Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[FRONTEND] Allow tl.{u}int{width} annotations to bypass opportunistic value-based JIT-specialization #3102

Merged
merged 2 commits into from
Feb 10, 2024

Conversation

ptillet
Copy link
Collaborator

@ptillet ptillet commented Feb 10, 2024

Triton's JIT currently specialize all integer arguments to int32 if they "fit", and int64 otherwise. This PR gives users the possibility to opt-out from this kind of specialization to avoid compiling too many kernels.

@ptillet ptillet force-pushed the phil/annotations-improvements branch from 73d2650 to f94c4ca Compare February 10, 2024 07:35
@ptillet ptillet marked this pull request as ready for review February 10, 2024 07:45
@ptillet ptillet merged commit b6e24b6 into main Feb 10, 2024
4 checks passed
@ptillet ptillet deleted the phil/annotations-improvements branch February 10, 2024 07:52
srush pushed a commit to srush/triton-edit that referenced this pull request Feb 22, 2024
…ic value-based JIT-specialization (triton-lang#3102)

Triton's JIT currently specialize all integer arguments to `int32` if
they "fit", and `int64` otherwise. This PR gives users the possibility
to opt-out from this kind of specialization to avoid compiling too many
kernels.
binarman pushed a commit to binarman/triton that referenced this pull request Apr 2, 2024
…ic value-based JIT-specialization (triton-lang#3102)

Triton's JIT currently specialize all integer arguments to `int32` if
they "fit", and `int64` otherwise. This PR gives users the possibility
to opt-out from this kind of specialization to avoid compiling too many
kernels.
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