[llvm] [NVPTX] Add float to tf32 conversion intrinsic (PR #121507)

[Alex MacLean via llvm-commits]

================
@@ -1466,6 +1466,15 @@ let TargetPrefix = "nvvm" in {
   def int_nvvm_e5m2x2_to_f16x2_rn_relu : ClangBuiltin<"__nvvm_e5m2x2_to_f16x2_rn_relu">,
       Intrinsic<[llvm_v2f16_ty], [llvm_i16_ty], [IntrNoMem, IntrNoCallback]>;
 
+// Convert Float to TF32
+def int_nvvm_cvt_float_to_tf32 : Intrinsic<[llvm_i32_ty],
+    [llvm_float_ty, // Input float
+     llvm_i8_ty,    // Flag for Rounding Modes
----------------
AlexMaclean wrote:

Can the rounding mode be expressed with metadata as is done with the existing constrained floating point intrinsic? Expressing enums with many values as ints can make things more difficult to read.

https://github.com/llvm/llvm-project/pull/121507