Add wrappers for the CUDA graph API. #877
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #65.
I'm pretty happy with how the API turned out. See the tests for two examples,
capture
/instantiate
/update
/launch
for full control,@captured
wrapping all that. I didn't manage to outperform regular kernel launches though, but that's probably due to my limited benchmarking of the APIs.Note that I started looking into this to implement a good
mapslices
implementation, but then I realized the function passed tomapslices
allocates its output on every iteration, so we can't use the graph API for it.