[clang] [llvm] [mlir] [AMDGPU] Verify AMDGPU required workgroup size matches flat workgroup size (PR #200989)

[Dmitry Sidorov via cfe-commits]

================
@@ -15,21 +15,21 @@
 ; CHECK-DAG: OpExecutionMode %[[#ENTRY2]] LocalSize 2 3 1
 ; CHECK-DAG: OpExecutionMode %[[#ENTRY3]] LocalSize 3 1 1
 
-define spir_kernel void @test1() !reqd_work_group_size !1 {
+define spir_kernel void @test1() "amdgpu-flat-work-group-size"="6,6" !reqd_work_group_size !1 {
----------------
MrSidims wrote:

Is presence of `"amdgpu-flat-work-group-size"="6,6"` required to make IR valid? If not, lets remove it. If yes - it's a bit odd to require vendor-specific annotation for generic triple. 

https://github.com/llvm/llvm-project/pull/200989