[WIP][X86] combineX86ShufflesRecursively - attempt to combine shuffles with larger types from EXTRACT_SUBVECTOR nodes #133947
+49,461
−48,366
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
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
2 times, most recently
from
cac4c3b to
b54089c
Compare
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
from
b54089c to
ed7b501
Compare
RKSimon
added a commit
to RKSimon/llvm-project
that referenced
this pull request
Apr 2, 2025
…/dst sizes Cleanup work for llvm#133947 - we need to handle VTRUNC nodes with large source vectors directly to allow us to widen the size of the shuffle combine We currently discard these results in combineX86ShufflesRecursively anyhow as we don't allow inputs from getTargetShuffleInputs to be larger than the shuffle value type
RKSimon
added a commit
that referenced
this pull request
Apr 3, 2025
…/dst sizes (#134161) Cleanup work for #133947 - we need to handle VTRUNC nodes with large source vectors directly to allow us to widen the size of the shuffle combine We currently discard these results in combineX86ShufflesRecursively anyhow as we don't allow inputs from getTargetShuffleInputs to be larger than the shuffle value type
RKSimon
added a commit
to RKSimon/llvm-project
that referenced
this pull request
Apr 3, 2025
…TOR(SRC1)) matching behind common one use bitcast checks No need to ignore one use checks for the INSERT_SUBVECTOR(SRC0, EXTRACT_SUBVECTOR(SRC1)) fold Noticed while working on the llvm#133947 regressions
RKSimon
added a commit
to RKSimon/llvm-project
that referenced
this pull request
Apr 3, 2025
…r value type. NFC. Prep work for llvm#133947
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
4 times, most recently
from
d575414 to
db33ee9
Compare
RKSimon
added a commit
that referenced
this pull request
Apr 7, 2025
…ffle operands. NFC. Merge loops to peek through free insert_subvector / bitcasts / extract_subvector. To keep this NFC I haven't reordered the peek throughs - this will done in a future patch to help with #133947 regressions
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
from
db33ee9 to
a71e9fb
Compare
RKSimon
added a commit
to RKSimon/llvm-project
that referenced
this pull request
Apr 9, 2025
…patterns should return bitcasted source values Noticed while investigating llvm#133947 regressions - if we peek through bitcasts we can lose track of oneuse/combined nodes Same current codegen as combineX86ShufflesRecursively still peeks through the bitcasts itself, but we will soon handle this consistently as another part of llvm#133947
RKSimon
added a commit
that referenced
this pull request
Apr 9, 2025
…patterns should return bitcasted source values (#134993) Noticed while investigating #133947 regressions - if we peek through bitcasts we can lose track of oneuse/combined nodes in shuffle combining Currently the same codegen as combineX86ShufflesRecursively still peeks through the bitcasts itself, but we will soon handle this consistently as another part of #133947
AllinLeeYL
pushed a commit
to AllinLeeYL/llvm-project
that referenced
this pull request
Apr 10, 2025
…patterns should return bitcasted source values (llvm#134993) Noticed while investigating llvm#133947 regressions - if we peek through bitcasts we can lose track of oneuse/combined nodes in shuffle combining Currently the same codegen as combineX86ShufflesRecursively still peeks through the bitcasts itself, but we will soon handle this consistently as another part of llvm#133947
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
from
a71e9fb to
96bdbb4
Compare
var-const
pushed a commit
to ldionne/llvm-project
that referenced
this pull request
Apr 17, 2025
…patterns should return bitcasted source values (llvm#134993) Noticed while investigating llvm#133947 regressions - if we peek through bitcasts we can lose track of oneuse/combined nodes in shuffle combining Currently the same codegen as combineX86ShufflesRecursively still peeks through the bitcasts itself, but we will soon handle this consistently as another part of llvm#133947
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
3 times, most recently
from
3139775 to
eec9df5
Compare
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
from
eec9df5 to
f0dd9c3
Compare
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
3 times, most recently
from
4b633b8 to
a1ab605
Compare
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
from
a1ab605 to
1d033a2
Compare
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
from
1d033a2 to
93bc1c2
Compare
…s with larger types from EXTRACT_SUBVECTOR nodes This replaces the rather limited combineX86ShuffleChainWithExtract function with handling for EXTRACT_SUBVECTOR node as we recurse down the shuffle chain, widening the shuffle mask to accommodate the larger value type. This will mainly help AVX2/AVX512 cases with cross-lane shuffles, but it also helps collapse some cases where the same subvector has gotten reused in multiple lanes. Exposed missing DemandedElts handling inside ISD::TRUNCATE nodes for ComputeNumSignBits
RKSimon
force-pushed
the
x86-combine-shuffles-extract
branch
from
93bc1c2 to
5867b13
Compare
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.
This replaces the rather limited combineX86ShuffleChainWithExtract function with handling for EXTRACT_SUBVECTOR node as we recurse down the shuffle chain, widening the shuffle mask to accomodate the larger value type.
This will mainly help AVX2/AVX512 cases with cross-lane shuffles, but it also helps collapse some cases where the same subvector has gotten reused in multiple lanes.
Exposed missing DemandedElts handling inside ISD::TRUNCATE nodes for ComputeNumSignBits
Fixes #143158