[clang] [llvm] [clang][NVPTX] Add support for scaled::n1::ue8m0 in FP8, FP6 and FP4 conversions (PR #227652)

[Durgadoss R via cfe-commits]

================
@@ -2974,11 +2974,11 @@ def : Pat<(int_nvvm_ui2f_rp i32:$a), (CVT_f32_u32 $a, CvtRP)>;
 
 // Utility to look up base and satfinite variants of a conversion intrinsic.
 // Missing variants are left unset so this can be used when only one exists.
-class ToCvtIntrinsics<string suffix> {
-  Intrinsic base = !if(!exists<Intrinsic>("int_nvvm_" # suffix),
-                       !cast<Intrinsic>("int_nvvm_" # suffix), ?);
-  Intrinsic sf = !if(!exists<Intrinsic>("int_nvvm_" # suffix # "_satfinite"),
-                     !cast<Intrinsic>("int_nvvm_" # suffix # "_satfinite"), ?);
+class ToCvtIntrinsics<string suffix, string tail = ""> {
+  Intrinsic base = !if(!exists<Intrinsic>("int_nvvm_" # suffix # tail),
----------------
durga4github wrote:

nit: can we do the concat once for both the cases?

https://github.com/llvm/llvm-project/pull/227652