enable unroll on top of vectorization in pointwise scheduler #3249
Conversation
This file contains 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
|
!build --pybench |
|
!build --pybench |
… test code diff" This reverts commit b23cb41.
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Outdated suggestions cannot be applied.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.
What's in this PR: Following #3067, enable unroll on top of vectorization in pointwise scheduler
Heuristics:
Unrollincreased instruction level parallelism but reduced threads-level paralelism, it is beneficial when(1) computation latency is close to memory latency
(2) threads-level paralelism is already saturated, e.g. each SM needs to process 8 * 64 warps
when these two conditions are met,
unrollincreasedPerformance:
(0) See doc for other hardwards.
(1) H100 [float16, no change for float32]
(1.1) Gelu Bias [1.05x speedup]
(1.2) Silu Mul [No change]
(1.3) Pointwise Mul [No change]
