[llvm] [NVPTX] Add TMA Bulk Copy Intrinsics (PR #138679)

[Durgadoss R via llvm-commits]

================
@@ -516,6 +516,9 @@ class CpAsyncBulkStr<bit mc, bit ch> {
   string S2G = "cp.async.bulk.global.shared::cta.bulk_group"
                # !if(ch, ".L2::cache_hint", "");
 
+  // Shared to Global memory with bytemask
+  string S2G_BM = S2G # ".cp_mask";
+
----------------
durga4github wrote:

yes, the mask suffix here is rather short ;-)
Otherwise, we generally keep these strings in a class since the final instr name tends to be rather long.

I feel using a bool for the mask and generating one S2G will be more concise and consistent with other uses here. I will try this out.

https://github.com/llvm/llvm-project/pull/138679