[clang] [llvm] [clang][NVPTX] Add missing half-precision add/mul/fma intrinsics (PR #170079)

[Alex MacLean via llvm-commits]

================
@@ -1490,16 +1490,25 @@ let TargetPrefix = "nvvm" in {
       def int_nvvm_fma_rn # ftz # variant # _f16x2 :
         PureIntrinsic<[llvm_v2f16_ty],
           [llvm_v2f16_ty, llvm_v2f16_ty, llvm_v2f16_ty]>;
-
-      def int_nvvm_fma_rn # ftz # variant # _bf16 : NVVMBuiltin,
-        PureIntrinsic<[llvm_bfloat_ty],
-          [llvm_bfloat_ty, llvm_bfloat_ty, llvm_bfloat_ty]>;
-
-      def int_nvvm_fma_rn # ftz # variant # _bf16x2 : NVVMBuiltin,
-        PureIntrinsic<[llvm_v2bf16_ty],
-          [llvm_v2bf16_ty, llvm_v2bf16_ty, llvm_v2bf16_ty]>;
     } // ftz
   } // variant
+  
+  foreach relu = ["", "_relu"] in { 
+    def int_nvvm_fma_rn # relu # _bf16 : NVVMBuiltin,
+      PureIntrinsic<[llvm_bfloat_ty],
+        [llvm_bfloat_ty, llvm_bfloat_ty, llvm_bfloat_ty]>;
+
+    def int_nvvm_fma_rn # relu # _bf16x2 : NVVMBuiltin,
+      PureIntrinsic<[llvm_v2bf16_ty],
+        [llvm_v2bf16_ty, llvm_v2bf16_ty, llvm_v2bf16_ty]>;
+  } // relu
+
+  // oob (out-of-bounds) - clamps the result to 0 if either of the operands is 
+  // OOB NaN value.
+  foreach relu = ["", "_relu"] in {
----------------
AlexMaclean wrote:

Nit: lets put these in the same `foreach relu =`

https://github.com/llvm/llvm-project/pull/170079