8000 FP32 output GEMM from FP16/BF16 inputs · pytorch/pytorch@50dd9cd · GitHub

Commit 50dd9cd

committed

FP32 output GEMM from FP16/BF16 inputs

ghstack-source-id: 375228f Pull Request resolved: #150812

1 parent b6929ae commit 50dd9cdCopy full SHA for 50dd9cd

6 files changed

+722

-122

lines changed

aten/src/ATen
- cuda
  - CUDABlas.cpp
  - CUDABlas.h
  - tunable
    - GemmCommon.h
- native
  - cuda
    - Blas.cpp
  - native_functions.yaml
test/cpp/tensorexpr
- test_external_calls.cpp

6 files changed

+722

-122

lines changed

Comments

(0)

0