[clang] [CIR][AMDGPU] Add support for AMDGCN grid_size builtins (PR #230798)

[Steffen Larsen via cfe-commits]

================
@@ -84,3 +84,21 @@ __device__ float test_logb(float x) {
 // OGCG: load float, ptr addrspace(4) %{{.*}}
 // OGCG: call contract{{.*}} @llvm.fabs.f32
 // OGCG: fcmp contract one float %{{.*}}, +inf
+
+__device__ unsigned test_grid_size_x() {
+  return __builtin_amdgcn_grid_size_x();
----------------
steffenlarsen wrote:

For the sake of coverage, could we also test the other dimensions?

https://github.com/llvm/llvm-project/pull/230798