[llvm] [VPlan] Impl VPlan-based pattern match for ExtendedRed and MulAccRed (PR #113903)

[Florian Hahn via llvm-commits]

================
@@ -9515,6 +9458,25 @@ LoopVectorizationPlanner::tryToBuildVPlanWithVPRecipes(VFRange &Range) {
   // Adjust the recipes for any inloop reductions.
   adjustRecipesForReductions(Plan, RecipeBuilder, Range.Start);
 
+  // Transform recipes to abstract recipes if it is leagal and beneficial and
+  // clamp the range for better cost estimation.
+  // TODO: Enable following transform when the EVL-version of extended-reduction
+  // and mulacc-reduction are implemented.
+  if (!CM.foldTailWithEVL()) {
+    VPCostContext CostCtx(CM.TTI, *CM.TLI, Legal->getWidestInductionType(), CM,
+                          CM.CostKind);
+    VPlanTransforms::runPass(VPlanTransforms::convertToAbstractRecipes, *Plan,
+                             CostCtx, Range);
+  }
+
+  // Update VF after convertToAbstractRecipes. Cannot set the VF here since
+  // `handleUncountableEarlyExit` will check the VF of the plan, need to set
+  // before it and update.
+  // TODO: Use a better method that only set the VF for plan once.
+  Plan->clearVF();
+  for (ElementCount VF : Range)
+    Plan->addVF(VF);
----------------
fhahn wrote:

Ah that's an unfortunate interaction. Let me see if we can break this up somehow. Maybe there's also a better way to handle the `handleUncountableEarlyExit` case.

In the meantime, can we remove the restricted VFs instead of clearing and re-adding?

https://github.com/llvm/llvm-project/pull/113903