[llvm] [InstCombine] Canonicalize gep scaling factor (PR #180216)

[Nikita Popov via llvm-commits]

================
@@ -85,8 +90,11 @@ define ptr @test_zero_sext_add_nsw(ptr %base, i32 %a) {
 ; CHECK-LABEL: define ptr @test_zero_sext_add_nsw(
 ; CHECK-SAME: ptr [[BASE:%.*]], i32 [[A:%.*]]) {
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = sext i32 [[A]] to i64
-; CHECK-NEXT:    [[P2:%.*]] = getelementptr i32, ptr [[BASE]], i64 [[TMP0]]
+; CHECK-NEXT:    [[P1:%.*]] = getelementptr i8, ptr [[BASE]], i64 -4
+; CHECK-NEXT:    [[INDEX:%.*]] = add nsw i32 [[A]], 1
+; CHECK-NEXT:    [[TMP0:%.*]] = sext i32 [[INDEX]] to i64
+; CHECK-NEXT:    [[TMP1:%.*]] = shl nsw i64 [[TMP0]], 2
+; CHECK-NEXT:    [[P2:%.*]] = getelementptr i8, ptr [[P1]], i64 [[TMP1]]
----------------
nikic wrote:

Regression

https://github.com/llvm/llvm-project/pull/180216