This repository was archived by the owner on Oct 7, 2025. It is now read-only.
[RFC] OpenXLA CPU Strategy #96
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
mattwalsh
approved these changes
Jan 18, 2024
copybara-service bot
pushed a commit
to jax-ml/jax
that referenced
this pull request
Jul 26, 2024
XLA:CPU is preparing to switch from compiling whole XLA program into a single LLVM function to a mode where each fusion/kernel will have its own entry point, and a thin runtime that will dispatch compute functions concurrently. This execution mode does not work very well with while loops with tiny computations and large number of iterations. Similar to GPU backend use vmap to avoid excessive runtime overheads. Context: openxla/community#96 PiperOrigin-RevId: 656119575
copybara-service bot
pushed a commit
to jax-ml/jax
that referenced
this pull request
Jul 26, 2024
XLA:CPU is preparing to switch from compiling whole XLA program into a single LLVM function to a mode where each fusion/kernel will have its own entry point, and a thin runtime that will dispatch compute functions concurrently. This execution mode does not work very well with while loops with tiny computations and large number of iterations. Similar to GPU backend use vmap to avoid excessive runtime overheads. Context: openxla/community#96 PiperOrigin-RevId: 656119575
copybara-service bot
pushed a commit
to jax-ml/jax
that referenced
this pull request
Jul 26, 2024
XLA:CPU is preparing to switch from compiling whole XLA program into a single LLVM function to a mode where each fusion/kernel will have its own entry point, and a thin runtime that will dispatch compute functions concurrently. This execution mode does not work very well with while loops with tiny computations and large number of iterations. Similar to GPU backend use vmap to avoid excessive runtime overheads. Context: openxla/community#96 PiperOrigin-RevId: 656199716
nitins17
pushed a commit
to google-ml-infra/jax-fork
that referenced
this pull request
Aug 27, 2024
XLA:CPU is preparing to switch from compiling whole XLA program into a single LLVM function to a mode where each fusion/kernel will have its own entry point, and a thin runtime that will dispatch compute functions concurrently. This execution mode does not work very well with while loops with tiny computations and large number of iterations. Similar to GPU backend use vmap to avoid excessive runtime overheads. Context: openxla/community#96 PiperOrigin-RevId: 656199716
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
3 participants
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.
We would love to hear your thoughts on the OpenXLA CPU Strategy. This is a master RFC on the overall plan and will have follow-up RFCs to flesh out specific details.
For easier discussion, we have put the full RFC in a Google Doc. Please feel free to leave comments/questions directly on the doc or post a longer-form comment here. The feedback period will be open until October 31, 2023. Thank you very much!