
Approved by: https://github.com/xw285cornell, https://github.com/pruthvistony

Co-authored-by: Andres Lugo <Andy.LugoReyes@amd.com>
Co-authored-by: Pruthvi Madugundu <pruthvigithub@gmail.com>
This commit is contained in:
Jeff Daily
2024-10-20 02:57:41 +00:00
committed by PyTorch MergeBot
parent 0a2407b93c
commit 3f3b692a00
20 changed files with 1645 additions and 32 deletions

4
.gitmodules vendored
View File

@ -127,3 +127,7 @@
[submodule "third_party/NVTX"]
path = third_party/NVTX
url = https://github.com/NVIDIA/NVTX.git
[submodule "third_party/composable_kernel"]
path = third_party/composable_kernel
url = https://github.com/ROCm/composable_kernel.git
branch = develop