Support prefetched inputs in Pallas pipelines#38165
Open
copybara-service[bot] wants to merge 1 commit into
Open
Support prefetched inputs in Pallas pipelines#38165copybara-service[bot] wants to merge 1 commit into
copybara-service[bot] wants to merge 1 commit into
Conversation
This change introduces support for pre-populated VMEM buffers (prefetched inputs) in Pallas pipelines. The changes introduced include: - Introduces `PrefetchedInput` to wrap a reference and its pre-allocated VMEM buffer. - Updates `emit_pipeline` execution logic to skip initial copy-in steps for these prefetched inputs, avoiding redundant DMA operations. - Adds comprehensive unit tests in `tpu_pallas_pipeline_test.py` for standard, lookahead, and trivial windowing configurations. By using pre-populated window buffer, we can avoid the head pipeline bubbles. The initial wait in can be moved outside the pipeline loop, potential hidden by other computations. PiperOrigin-RevId: 921750865
e30ff9d to
50288fd
Compare
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Support prefetched inputs in Pallas pipelines
This change introduces support for pre-populated VMEM buffers (prefetched inputs) in Pallas pipelines. The changes introduced include:
PrefetchedInputto wrap a reference and its pre-allocated VMEM buffer.emit_pipelineexecution logic to skip initial copy-in steps for these prefetched inputs, avoiding redundant DMA operations.tpu_pallas_pipeline_test.pyfor standard, lookahead, and trivial windowing configurations.By using pre-populated window buffer, we can avoid the head pipeline bubbles. The initial wait in can be moved outside the pipeline loop, potential hidden by other computations.