Add LoopVectorizer support for llvm.vector.partial.reduce.fadd
#163975
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
SamTebbs33
reviewed
Oct 20, 2025
llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-fdot-product.ll
Show resolved
Hide resolved
With this intrinsic, and supporting SelectionDAG nodes, we can better make use of instructions such as AArch64's `FDOT`.
Corrected LangRef typos, improved const comparisons for fadd, and add direct tests.
This reverts commit 319852132602f685aea6228f10418370fd530aa7.
dheaton-arm
force-pushed
the
loopvectorize-partial-reduce-fadd
branch
from
c27c92c to
7005049
Compare
SamTebbs33
approved these changes
Oct 22, 2025
Comment on lines
+1064
to
+1101
|
|
||
|
|
||
| ;. | ||
| ; CHECK-INTERLEAVE1: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} | ||
| ; CHECK-INTERLEAVE1: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} | ||
| ; CHECK-INTERLEAVE1: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} | ||
| ; CHECK-INTERLEAVE1: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]} | ||
| ; CHECK-INTERLEAVE1: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]} | ||
| ; CHECK-INTERLEAVE1: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]} | ||
| ; CHECK-INTERLEAVE1: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]} | ||
| ; CHECK-INTERLEAVE1: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]} | ||
| ; CHECK-INTERLEAVE1: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]} | ||
| ; CHECK-INTERLEAVE1: [[LOOP9]] = distinct !{[[LOOP9]], [[META2]], [[META1]]} | ||
| ;. | ||
| ; CHECK-INTERLEAVED: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} | ||
| ; CHECK-INTERLEAVED: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} | ||
| ; CHECK-INTERLEAVED: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} | ||
| ; CHECK-INTERLEAVED: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]} | ||
| ; CHECK-INTERLEAVED: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]} | ||
| ; CHECK-INTERLEAVED: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]} | ||
| ; CHECK-INTERLEAVED: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]} | ||
| ; CHECK-INTERLEAVED: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]} | ||
| ; CHECK-INTERLEAVED: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]} | ||
| ; CHECK-INTERLEAVED: [[LOOP9]] = distinct !{[[LOOP9]], [[META2]], [[META1]]} | ||
| ;. | ||
| ; CHECK-MAXBW: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} | ||
| ; CHECK-MAXBW: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} | ||
| ; CHECK-MAXBW: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} | ||
| ; CHECK-MAXBW: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]} | ||
| ; CHECK-MAXBW: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]} | ||
| ; CHECK-MAXBW: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]} | ||
| ; CHECK-MAXBW: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]} | ||
| ; CHECK-MAXBW: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]} | ||
| ; CHECK-MAXBW: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]} | ||
| ; CHECK-MAXBW: [[LOOP9]] = distinct !{[[LOOP9]], [[META2]], [[META1]]} | ||
| ; CHECK-MAXBW: [[LOOP10]] = distinct !{[[LOOP10]], [[META1]], [[META2]]} | ||
| ; CHECK-MAXBW: [[LOOP11]] = distinct !{[[LOOP11]], [[META2]], [[META1]]} | ||
| ;. |
There was a problem hiding this comment.
I know it's a bit of a pain but we tend to remove these. You can add --check-globals none to the update script command to stop them from being generated.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
This PR follows on from #159776, and therefore is dependent on it.
See 8236481 for the first relevant changes for this PR (everything prior is #159776).
Allows the Loop Vectorizer to generate
llvm.vector.partial.reduce.faddintrinsics when sequences which match its requirements are found.