[PATCH] D13447: [AArch64]Fix bug in DAG combine for ADDV
James Molloy via llvm-commits
llvm-commits at lists.llvm.org
Mon Oct 5 13:29:30 PDT 2015
jmolloy requested changes to this revision.
jmolloy added a comment.
This revision now requires changes to proceed.
Hi Jun,
I don't think this is the right fix. We generate oversized vectors in the loop vectorizer, and we expect them to work well. The right way to solve this PR is by implementing vector splitting for the [US]ADDV and [SU][MIN,MAX]V nodes.
Cheers,
James
http://reviews.llvm.org/D13447
More information about the llvm-commits
mailing list