[PATCH] D30211: [LV] Merge floating point and integer induction widening code

[Matthew Simpson via Phabricator via llvm-commits]

mssimpso added inline comments.


================
Comment at: lib/Transforms/Vectorize/LoopVectorize.cpp:4741
     // Handle the pointer induction variable case.
     assert(P->getType()->isPointerTy() && "Unexpected type.");
     // This is the normalized GEP that starts counting at zero.
----------------
delena wrote:
> mssimpso wrote:
> > delena wrote:
> > > Is the pointer induction really different from int and FP induction?
> > > May be you can put everything under widenInduction?
> > It's not, and I'm planning to work on this soon. Thanks!
> I'd put them all in one patch and call the function widenInduction(). but it is not a stumbling-block anyway.
Would you mind if we did this in a separate patch? I was trying to focus this one on floating-point, while keeping everything thing else as NFC as possible. Once committed, we can make sure no performance issues crop up for anyone before moving on to the pointer induction variables. The suggestion makes sense - I just would like to keep things small for review and codgen investigations should the need arise.


https://reviews.llvm.org/D30211