[llvm-branch-commits] [llvm] [LV] Optimize partial reduction extends before handling inloop subs (PR #199665)

[Sander de Smalen via llvm-branch-commits]

================
@@ -380,3 +380,60 @@ for.body:
 exit:
   ret i32 %sum.1
 }
+
+define i32 @sub_add_chain_unsigned_absolute_difference(ptr noalias %x, ptr noalias %y) {
----------------
sdesmalen-arm wrote:

nit: maybe add a comment describing what this is testing specifically?

https://github.com/llvm/llvm-project/pull/199665