[llvm] 7c0089d - [Matrix] Check if iterator is at beginning of BB in optimizeTranspose.
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 14 13:37:26 PDT 2022
Author: Florian Hahn
Date: 2022-06-14T21:37:02+01:00
New Revision: 7c0089d7351cf7ab067e987aa2e32db4b07b3adf
URL: https://github.com/llvm/llvm-project/commit/7c0089d7351cf7ab067e987aa2e32db4b07b3adf
DIFF: https://github.com/llvm/llvm-project/commit/7c0089d7351cf7ab067e987aa2e32db4b07b3adf.diff
LOG: [Matrix] Check if iterator is at beginning of BB in optimizeTranspose.
If an instruction at the beginning of a block is erased, this may
trigger crash due to dereferencing an invalid iterator.
Check if II is at the end before dereferencing it.
Reviewed By: thegameg
Differential Revision: https://reviews.llvm.org/D127736
Added:
llvm/test/Transforms/LowerMatrixIntrinsics/transpose-opts-iterator-invalidation.ll
Modified:
llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp b/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
index d67a314906134..65235a3398e7f 100644
--- a/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
+++ b/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
@@ -704,10 +704,10 @@ class LowerMatrixIntrinsics {
// We may remove II. By default continue on the next/prev instruction.
++II;
// If we were to erase II, move again.
- auto EraseFromParent = [&II](Value *V) {
+ auto EraseFromParent = [&II, &BB](Value *V) {
auto *Inst = cast<Instruction>(V);
if (Inst->use_empty()) {
- if (Inst == &*II) {
+ if (II != BB.rend() && Inst == &*II) {
++II;
}
Inst->eraseFromParent();
diff --git a/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-opts-iterator-invalidation.ll b/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-opts-iterator-invalidation.ll
new file mode 100644
index 0000000000000..5c947ccaca8b2
--- /dev/null
+++ b/llvm/test/Transforms/LowerMatrixIntrinsics/transpose-opts-iterator-invalidation.ll
@@ -0,0 +1,61 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -passes=lower-matrix-intrinsics -S %s | FileCheck %s
+
+define <2 x double> @test(<4 x double> %a, <2 x double> %b, i1 %c) {
+; CHECK-LABEL: @test(
+; CHECK-NEXT: entry:
+; CHECK-NEXT: br i1 [[C:%.*]], label [[EXIT_1:%.*]], label [[EXIT_2:%.*]]
+; CHECK: exit.1:
+; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <4 x double> [[A:%.*]], <4 x double> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT: [[SPLIT1:%.*]] = shufflevector <4 x double> [[A]], <4 x double> poison, <2 x i32> <i32 2, i32 3>
+; CHECK-NEXT: [[SPLIT2:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT: [[BLOCK:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP0:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x double> poison, double [[TMP0]], i32 0
+; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT]], <1 x double> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP1:%.*]] = fmul <1 x double> [[BLOCK]], [[SPLAT_SPLAT]]
+; CHECK-NEXT: [[BLOCK3:%.*]] = shufflevector <2 x double> [[SPLIT1]], <2 x double> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 1
+; CHECK-NEXT: [[SPLAT_SPLATINSERT4:%.*]] = insertelement <1 x double> poison, double [[TMP2]], i32 0
+; CHECK-NEXT: [[SPLAT_SPLAT5:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT4]], <1 x double> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP3:%.*]] = fmul <1 x double> [[BLOCK3]], [[SPLAT_SPLAT5]]
+; CHECK-NEXT: [[TMP4:%.*]] = fadd <1 x double> [[TMP1]], [[TMP3]]
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <1 x double> [[TMP4]], <1 x double> poison, <2 x i32> <i32 0, i32 undef>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <2 x double> undef, <2 x double> [[TMP5]], <2 x i32> <i32 2, i32 1>
+; CHECK-NEXT: [[BLOCK6:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> <i32 1>
+; CHECK-NEXT: [[TMP7:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 0
+; CHECK-NEXT: [[SPLAT_SPLATINSERT7:%.*]] = insertelement <1 x double> poison, double [[TMP7]], i32 0
+; CHECK-NEXT: [[SPLAT_SPLAT8:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT7]], <1 x double> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP8:%.*]] = fmul <1 x double> [[BLOCK6]], [[SPLAT_SPLAT8]]
+; CHECK-NEXT: [[BLOCK9:%.*]] = shufflevector <2 x double> [[SPLIT1]], <2 x double> poison, <1 x i32> <i32 1>
+; CHECK-NEXT: [[TMP9:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 1
+; CHECK-NEXT: [[SPLAT_SPLATINSERT10:%.*]] = insertelement <1 x double> poison, double [[TMP9]], i32 0
+; CHECK-NEXT: [[SPLAT_SPLAT11:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT10]], <1 x double> poison, <1 x i32> zeroinitializer
+; CHECK-NEXT: [[TMP10:%.*]] = fmul <1 x double> [[BLOCK9]], [[SPLAT_SPLAT11]]
+; CHECK-NEXT: [[TMP11:%.*]] = fadd <1 x double> [[TMP8]], [[TMP10]]
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <1 x double> [[TMP11]], <1 x double> poison, <2 x i32> <i32 0, i32 undef>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <2 x double> [[TMP6]], <2 x double> [[TMP12]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT: [[TMP14:%.*]] = extractelement <2 x double> [[TMP13]], i64 0
+; CHECK-NEXT: [[TMP15:%.*]] = insertelement <1 x double> undef, double [[TMP14]], i64 0
+; CHECK-NEXT: [[TMP16:%.*]] = extractelement <2 x double> [[TMP13]], i64 1
+; CHECK-NEXT: [[TMP17:%.*]] = insertelement <1 x double> undef, double [[TMP16]], i64 0
+; CHECK-NEXT: [[TMP18:%.*]] = shufflevector <1 x double> [[TMP15]], <1 x double> [[TMP17]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT: ret <2 x double> [[TMP18]]
+; CHECK: exit.2:
+; CHECK-NEXT: ret <2 x double> zeroinitializer
+;
+entry:
+ %mult = call <2 x double> @llvm.matrix.multiply.v2f64.v4f64.v2f64(<4 x double> %a, <2 x double> %b, i32 2, i32 2, i32 1)
+ br i1 %c, label %exit.1, label %exit.2
+
+exit.1:
+ %trans = call <2 x double> @llvm.matrix.transpose.v2f64(<2 x double> %mult, i32 2, i32 1)
+ ret <2 x double> %trans
+
+exit.2:
+ ret <2 x double> zeroinitializer
+}
+
+declare <2 x double> @llvm.matrix.transpose.v2f64(<2 x double>, i32 immarg, i32 immarg)
+
+declare <2 x double> @llvm.matrix.multiply.v2f64.v4f64.v2f64(<4 x double>, <2 x double>, i32 immarg, i32 immarg, i32 immarg)
More information about the llvm-commits
mailing list