leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

Updates for 3.0 (#857)

Browse Source 
Co-authored-by: Aniket Shivam <ashivam@nvidia.com>
This commit is contained in:
ANIKET SHIVAM
2023-03-09 15:27:40 -05:00
committed by GitHub
co-authored by Aniket Shivam
parent a68e2f95f0
commit c4f6b8c6bc
7 changed files with 131 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+3 -1
View File
@@ -311,7 +311,9 @@ $ make cutlass_profiler -j16
By default, only one tile size is instantiated for each data type, math instruction, and layout.
To instantiate all, set the following environment variable when running CMake from an empty `build/` directory.
Beware, this results in *thousands* of kernels and long build times.
Beware, this results in *tens of thousands* of kernels and long build times.
This would also result in a large binary size and on some platforms linker to fail on building the library.
Therefore, it's highly recommended to generate only a subset of kernels as demonstrated in the sub-section below.
```bash
$ cmake .. -DCUTLASS_NVCC_ARCHS=90a -DCUTLASS_LIBRARY_KERNELS=all
...