[llvm] 5ed6229 - [VectorCombine] Add scalarizeLoadExtract infinite loop test from #120984 regression

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2025-01-02T17:23:12Z
New Revision: 5ed6229019de43df0ff4b3e73097781e0f1a6651

URL: https://github.com/llvm/llvm-project/commit/5ed6229019de43df0ff4b3e73097781e0f1a6651
DIFF: https://github.com/llvm/llvm-project/commit/5ed6229019de43df0ff4b3e73097781e0f1a6651.diff

LOG: [VectorCombine] Add scalarizeLoadExtract infinite loop test from #120984 regression

scalarizeLoadExtract replaces instructions up the use list, which can result in the vectorcombine worklist adding users back to the worklist when they should really be erased first.

Added: 
    llvm/test/Transforms/VectorCombine/X86/load-extractelement-scalarization.ll

Modified: 
    

Removed: 
    


################################################################################
diff  --git a/llvm/test/Transforms/VectorCombine/X86/load-extractelement-scalarization.ll b/llvm/test/Transforms/VectorCombine/X86/load-extractelement-scalarization.ll
new file mode 100644
index 00000000000000..0acfeccb92ef72
--- /dev/null
+++ b/llvm/test/Transforms/VectorCombine/X86/load-extractelement-scalarization.ll
@@ -0,0 +1,26 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=sse2 | FileCheck %s
+; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=avx2 | FileCheck %s
+
+; infinite loop if we add the erased instructions to the work list in the wrong order.
+define void @multiple_extract(ptr %p) {
+; CHECK-LABEL: @multiple_extract(
+; CHECK-NEXT:    [[VP:%.*]] = load ptr, ptr [[P:%.*]], align 8
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds <2 x i32>, ptr [[VP]], i32 0, i64 0
+; CHECK-NEXT:    [[E0:%.*]] = load i32, ptr [[TMP1]], align 16
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds <2 x i32>, ptr [[VP]], i32 0, i64 1
+; CHECK-NEXT:    [[E1:%.*]] = load i32, ptr [[TMP2]], align 4
+; CHECK-NEXT:    store i32 [[E0]], ptr [[P]], align 4
+; CHECK-NEXT:    [[P1:%.*]] = getelementptr inbounds nuw i8, ptr [[P]], i64 4
+; CHECK-NEXT:    store i32 [[E1]], ptr [[P1]], align 4
+; CHECK-NEXT:    ret void
+;
+  %vp = load ptr, ptr %p, align 8
+  %v = load <2 x i32>, ptr %vp, align 16
+  %e0 = extractelement <2 x i32> %v, i64 0
+  %e1 = extractelement <2 x i32> %v, i64 1
+  store i32 %e0, ptr %p, align 4
+  %p1 = getelementptr inbounds nuw i8, ptr %p, i64 4
+  store i32 %e1, ptr %p1, align 4
+  ret void
+}