Skip to content

kokkos: enable +cuda_lambda by default for 4.0.00+#38565

Closed
brian-kelley wants to merge 3 commits intospack:developfrom
brian-kelley:cudaLambdaDefault
Closed

kokkos: enable +cuda_lambda by default for 4.0.00+#38565
brian-kelley wants to merge 3 commits intospack:developfrom
brian-kelley:cudaLambdaDefault

Conversation

@brian-kelley
Copy link
Copy Markdown
Contributor

Kokkos_ENABLE_CUDA_LAMBDA (the CMake option) has been ON by default since version 4.0.00, so this makes the corresponding variant cuda_lambda have the same default behavior.

Tested by installing the following recipes:

# cuda_lambda correctly disabled:
[email protected] +cuda +wrapper cuda_arch=61

# cuda_lambda correctly enabled:
[email protected] +cuda +wrapper +cuda_lambda cuda_arch=61
[email protected] +cuda +wrapper cuda_arch=61
kokkos@master +cuda +wrapper cuda_arch=61
kokkos@develop +cuda +wrapper cuda_arch=61

@brian-kelley brian-kelley marked this pull request as draft June 28, 2023 00:03
@brian-kelley
Copy link
Copy Markdown
Contributor Author

brian-kelley commented Jun 28, 2023

@eugeneswalker It turns out that this does not actually work the way I thought - it's not currently possible to have version-dependent defaults for variants. This feature was also requested by someone else in #38302 though, and it's actively being worked on.

So this is just going to be a draft PR until that feature reaches the develop branch, and then this can be merged.

Copy link
Copy Markdown
Contributor

@masterleinad masterleinad left a comment

Choose a reason for hiding this comment

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

@brian-kelley Can you try again after #40326 and #40998 have been merged?

@brian-kelley
Copy link
Copy Markdown
Contributor Author

@masterleinad Yes, the default behavior works now (lambda off by default pre-4.0, but on for later versions).

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

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants