[llvm] [AMDGPU][AggressiveInstCombine] Narrow 64 bit math to 32 bit if profitable (PR #130577)

[via llvm-commits]

================
@@ -0,0 +1,210 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -passes=aggressive-instcombine < %s | FileCheck %s
+
+; REQUIRES: amdgpu-registered-target
+
+define i64 @narrow_add(i64 noundef %a, i64 noundef %b) {
+; CHECK-LABEL: define i64 @narrow_add(
+; CHECK-SAME: i64 noundef [[A:%.*]], i64 noundef [[B:%.*]]) #[[ATTR0:[0-9]+]] {
----------------
Shoreshen wrote:

Yes you are very right about this, InstCombine will revert the change. I will move this to `llvm-project/llvm/lib/CodeGen/CodeGenPrepare.cpp`

https://github.com/llvm/llvm-project/pull/130577