[llvm] [RISCV][llvm] Support VFADD, VFSUB, VFMUL codegen for Zvfbfa (PR #170612)

[Brandon Wu via llvm-commits]

================
@@ -523,6 +523,71 @@ multiclass VPatConversionVF_WF_BF16<string intrinsic, string instruction,
   }
 }
 
+multiclass VPatBinaryFPSDNode_VV_VF_RM_BF16<SDPatternOperator vop,
----------------
4vtomat wrote:

Yeah, I actually thought about it and found I also defined new ones in previous patches so I just go with it lol
But I think you're right, we should not define a new one so that it's easier to maintain

https://github.com/llvm/llvm-project/pull/170612