[llvm] [AMDGPU][wmma] - Add tied wmma intrinsic (PR #69903)

[Nicolai Hähnle via llvm-commits]

================
@@ -865,22 +865,26 @@ def WMMAOpcode3AddrMappingTable : WMMAMappingTable {
 //    it converts the default pseudo to the pseudo where src2 is not the same as vdst.
 // 3) @earlyclobber on the destination satisfies the constraint during RA.
 
-multiclass WMMAInst<string Suffix, string Instr, VOPProfile P, SDPatternOperator node = null_frag, RegisterOperand _Src01RC64 = VRegSrc_256, WMMAType Type> {
+multiclass WMMAInst<string Suffix, string Instr, VOPProfile P, SDPatternOperator node = null_frag, RegisterOperand _Src01RC64 = VRegSrc_256, WMMAType Type, bit convertibleTo3Addr> {
 
   defvar WMMAConstraints2Addr = "@earlyclobber $vdst,$vdst = $src2";
   defvar WMMAConstraints3Addr = "@earlyclobber $vdst";
 
   defvar WMMAProfile = VOPProfileWMMA<P, Suffix, _Src01RC64, Type.hasClamp, Type.hasOpsel>;
   let Mnemonic = Instr, mayRaiseFPException = 0, ReadsModeReg = 0 in {
-    let Constraints = WMMAConstraints2Addr, isConvertibleToThreeAddress = 1 in {
+    let Constraints = WMMAConstraints2Addr, isConvertibleToThreeAddress = convertibleTo3Addr in {
       def _twoaddr # Suffix : VOP3P_Pseudo<Instr # Suffix, WMMAProfile>;
     }
-    let Constraints = WMMAConstraints3Addr, SchedRW = [Write32Bit, Write32Bit] in {
-      def _threeaddr # Suffix : VOP3P_Pseudo<Instr # Suffix, WMMAProfile>;
-    }
   }
-  def : WMMAOpcodeMapping<!cast<Instruction>(NAME # _twoaddr # Suffix),
+  if convertibleTo3Addr then {
+    let Mnemonic = Instr, mayRaiseFPException = 0, ReadsModeReg = 0 in {
----------------
nhaehnle wrote:

I think that's explained by the implementation of `if` in TableGen being a bit of a hack :)

https://github.com/llvm/llvm-project/pull/69903