[AMD] Use Linear Layout convertions for AMDWmma

Enable LL conwertions for WMMA as well as for MFMA layouts.

See also: #5210

Signed-off-by: Ilya Veselov <[email protected]>

lib/Conversion/TritonGPUToLLVM/ConvertLayoutOpToLLVM.cpp

@@ -376,7 +376,7 @@ struct ConvertLayoutOpUsingLinearLayoutsConversion
     // completed before we can remove the layoutIsOK check:
     // 1. Support for AMD's WMMA
     std::function<bool(Attribute)> layoutIsOK = [&](Attribute layout) {
-      if (isa<NvidiaMmaEncodingAttr, AMDMfmaEncodingAttr>(layout)) {
+      if (isa<MmaEncodingTrait>(layout)) {
         return !useLegacyMMAConversion;
       }
       if (auto dotOperand = dyn_cast<DotOperandEncodingAttr>(layout)) {