[llvm] 202137d - [SLP]Fix a crash on matching gather operands of phi nodes in loops
Alexey Bataev via llvm-commits
llvm-commits at lists.llvm.org
Wed Mar 12 14:46:10 PDT 2025
Author: Alexey Bataev
Date: 2025-03-12T14:46:00-07:00
New Revision: 202137dbead88004c956595106ab6d80d597508c
URL: https://github.com/llvm/llvm-project/commit/202137dbead88004c956595106ab6d80d597508c
DIFF: https://github.com/llvm/llvm-project/commit/202137dbead88004c956595106ab6d80d597508c.diff
LOG: [SLP]Fix a crash on matching gather operands of phi nodes in loops
If the gather operands in phi nodes are matching and phi nodes may build
up a loop, it may cause a compiler crash with the incorrect def-use
chain. Patch fixes this crash.
Added:
llvm/test/Transforms/SLPVectorizer/X86/phi-nodes-incoming-same-blocks.ll
Modified:
llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
llvm/test/Transforms/SLPVectorizer/X86/delayed-gather-emission.ll
llvm/test/Transforms/SLPVectorizer/X86/matching-gather-nodes-phi-users.ll
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 3ffcdb550f8ff..a2200f283168d 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -14104,9 +14104,7 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(
TEUseEI.UserTE->getOpcode() == Instruction::PHI &&
UseEI.UserTE->State == TreeEntry::Vectorize &&
UseEI.UserTE->getOpcode() == Instruction::PHI &&
- TEUseEI.UserTE != UseEI.UserTE &&
- TEUseEI.UserTE->getMainOp()->getParent() ==
- UseEI.UserTE->getMainOp()->getParent())
+ TEUseEI.UserTE != UseEI.UserTE)
continue;
// If 2 gathers are operands of the same entry (regardless of whether
// user is PHI or else), compare operands indices, use the earlier one
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/delayed-gather-emission.ll b/llvm/test/Transforms/SLPVectorizer/X86/delayed-gather-emission.ll
index 5562291dbb6be..249b3f9329319 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/delayed-gather-emission.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/delayed-gather-emission.ll
@@ -29,9 +29,9 @@ define void @test() {
; CHECK-NEXT: tail call void @foo(float [[MUL]])
; CHECK-NEXT: [[I2:%.*]] = load float, ptr poison, align 4
; CHECK-NEXT: [[TOBOOL:%.*]] = fcmp une float [[I2]], 0.000000e+00
+; CHECK-NEXT: [[TMP10]] = insertelement <2 x float> [[TMP2]], float [[I2]], i32 0
; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <2 x float> [[TMP5]], <2 x float> poison, <2 x i32> <i32 poison, i32 0>
; CHECK-NEXT: [[TMP9]] = insertelement <2 x float> [[TMP8]], float [[I2]], i32 0
-; CHECK-NEXT: [[TMP10]] = shufflevector <2 x float> [[TMP9]], <2 x float> [[TMP2]], <2 x i32> <i32 0, i32 3>
; CHECK-NEXT: br i1 [[TOBOOL]], label [[BB1]], label [[BB2]]
;
entry:
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/matching-gather-nodes-phi-users.ll b/llvm/test/Transforms/SLPVectorizer/X86/matching-gather-nodes-phi-users.ll
index 7960278d2b21d..3c0891349f79a 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/matching-gather-nodes-phi-users.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/matching-gather-nodes-phi-users.ll
@@ -8,7 +8,7 @@
; YAML: Function: test
; YAML: Args:
; YAML: - String: 'Stores SLP vectorized with cost '
-; YAML: - Cost: '-6'
+; YAML: - Cost: '-3'
; YAML: - String: ' and with tree size '
; YAML: - TreeSize: '16'
; YAML: ...
