[llvm] [AMDGPU] Do not fold into v_accvpr_mov/write/read (PR #120475)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,124 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -O3 -mtriple=amdgcn--amdhsa -mcpu=gfx942 %s -o - | FileCheck %s --check-prefixes=GFX942
+; RUN: llc -O3 -mtriple=amdgcn--amdhsa -mcpu=gfx908 %s -o - | FileCheck %s --check-prefixes=GFX908
+
----------------
arsenm wrote:

```suggestion
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx942 %s -o - | FileCheck %s --check-prefixes=GFX942
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx908 %s -o - | FileCheck %s --check-prefixes=GFX908

```

-O3 rarely makes a difference 

https://github.com/llvm/llvm-project/pull/120475