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

[Artem Belevich via llvm-commits]

================
@@ -725,6 +728,12 @@ let hasSideEffects = false in {
 
   def CVT_f16x2_e4m3x2 : CVT_f16x2_fp8<"e4m3">;
   def CVT_f16x2_e5m2x2 : CVT_f16x2_fp8<"e5m2">;
+
+  // Float to TF32 conversions.
+  def CVT_tf32_f32 : NVPTXInst<(outs Int32Regs:$dst),
+                     (ins Float32Regs:$src, CvtMode:$mode),
+                     !strconcat("cvt${mode:base}${mode:relu}${mode:satfinite}.",
----------------
Artem-B wrote:

For this limited use case, instead of implementing a separate mode printer, you could just pass the necessary relu/satfinite modifier string for the instruction directly, as the record argument. E.g. instead of `CVT_tf32_f32 $a, CvtRN_RELU` you would use `CVT_tf32_f32 $a, ".relu"` and instead of `CvtRNA_SATFINITE` you'd pass `".rna.satfinite"`.

The extra mode here just complicates things without buying you anything.


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