Skip to content

Add cross-compilation option for all the supported architecture - #235

Open
Caenorst wants to merge 1 commit into
NVlabs:mainfrom
Caenorst:cfujitsang/setup_cross_compilation
Open

Add cross-compilation option for all the supported architecture#235
Caenorst wants to merge 1 commit into
NVlabs:mainfrom
Caenorst:cfujitsang/setup_cross_compilation

Conversation

@Caenorst

@Caenorst Caenorst commented Mar 2, 2026

Copy link
Copy Markdown

This is adding a NVDIFFRAST_CROSS_COMPILE_ALL environment variable which will automatically cross-compile nvdiffrast across all the supported architecture (from the cuda used by torch).

This is not modifying the default behavior which requires TORCH_CUDA_ARCH_LIST but if the user wants to lazily cross-compile then it make things easier.

This is especially useful for CI, or making wheels to be shared.

… assignment in the dockerfile

Signed-off-by: Clement Fuji Tsang <cfujitsang@nvidia.com>
@Caenorst
Caenorst force-pushed the cfujitsang/setup_cross_compilation branch from 665a026 to 87f26c4 Compare March 3, 2026 04:21
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.

1 participant