[llvm] [AMDGPU] Insert MFMA anti-hints in GCNPreRAOptimizations (PR #218075)

[Matt Arsenault via llvm-commits]

================
@@ -0,0 +1,135 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 6
+# RUN: llc -mtriple=amdgpu9.42-amd-amdhsa -run-pass=amdgpu-pre-ra-optimizations,greedy,machineverifier,virtregrewriter %s -o - | FileCheck %s --check-prefix=CHECK
+# RUN: llc -mtriple=amdgpu9.42-amd-amdhsa -run-pass=amdgpu-pre-ra-optimizations,greedy,machineverifier,virtregrewriter -amdgpu-anti-hints=false %s -o - | FileCheck %s --check-prefix=CHECK-NO-ANTIHINT
+
+--- |
+  define amdgpu_kernel void @test_anti_hints() #0 { ret void }
+  define amdgpu_kernel void @test_anti_hints_multi_mfma() #0 { ret void }
+
+  attributes #0 = {nounwind "amdgpu-agpr-alloc"="0" "frame-pointer"="none"}
----------------
arsenm wrote:

```suggestion
  attributes #0 = { nounwind "amdgpu-agpr-alloc"="0" "frame-pointer"="none" }
```

https://github.com/llvm/llvm-project/pull/218075