[PATCH] Loop Versioning for LICM

[silviu.baranga at arm.com]

In http://reviews.llvm.org/D9151#176258, @ashutosh.nema wrote:

> Didn’t understood this point completely, which structure you mentioning about ?
>  We are already using LAA(Loop access analysis) as the part of loop versioning.
>  It’s used for feasibility checking and memcheck generation.


I was thinking about the control flow generated by  InnerLoopVectorizer::createEmptyLoop() (see ASCII art in LoopVectorize.cpp) to execute either the vectorized version or the original version of the loop, which seems to be a more complex version of what is being added here. The vectorizer needs to be able to execute some iterations in the vector loop and the execute the scalar remainder, which is why it needs to be more complex.

Does it sound sensible?

Thanks,
Silviu


REPOSITORY
  rL LLVM

http://reviews.llvm.org/D9151

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/