[clang] [llvm] [CUDA][NVPTX] Add dp4a/dp2a builtins; define CUDA-13.4's __nv_idp*_impl externs (PR #233434)

[Ilya Biryukov via cfe-commits]

================
@@ -1939,9 +1939,11 @@ let TargetPrefix = "nvvm" in {
   //
   foreach a_type = ["s", "u"] in {
     foreach b_type = ["s", "u"] in {
-      def int_nvvm_idp4a_ # a_type # _ # b_type :
+      def int_nvvm_idp4a_ # a_type # _ # b_type : NVVMBuiltin,
           PureIntrinsic<[llvm_i32_ty],
               [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty]>;
+      // The i1 operand selects the high (true) or low (false) half of the
+      // second operand. Exposed to clang as separate _lo/_hi builtins.
----------------
ilya-biryukov wrote:

Removed it from here.

https://github.com/llvm/llvm-project/pull/233434