[SLP]Do extra analysis int minbitwidth if some checks return false. #84363
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
Created using spr 1.3.5
|
@llvm/pr-subscribers-llvm-transforms Author: Alexey Bataev (alexey-bataev) ChangesThe instruction itself can be considered good for minbitwidth casting, Full diff: https://github.com/llvm/llvm-project/pull/84363.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 1889bc09e85028..c6502c9f77f001 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -13942,7 +13942,7 @@ bool BoUpSLP::collectValuesToDemote(
!collectValuesToDemote(I->getOperand(1), IsProfitableToDemoteRoot,
BitWidth, ToDemote, DemotedConsts, Visited,
Level2, IsProfitableToDemote))
- return false;
+ return IsProfitableToDemote && IsPotentiallyTruncated(I, BitWidth);
MaxDepthLevel = std::max(Level1, Level2);
break;
}
@@ -13958,7 +13958,7 @@ bool BoUpSLP::collectValuesToDemote(
!collectValuesToDemote(SI->getFalseValue(), IsProfitableToDemoteRoot,
BitWidth, ToDemote, DemotedConsts, Visited,
Level2, IsProfitableToDemote))
- return false;
+ return IsProfitableToDemote && IsPotentiallyTruncated(I, BitWidth);
MaxDepthLevel = std::max(Level1, Level2);
break;
}
@@ -13971,7 +13971,7 @@ bool BoUpSLP::collectValuesToDemote(
if (!collectValuesToDemote(IncValue, IsProfitableToDemoteRoot, BitWidth,
ToDemote, DemotedConsts, Visited,
MaxDepthLevel, IsProfitableToDemote))
- return false;
+ return IsProfitableToDemote && IsPotentiallyTruncated(I, BitWidth);
break;
}
diff --git a/llvm/test/Transforms/SLPVectorizer/AArch64/horizontal.ll b/llvm/test/Transforms/SLPVectorizer/AArch64/horizontal.ll
index 1986b51ec94828..02d1f9f60d0ca1 100644
--- a/llvm/test/Transforms/SLPVectorizer/AArch64/horizontal.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AArch64/horizontal.ll
@@ -228,7 +228,7 @@ for.end: ; preds = %for.end.loopexit, %
; YAML-NEXT: Function: test_unrolled_select
; YAML-NEXT: Args:
; YAML-NEXT: - String: 'Vectorized horizontal reduction with cost '
-; YAML-NEXT: - Cost: '-36'
+; YAML-NEXT: - Cost: '-40'
; YAML-NEXT: - String: ' and with tree size '
; YAML-NEXT: - TreeSize: '10'
@@ -246,15 +246,17 @@ define i32 @test_unrolled_select(ptr noalias nocapture readonly %blk1, ptr noali
; CHECK-NEXT: [[P2_045:%.*]] = phi ptr [ [[BLK2:%.*]], [[FOR_BODY_LR_PH]] ], [ [[ADD_PTR88:%.*]], [[IF_END_86]] ]
; CHECK-NEXT: [[P1_044:%.*]] = phi ptr [ [[BLK1:%.*]], [[FOR_BODY_LR_PH]] ], [ [[ADD_PTR:%.*]], [[IF_END_86]] ]
; CHECK-NEXT: [[TMP0:%.*]] = load <8 x i8>, ptr [[P1_044]], align 1
-; CHECK-NEXT: [[TMP1:%.*]] = zext <8 x i8> [[TMP0]] to <8 x i32>
+; CHECK-NEXT: [[TMP1:%.*]] = zext <8 x i8> [[TMP0]] to <8 x i16>
; CHECK-NEXT: [[TMP2:%.*]] = load <8 x i8>, ptr [[P2_045]], align 1
-; CHECK-NEXT: [[TMP3:%.*]] = zext <8 x i8> [[TMP2]] to <8 x i32>
-; CHECK-NEXT: [[TMP4:%.*]] = sub nsw <8 x i32> [[TMP1]], [[TMP3]]
-; CHECK-NEXT: [[TMP5:%.*]] = icmp slt <8 x i32> [[TMP4]], zeroinitializer
-; CHECK-NEXT: [[TMP6:%.*]] = sub nsw <8 x i32> zeroinitializer, [[TMP4]]
-; CHECK-NEXT: [[TMP7:%.*]] = select <8 x i1> [[TMP5]], <8 x i32> [[TMP6]], <8 x i32> [[TMP4]]
-; CHECK-NEXT: [[TMP8:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP7]])
-; CHECK-NEXT: [[OP_RDX]] = add i32 [[TMP8]], [[S_047]]
+; CHECK-NEXT: [[TMP3:%.*]] = zext <8 x i8> [[TMP2]] to <8 x i16>
+; CHECK-NEXT: [[TMP4:%.*]] = sub <8 x i16> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = trunc <8 x i16> [[TMP4]] to <8 x i1>
+; CHECK-NEXT: [[TMP6:%.*]] = icmp slt <8 x i1> [[TMP5]], zeroinitializer
+; CHECK-NEXT: [[TMP7:%.*]] = sub <8 x i16> zeroinitializer, [[TMP4]]
+; CHECK-NEXT: [[TMP8:%.*]] = select <8 x i1> [[TMP6]], <8 x i16> [[TMP7]], <8 x i16> [[TMP4]]
