[llvm] [AMDGPU][CodeGen] Improve handling of memcpy for -Os/-Oz compilations (PR #87632)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,642 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx908 %s -o - | FileCheck %s
----------------
arsenm wrote:

use -mtriple=amdgcn-amd-amdhsa so we don't get the weird scratch ABI 

https://github.com/llvm/llvm-project/pull/87632