[llvm] [VPlan] Fix `pseudoprobe` being treated a vector intrinsic (PR #185347)

[David Sherwood via llvm-commits]

================
@@ -92,7 +92,13 @@ bool VPlanTransforms::tryToConvertVPInstructionsToVPRecipes(
                                            Ingredient.getDebugLoc());
         } else if (CallInst *CI = dyn_cast<CallInst>(Inst)) {
           Intrinsic::ID VectorID = getVectorIntrinsicIDForCall(CI, &TLI);
-          if (VectorID == Intrinsic::not_intrinsic)
+          if (VectorID == Intrinsic::not_intrinsic ||
----------------
david-arm wrote:

Instead of this change, if you add the pseudoprobe intrinsic to the list in `LoopVectorizationCostModel::collectLoopUniforms` does this also fix your test? The list is here:

```
  // Scan the loop for instructions which are either a) known to have only
  // lane 0 demanded or b) are uses which demand only lane 0 of their operand.
  for (auto *BB : TheLoop->blocks())
    for (auto &I : *BB) {
      if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(&I)) {
        switch (II->getIntrinsicID()) {
        case Intrinsic::sideeffect:
        case Intrinsic::experimental_noalias_scope_decl:
        case Intrinsic::assume:
        case Intrinsic::lifetime_start:
        case Intrinsic::lifetime_end:
          if (TheLoop->hasLoopInvariantOperands(&I))
            AddToWorklistIfAllowed(&I);
          break;
        default:
          break;
        }
      }
```

I think it probably needs adding to the list in `VPRecipeBuilder::handleReplication` as well for the scalable vectors case, e.g.

```
  if (!IsUniform && Range.Start.isScalable() && isa<IntrinsicInst>(I)) {
    switch (cast<IntrinsicInst>(I)->getIntrinsicID()) {
    case Intrinsic::assume:
    case Intrinsic::lifetime_start:
    case Intrinsic::lifetime_end:
```

https://github.com/llvm/llvm-project/pull/185347