+; CHECK-NEXT: [[TMP9:%.*]] = zext <8 x i16> [[TMP8]] to <8 x i32>
+; CHECK-NEXT: [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP9]])
+; CHECK-NEXT: [[OP_RDX]] = add i32 [[TMP10]], [[S_047]]
; CHECK-NEXT: [[CMP83:%.*]] = icmp slt i32 [[OP_RDX]], [[LIM:%.*]]
; CHECK-NEXT: br i1 [[CMP83]], label [[IF_END_86]], label [[FOR_END_LOOPEXIT:%.*]]
; CHECK: if.end.86:
|
Created using spr 1.3.5
Created using spr 1.3.5
alexey-bataev
deleted the
users/alexey-bataev/spr/slpdo-extra-analysis-int-minbitwidth-if-some-checks-return-false
branch
alexey-bataev
added a commit
that referenced
this pull request
Mar 14, 2024
The instruction itself can be considered good for minbitwidth casting, even if one of the operand checks returns false. Reviewers: RKSimon Reviewed By: RKSimon Pull Request: #84363
|
This caused a crash building libjpeg's jdsample.c. Here's a reduced testcase: Compile with: (I didn't try to reduce the flags more than removing preprocessing and warning flags) A debug build shows this: |
alexey-bataev
added a commit
that referenced
this pull request
Mar 19, 2024
The instruction itself can be considered good for minbitwidth casting, even if one of the operand checks returns false. Reviewers: RKSimon Reviewed By: RKSimon Pull Request: #84363
|
Committed in da118c9 |
|
I found two new cases that trigger failed asserts, based on the new relanded commit, in openh264 and libschroedinger. Those cases are: typedef int a;
short *b;
short c;
void d() {
a e, f;
e = 0;
for (; e < 16; e += 4) {
b[e] = (f ^ (f ^ b[e] - f) * c >> 16) - f;
f = b[e + 1] >> 1;
b[e + 1] = f ^ ((f ^ b[e + 1] - f) * c >> 16) - f;
b[e + 2] = f ^ ((f ^ b[e + 2] - f) * c >> 16) - f;
b[e + 3] = f ^ ((f ^ b[e + 3] - f) * c >> 16) - f;
}
}typedef struct {
short a
} b;
int c, d;
b *e;
b f, g;
short h, i, j;
void k() {
short l;
d = 0;
for (; d < 8; d++) {
f = e[d];
l = f.a < -1 ? -1 : f.a > 1 ? 1 : f.a;
h = f.a < 0 ? 0 : f.a;
i = h * c;
j = i >> 2;
g.a = j * l;
e[d] = g;
}
}Both can be triggered with the following command - but they appear to trigger failed asserts on a couple different source lines: |
|
Thanks, reverting it for the proper fix. |
alexey-bataev
added a commit
that referenced
this pull request
Mar 20, 2024
The instruction itself can be considered good for minbitwidth casting, even if one of the operand checks returns false. Reviewers: RKSimon Reviewed By: RKSimon Pull Request: #84363
Fixed in 81d9ed6 |
Thanks - from a quick rerun it seems like all these cases are fixed now, and large parts of my builds pass (knowing conclusively takes much longer though). |
|
Hi @alexey-bataev It fails like: |
|
Thanks, will fix it ASAP. |
Fixed in 34f0a8a |
Yep, thanks! |
chencha3
pushed a commit
to chencha3/llvm-project
that referenced
this pull request
Mar 23, 2024
The instruction itself can be considered good for minbitwidth casting, even if one of the operand checks returns false. Reviewers: RKSimon Reviewed By: RKSimon Pull Request: llvm#84363
chencha3
pushed a commit
to chencha3/llvm-project
that referenced
this pull request
Mar 23, 2024
…false." This reverts commit da118c9 to fix crashes reported in llvm#84363.
chencha3
pushed a commit
to chencha3/llvm-project
that referenced
this pull request
Mar 23, 2024
The instruction itself can be considered good for minbitwidth casting, even if one of the operand checks returns false. Reviewers: RKSimon Reviewed By: RKSimon Pull Request: llvm#84363
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.
The instruction itself can be considered good for minbitwidth casting,
even if one of the operand checks returns false.