[PATCH] D67977: [InstCombine] Don't apply SimplifyDemandedVectorElts to ConstantAggregateZero GEP indices

[Philip Reames via Phabricator via llvm-commits]

reames added a comment.

I'm fine with this landing as a workaround, but it really feels like the right fix is on the other transform.  <0, undef> is effectively a zero index (or can be chosen as such), and our transforms should be aware of that.

More generally, I'm increasingly thinking the implicit broadcast feature of vector geps is a mistake.  I'm tempted to just change the LangRef and require a separate explicit broadcast via a shuffle.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D67977/new/

https://reviews.llvm.org/D67977