[PATCH] D82418: [SLP] Limit GEP lists based on width of index computation.

[Sanjay Patel via Phabricator via llvm-commits]

spatel added inline comments.


================
Comment at: llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp:7505-7506
 
     // Process the GEP list in chunks suitable for the target's supported
     // vector size. If a vector register can't hold 1 element, we are done.
     unsigned MaxVecRegSize = R.getMaxVecRegSize();
----------------
fhahn wrote:
> spatel wrote:
> > We should update this comment with something like the text in this patch description:
> >   // We are trying to vectorize the index computations, so we base
> >   // the vectorization factor on the size of the loaded values rather
> >   // than the size of the GEP itself (the target's pointer size).
> > ?
> > 
> I've updated the comment as suggested, with slight edits to make it if better with the maximum number of elements below. WDYT? I'll commit it in a bit, unless there are further suggestions.
That text works for me - thanks! 


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D82418/new/

https://reviews.llvm.org/D82418