[clang] [llvm] [mlir] [OpenMP] Reserve the main thread's warp for generic mode kernels (PR #218790)

[Larry Meadows via cfe-commits]

================
@@ -35,29 +35,32 @@ void func() {
 }
 
 // SPIRV: attributes #0
-// SPIRV-SAME: "nvvm.maxntid"="20"
-// SPIRV-SAME: "omp_target_thread_limit"="20" 
+// SPIRV-SAME: "nvvm.maxntid"="84"
+// SPIRV-SAME: "omp_target_thread_limit"="84" 
 // SPIRV: attributes #4
 // SPIRV-SAME: "amdgpu-waves-per-eu"="3,7"
 // SPIRV-SAME: "nvvm.maxntid"="17"
 // SPIRV-SAME: "omp_target_thread_limit"="17"
 
+// The first two kernels are generic mode and reserve room for the main
----------------
lfmeadow wrote:

I spanked the AI. They are gone.

https://github.com/llvm/llvm-project/pull/218790