[VPlan] Record whether scalar IVs are need in induction recipe. (NFC)

[fhahn]

Summary

This explicitly records whether a scalar IV is needed in the
VPWidenIntOrFpInductionRecipe, to remove a dependence on the cost-model
during its ::execute.

It will also be used in D116123 to determine if a vector phi will be
generated.

[fhahn]

Rename to needsVectorIV

[Ayal]

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

2585–2586

While we're here, comment above is out of sync with code below.

8600

!NeedsScalarIV || !NeedsScalarIVOnly - suffice to have only !NeedsScalarIVOnly to feed NeedsVectorIV ?

Refactor the code that constructs a VPWidenIntOrFpInductionRecipe to be used by both tryToOptimizeInduction[PHI,Truncate]?

[fhahn]

Address latest comments, thanks! I unified the VPWidenIntOrFpInductionRecipe* creating in a single helper. Comment also updated.

[fhahn]

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

2585–2586

Updated, thanks!

8600

Good point, thanks! Looks like !NeedsScalarIVOnly is indeed sufficient. Updated and also moved to common createInductionRecipe helper!

[Ayal]

This is fine, couple of final nits; thanks for accommodating!

llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

2587

Suffice to ask if !Def->needsScalarIV() as that implies Def->needsVectorIV() - they cannot both be false if Def is not dead - but being explicit is clearer?

2595

Suffice to ask if Def->needsVectorIV() as we already know that Def->needsScalarIV() here - but being explicit is clearer?

8525

"create[Widen]InductionRecipe" ?

8547

Nice to see the above encapsulated here!

[fhahn]

Address latest comments, thanks! I am planning to land this soon.