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

[Nikita Popov via llvm-commits]

================
@@ -25,7 +25,10 @@
 
 define i1 @test1(i32 %X) {
 ; CHECK-LABEL: @test1(
-; CHECK-NEXT:    [[R:%.*]] = icmp eq i32 [[X:%.*]], 9
+; CHECK-NEXT:    [[TMP1:%.*]] = shl nsw i32 [[X:%.*]], 1
+; CHECK-NEXT:    [[P:%.*]] = getelementptr inbounds i8, ptr @G16, i32 [[TMP1]]
+; CHECK-NEXT:    [[Q:%.*]] = load i16, ptr [[P]], align 2
+; CHECK-NEXT:    [[R:%.*]] = icmp eq i16 [[Q]], 0
----------------
nikic wrote:

Regression (foldCmpLoadFromIndexedGlobal).

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