[llvm] [LV] Thread an explicit CM param through VPlan-building/costing helpers (PR #224836)

[Kiran Chandramohan via llvm-commits]

================
@@ -896,15 +896,18 @@ class LoopVectorizationPlanner {
   /// A builder used to construct the current plan.
   VPBuilder Builder;
 
-  /// Computes the cost of \p Plan for vectorization factor \p VF.
+  /// Computes the cost of \p Plan for vectorization factor \p VF, using
+  /// \p EnabledCM as the cost model (the main loop's cost model, or a
+  /// separate one when costing a tail-folded epilogue).
   ///
   /// The current implementation requires access to the
   /// LoopVectorizationLegality to handle inductions and reductions, which is
   /// why it is kept separate from the VPlan-only cost infrastructure.
   ///
   /// TODO: Move to VPlan::cost once the use of LoopVectorizationLegality has
   /// been retired.
-  InstructionCost cost(VPlan &Plan, ElementCount VF, VPRegisterUsage *RU) const;
+  InstructionCost cost(VPlan &Plan, ElementCount VF, VPRegisterUsage *RU,
----------------
kiranchandramohan wrote:

Would grouping the cost-model and plans via a VPlanPlanningContext as in https://github.com/llvm/llvm-project/pull/226415 help?

```
  VPlanPlanningContext MainCtx(createMainCostModel());
  LVP.plan(MainCtx, UserVF, UserIC);
  auto [MainVF, MainPlan] = LVP.computeBestVF(MainCtx);

  VPlanPlanningContext EpilogueCtx(
      createEpilogueCostModel(..));
  LVP.plan(EpilogueCtx, EpilogueVF, EpilogueIC);
  LVP.computeBestVF(EpilogueCtx);
  ```

https://github.com/llvm/llvm-project/pull/224836