feat(RISCVCombines): add combines to speed-up ChaCha20 - #1344
Open
tobias-rothmann wants to merge 9 commits into
Open
tobias-rothmann wants to merge 9 commits into
tobias-rothmann wants to merge 9 commits into
Conversation
tobias-rothmann
changed the title
tobias-rothmann
changed the title
tobias-rothmann
force-pushed
the
tr-zextw-xor-roriw-combine
branch
2 times, most recently
from
September 3, 2026 13:19
c7f46ad to
897c18b
Compare
tobias-rothmann
changed the title
tobias-rothmann
marked this pull request as ready for review
September 3, 2026 13:20
tobias-rothmann
force-pushed
the
tr-zextw-xor-roriw-combine
branch
from
September 3, 2026 13:24
897c18b to
1eb01e2
Compare
tobias-rothmann
changed the title
tobias-rothmann
changed the title
tobias-rothmann
marked this pull request as draft
September 3, 2026 13:34
tobias-rothmann
force-pushed
the
tr-zextw-xor-roriw-combine
branch
2 times, most recently
from
September 4, 2026 13:46
f2ee4fc to
9c361d5
Compare
tobias-rothmann
force-pushed
the
tr-zextw-xor-roriw-combine
branch
from
September 4, 2026 14:28
3155a69 to
55e829f
Compare
tobias-rothmann
marked this pull request as ready for review
September 4, 2026 14:31
Contributor
|
thanks a lot for the PR, the code looks good to me, I left a couple comments! I will ask a second look by @regehr and @naveen-seth - do you guys think these rewrites will interfere with the legalization work? |
| LLVM: `CastInst::isEliminableCastPair` folds a `ZExt` followed by a `ZExt` | ||
| to the first cast. | ||
| https://github.com/llvm/llvm-project/blob/c536b0aa030474672e293dccdb27b97c36c4e1af/llvm/lib/IR/Instructions.cpp#L2922-L2967 -/ | ||
| private def zextw_zextb_pattern : Puddle.Pattern OpCode := |
Contributor
There was a problem hiding this comment.
why is this definition private? (same for all the defs in this file)
Author
There was a problem hiding this comment.
I'm not sure. I don't think there's a coherent pattern, but it seems like you only want to expose the compiled version for some combines. I've removed the "private" tag from all definitions now.
Contributor
|
Upstream seems to make these optimizations during the legalization step, but I'd prefer landing this and to then move/revert this later once we have gotten further on legalization. |
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
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.
Adds verified RISC-V combines:
roriw (xor (zextw x) (zextw y)), imm→roriw (xor x y), imm, removing redundant extensions in the ChaCha20 idiom while preserving shared full-width XOR uses.zextw (zextb x)→zextb x, including fixed-register retargeting.Includes correctness proofs and regression tests for matching, negative, shared-use, and fixed-register cases.