[llvm] [InstCombine] Fold mul (shr exact (X, N)), 2^N + 1 -> add (X , shr exact (X, N)) (PR #112407)

[Yingwei Zheng via llvm-commits]

================
@@ -1077,4 +1077,170 @@ entry:
   ret i32 %shr
 }
 
+define i32 @ashr_shift_mul(i32 noundef %x) {
----------------
dtcxzyw wrote:

`noundef` is not needed.

https://github.com/llvm/llvm-project/pull/112407