[PATCH] D22377: [SCEV] trip count calculation for loops with unknown stride

[Pankaj Chawla via llvm-commits]

pankajchawla added a comment.

Hi Sanjoy,

You example seems correct. Ideally, we should check the wrap flags on the phi associated with the IV but I think an alternative is to check that the stride is not known to be either positive or negative. The wrap flags cannot be added by ScalarEvolution for unknown strides. This is the case we were originally trying to fix anyway. If you agree, I will include this check in my next revision.

I will also make the changes you suggested in your inline comments.

Thanks,
Pankaj


https://reviews.llvm.org/D22377