[llvm] [DAGCombiner][VP] Add DAGCombine for VP_MUL (PR #80105)

[Shao-Ce SUN via llvm-commits]

================
@@ -26421,6 +26433,10 @@ SDValue DAGCombiner::visitVPOp(SDNode *N) {
       return visitFMA<VPMatchContext>(N);
     case ISD::VP_SELECT:
       return visitVP_SELECT(N);
+    case ISD::VP_MUL:
----------------
sunshaoce wrote:

I think this is a great PR for Loop Vectorize. We have too few optimizations for ISD::VP_* , similar optimizations are akin to this one. I tested optimizing VP_UDIV to VP_MUL on the RVV1.0 development board and indeed gained benefits.

https://github.com/llvm/llvm-project/pull/80105