[PATCH] D41096: [X86] Initial support for prefer-vector-width function attribute

[Craig Topper via Phabricator via llvm-commits]

craig.topper added a comment.

@spatel based on my conversation with Chandler and Eric, I think they said we should have a pass before the vectorizers that detects any explicit vector code in the IR, If it finds any vector code, the pass would override the prefer-vector-width attribute to the length of the widest vector in that code.


https://reviews.llvm.org/D41096