[clang] clang/AMDGPU: Migrate cc1 tests to subarch triples (10) (PR #211955)

[Matt Arsenault via cfe-commits]

================
@@ -1,6 +1,6 @@
 // REQUIRES: amdgpu-registered-target
 // gfx1200 has neither vmem-to-lds-load-insts nor asynccnt, so asyncmark is unavailable.
-// RUN: %clang_cc1 -O0 -cl-std=CL2.0 -triple amdgcn-amd-amdhsa -target-cpu gfx1200 -verify -S -o - %s
+// RUN: %clang_cc1 -O0 -cl-std=CL2.0 -triple amdgpu12.50-amd-amdhsa -verify -S -o - %s
----------------
arsenm wrote:

```suggestion
// RUN: %clang_cc1 -O0 -cl-std=CL2.0 -triple amdgpu12.00-amd-amdhsa -verify -S -o - %s
```

https://github.com/llvm/llvm-project/pull/211955