[llvm] [NVVM][NVPTX] Support decompress_b feature for tcgen05.mma intrinsics (PR #216312)

[Durgadoss R via llvm-commits]

================
@@ -6802,6 +6806,189 @@ foreach sparse = [0, 1] in {
   } // space
 } // sparse
 
+class Tcgen05MMADecompressBase<string ASpace, string Kind, int CtaGroup,
+                               string CollectorUsageA, string CollectorUsageB> :
+        Tcgen05MMABase</*IsSparse=*/ 0, ASpace, Kind, CtaGroup,
+                       CollectorUsageA, CollectorUsageB> {
+  let Predicates = [hasRubinFamilySupport];
+
+  let KindVal = !cond(
+                  !eq(Kind, "f8f6f4") : 0
+                );
----------------
durga4github wrote:

nit:
can all these not fit in one line?

https://github.com/llvm/llvm-project/pull/216312