[llvm] [LV] Optimise code created by createElementCount (PR #216981)

[Luke Lau via llvm-commits]

================
@@ -375,15 +375,16 @@ class VPBuilder {
 
   VPValue *createElementCount(Type *Ty, ElementCount EC) {
     VPlan &Plan = getPlan();
-    VPValue *RuntimeEC = Plan.getConstantInt(Ty, EC.getKnownMinValue());
     if (EC.isScalable()) {
       VPValue *VScale = createVScale(Ty);
-      RuntimeEC = EC.getKnownMinValue() == 1
-                      ? VScale
-                      : createOverflowingOp(Instruction::Mul,
-                                            {VScale, RuntimeEC}, {true, false});
+      if (EC.getKnownMinValue() == 1)
+        return VScale;
+      assert(isPowerOf2_32(EC.getKnownMinValue()));
----------------
lukel97 wrote:

I guess in practice createElementCount is only ever passed the VF, but in theory an ElementCount can have a non power-of-2 min value and we would diverge from IRBuilder::createElementCount here.

Would a more general approach that's more similar to IRBuilder be to do a peephole fold in VPBuilder::createOverflowingOp that folds a power-of-2 mul to a shl? Not sure what the blast radius of that is. 

https://github.com/llvm/llvm-project/pull/216981