[PATCH] D26675: [LoopUnroll] Modify a comment to clarify the usage of TripCount. NFC.

[Michael Zolotukhin via Phabricator via llvm-commits]

mzolotukhin added a comment.

Hi Haicheng,

The comment is still confusing, at least to me. For instance, this sentence:

  ... Parameter TripCount
  may contain the exact number or the upper bound.

repeats what we've said in:

  ... In other words, control may exit the loop prior to TripCount 
  iterations via an early branch, but control may not exit the loop from the
  LatchBlock's terminator prior to TripCount iterations.

At least, it looks very much the same.

AFAIU, now we started to handle cases where the loop may exit from the LatchBlock's terminator prior to TripCount iterations - am I right here? If so, can we somehow reflect it in the comment (maybe reword it entirely)?

Thanks,
Michael


Repository:
  rL LLVM

https://reviews.llvm.org/D26675