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/phi-nodes-incoming-same-blocks.ll b/llvm/test/Transforms/SLPVectorizer/X86/phi-nodes-incoming-same-blocks.ll
new file mode 100644
index 0000000000000..048d2814b9abb
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/phi-nodes-incoming-same-blocks.ll
@@ -0,0 +1,63 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -mcpu=cascadelake < %s | FileCheck %s
+
+define void @test(ptr %0, i1 %1, i1 %2) {
+; CHECK-LABEL: define void @test(
+; CHECK-SAME: ptr [[TMP0:%.*]], i1 [[TMP1:%.*]], i1 [[TMP2:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: br label %[[BB4:.*]]
+; CHECK: [[BB4]]:
+; CHECK-NEXT: [[TMP5:%.*]] = phi <2 x i32> [ [[TMP12:%.*]], %[[TMP7:.*]] ], [ zeroinitializer, [[TMP3:%.*]] ]
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <2 x i32> [[TMP5]], <2 x i32> poison, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
+; CHECK-NEXT: br i1 [[TMP1]], label %[[TMP7]], label %[[BB14:.*]]
+; CHECK: [[TMP7]]:
+; CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[TMP0]], align 8
+; CHECK-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[TMP8]], i64 16
+; CHECK-NEXT: [[TMP10:%.*]] = load <2 x i32>, ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP11:%.*]] = or <2 x i32> [[TMP10]], splat (i32 1)
+; CHECK-NEXT: [[TMP12]] = shufflevector <2 x i32> [[TMP11]], <2 x i32> <i32 1, i32 poison>, <2 x i32> <i32 2, i32 1>
+; CHECK-NEXT: [[TMP13:%.*]] = call <4 x i32> @llvm.vector.insert.v4i32.v2i32(<4 x i32> <i32 0, i32 0, i32 poison, i32 poison>, <2 x i32> [[TMP11]], i64 2)
+; CHECK-NEXT: br i1 [[TMP2]], label %[[BB15:.*]], label %[[BB4]]
+; CHECK: [[BB14]]:
+; CHECK-NEXT: br label %[[BB15]]
+; CHECK: [[BB15]]:
+; CHECK-NEXT: [[TMP16:%.*]] = phi <4 x i32> [ [[TMP6]], %[[BB14]] ], [ [[TMP13]], %[[TMP7]] ]
+; CHECK-NEXT: [[TMP17:%.*]] = load volatile ptr, ptr null, align 8
+; CHECK-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP17]], i64 176
+; CHECK-NEXT: store <4 x i32> [[TMP16]], ptr [[TMP18]], align 8
+; CHECK-NEXT: ret void
+;
+ br label %4
+
+4:
+ %5 = phi i32 [ %14, %7 ], [ 0, %3 ]
+ %6 = phi i32 [ 1, %7 ], [ 0, %3 ]
+ br i1 %1, label %7, label %15
+
+7:
+ %8 = load ptr, ptr %0, align 8
+ %9 = getelementptr i8, ptr %8, i64 16
+ %10 = load i32, ptr %9, align 1
+ %11 = or i32 %10, 1
+ %12 = getelementptr i8, ptr %8, i64 20
+ %13 = load i32, ptr %12, align 1
+ %14 = or i32 %13, 1
+ br i1 %2, label %16, label %4
+
+15:
+ br label %16
+
+16:
+ %17 = phi i32 [ %6, %15 ], [ 0, %7 ]
+ %18 = phi i32 [ %6, %15 ], [ %11, %7 ]
+ %19 = phi i32 [ %5, %15 ], [ %14, %7 ]
+ %20 = load volatile ptr, ptr null, align 8
+ %21 = getelementptr i8, ptr %20, i64 176
+ store i32 %17, ptr %21, align 8
+ %22 = getelementptr i8, ptr %20, i64 180
+ store i32 %17, ptr %22, align 4
+ %23 = getelementptr i8, ptr %20, i64 184
+ store i32 %18, ptr %23, align 8
+ %24 = getelementptr i8, ptr %20, i64 188
+ store i32 %19, ptr %24, align 4
+ ret void
+}
More information about the llvm-commits
mailing list