Skip to content

Conversation

@vinx13
Copy link
Member

@vinx13 vinx13 commented Jan 2, 2024

CI images should also be updated to install cmake 3.24

cc @tqchen @junrushao @Lunderberg

@Lunderberg
Copy link
Contributor

I like the change, and this makes sense to me.

@Lunderberg
Copy link
Contributor

Since #16302 passed CI, and is now merged in as a temporary fix, can this PR also revert the changes from it?

@vinx13 vinx13 force-pushed the unity-cmake-native-cuda branch from 621fc09 to 47a3c9d Compare January 2, 2024 21:44
@vinx13 vinx13 marked this pull request as ready for review January 9, 2024 16:55
Copy link
Contributor

@Lunderberg Lunderberg left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thank you for the changes, and for handling the changes between TVM and the cutlass kernels!

@vinx13 vinx13 force-pushed the unity-cmake-native-cuda branch from 113f38d to 1d600da Compare January 10, 2024 21:14
@junrushao junrushao merged commit 474c06b into apache:unity Jan 10, 2024
@jwfromm
Copy link
Contributor

jwfromm commented Jan 11, 2024

Late to the party but just wanted to say i love this change. Thanks Wuwei!

@Lunderberg
Copy link
Contributor

It looks like cmake's GPU detection isn't perfect (tested with cmake 3.26.4 and 3.27.7), and this can cause the following error to occur when trying to determine the cuda architectures to use.

CMake Error in CMakeLists.txt:                                    
  CUDA_ARCHITECTURES is set to "native", but no GPU was detected. 

I don't have a good long-term solution for this issue, but a current workaround is to explicitly call set(CMAKE_CUDA_ARCHITECTURES 75) in your config.cmake.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants