[llvm] [AMDGPU] Gate TBUFFER merging on relaxed OOB mode (PR #212695)

[Harrison Hao via llvm-commits]

================
@@ -1,6 +1,63 @@
 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
 # RUN: llc -mtriple=amdgpu9.00 -verify-machineinstrs -run-pass si-load-store-opt -o - %s | FileCheck -check-prefix=GFX9 %s
 
+--- |
+  define void @gfx9_tbuffer_load_x_xyz() { entry: ret void }
----------------
harrisonGPU wrote:

Sorry, I meant the module flag, not function attributes.
These are not full Machine IR tests. They are a collection of small tests for TBuffer combining, covering many different scenarios. That's why there are so many dummy functions.
If we don't want these dummy functions, I think we would need a command-line option instead of using a module flag. Is that correct?
I remember we originally decided not to introduce a command-line option for TBuffer combining, so we ended up using a module flag instead.

https://github.com/llvm/llvm-project/pull/212695