[llvm] [AMDGPU][True16] remove waitcnt workaround for D16Writes32BitVgpr and disable D16 instructions in true16 mode (PR #193074)

[Joe Nash via llvm-commits]

================
@@ -3,7 +3,7 @@
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji < %s | FileCheck -check-prefixes=GCN,CIVI,CIVI-HSA %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GFX9 %s
 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GCN,GFX10,GFX10PLUS %s
-; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GCN,GFX10PLUS,GFX11-TRUE16 %s
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16,-d16-write-vgpr32 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GCN,GFX10PLUS,GFX11-TRUE16 %s
----------------
Sisyph wrote:

What's the intention here? It actually makes me thing we should add a runline to each of the flat,global,ds,scratch test files to make sure the d16 load patterns still have test coverage.

https://github.com/llvm/llvm-project/pull/193074