youngkingdom/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Make cutlass::gemm::device::GemmArray usable (#295)

Browse Source 
* Fix the build of cutlass/gemm/device/gemm_array.h and add a demo for GemmArray

* Add a reference to GemmArray to the docs

Co-authored-by: Ivan Komarov <dfyz@yandex-team.ru>
This commit is contained in:
Ivan Komarov
2022-02-18 04:01:05 +03:00
committed by GitHub
parent 3cfa5db2a2
commit e96f00586c
3 changed files with 127 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
media/docs/gemm_api.md
View File

@ -81,6 +81,7 @@ has semantics similar to cuBLAS.
The device-wide GEMM API is embodied by the following operators:
- [cutlass::gemm::device::Gemm](/include/cutlass/gemm/device/gemm.h) - basic GEMM operation
- [cutlass::gemm::device::GemmArray](/include/cutlass/gemm/device/gemm_array.h) - batched GEMM operation in which input matrices are read from arrays of pointers
- [cutlass::gemm::device::GemmBatched](/include/cutlass/gemm/device/gemm_batched.h) - batched GEMM operation in which input matrices are separated by a constant stride
- [cutlass::gemm::device::GemmSplitKParallel](/include/cutlass/gemm/device/gemm_splitk_parallel.h) - GEMM operation that partitions the GEMM K dimension then launches a separate reduction kernel