[nvidia backend] Replace cvt instructions with bitwise operations in s8->bf16 conversions #4563
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
Hopper has very low throughput of conversion instructions that cause this operations to quickly become an ALU bottleneck. Restating it in terms of bitwise ops and SIMD bf16 instructions increases the throughput significantly and translates to meaningful speedups (e.g. 10% end-to-end on one matmul I was looking at).
chsigg
force-pushed
the
export_cl666252733
branch
from
56f40ca to
f5fb50f
Compare
bertmaher
pushed a commit
to bertmaher/triton
that referenced
this pull request
Dec 10, 2024
…s8->bf16 conversions (triton-lang#4563) Hopper has very low throughput of conversion instructions that cause this operations to quickly become an ALU bottleneck. Restating it in terms of bitwise ops and SIMD bf16 instructions increases the throughput significantly and translates to meaningful speedups (e.g. 10% end-to-end on one matmul I was looking at). Co-authored-by: Adam Paszke <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
Hopper has very low throughput of conversion instructions that cause this operations to quickly become an ALU bottleneck. Restating it in terms of bitwise ops and SIMD bf16 instructions increases the throughput significantly and translates to meaningful speedups (e.g. 10% end-to-end on one matmul I was looking at).