[NVPTX] Add a clang builtin for the warpsize intrinsic (#110316)

Summary:
There's an intrinsic for the warp size, we want to expose this to make
the interface proposed in
#110179 more generic.

Author: jhuber6

clang/include/clang/Basic/BuiltinsNVPTX.def

@@ -139,6 +139,7 @@ TARGET_BUILTIN(__nvvm_is_explicit_cluster, "b", "nc", AND(SM_90, PTX78))
 BUILTIN(__nvvm_read_ptx_sreg_laneid, "i", "nc")
 BUILTIN(__nvvm_read_ptx_sreg_warpid, "i", "nc")
 BUILTIN(__nvvm_read_ptx_sreg_nwarpid, "i", "nc")
+BUILTIN(__nvvm_read_ptx_sreg_warpsize, "i", "nc")
 
 BUILTIN(__nvvm_read_ptx_sreg_smid, "i", "nc")
 BUILTIN(__nvvm_read_ptx_sreg_nsmid, "i", "nc")

clang/test/CodeGen/builtins-nvptx.c

@@ -114,15 +114,17 @@ __device__ int read_ids() {
 // CHECK: call i32 @llvm.nvvm.read.ptx.sreg.smid()
 // CHECK: call i32 @llvm.nvvm.read.ptx.sreg.nsmid()
 // CHECK: call i32 @llvm.nvvm.read.ptx.sreg.gridid()
+// CHECK: call i32 @llvm.nvvm.read.ptx.sreg.warpsize()
 
   int a = __nvvm_read_ptx_sreg_laneid();
   int b = __nvvm_read_ptx_sreg_warpid();
   int c = __nvvm_read_ptx_sreg_nwarpid();
   int d = __nvvm_read_ptx_sreg_smid();
   int e = __nvvm_read_ptx_sreg_nsmid();
   int f = __nvvm_read_ptx_sreg_gridid();
+  int g = __nvvm_read_ptx_sreg_warpsize();
 
-  return a + b + c + d + e + f;
+  return a + b + c + d + e + f + g;
 
 }