[llvm] [VPlan] Speed up VPSlotTracker by using ModuleSlotTracker (PR #139881)

[David Sherwood via llvm-commits]

================
@@ -382,14 +383,25 @@ class VPSlotTracker {
   /// Number to assign to the next VPValue without underlying value.
   unsigned NextSlot = 0;
 
+  /// Cache slot indexes to avoid recomputing them on each printAsOperand call.
+  std::unique_ptr<ModuleSlotTracker> MST;
+
   void assignName(const VPValue *V);
   void assignNames(const VPlan &Plan);
   void assignNames(const VPBasicBlock *VPBB);
 
 public:
   VPSlotTracker(const VPlan *Plan = nullptr) {
-    if (Plan)
+    if (Plan) {
+      // This check is required to support unit tests with incomplete IR.
+      if (Function *F =
+              Plan->getScalarHeader()->getIRBasicBlock()->getParent()) {
+        Module *M = F->getParent();
+        MST = std::make_unique<ModuleSlotTracker>(M);
----------------
david-arm wrote:

It looks like we're invoking this from these functions:

```
raw_ostream &llvm::operator<<(raw_ostream &OS, const VPRecipeBase &R) {
  const VPBasicBlock *Parent = R.getParent();
  VPSlotTracker SlotTracker(Parent ? Parent->getPlan() : nullptr);
  R.print(OS, "", SlotTracker);
  return OS;
}

void VPValue::dump() const {
  const VPRecipeBase *Instr = dyn_cast_or_null<VPRecipeBase>(this->Def);
  VPSlotTracker SlotTracker(
      (Instr && Instr->getParent()) ? Instr->getParent()->getPlan() : nullptr);
  print(dbgs(), SlotTracker);
  dbgs() << "\n";
}

void VPDef::dump() const {
  const VPRecipeBase *Instr = dyn_cast_or_null<VPRecipeBase>(this);
  VPSlotTracker SlotTracker(
      (Instr && Instr->getParent()) ? Instr->getParent()->getPlan() : nullptr);
  print(dbgs(), "", SlotTracker);
  dbgs() << "\n";
}

void VPBlockBase::print(raw_ostream &O) const {
  VPSlotTracker SlotTracker(getPlan());
  print(O, "", SlotTracker);
}
```

Isn't this actually making the general case a lot slower? It looks like `incorporateFunction` could be quite expensive. Wouldn't it be faster to only create this once, perhaps by adding the ModuleSlotTracker to VPlan and guarded by `#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)`?

https://github.com/llvm/llvm-project/pull/139881