[PATCH] D59703: Optimize masked.loads and masked.gathers with a single active lane

[Philip Reames via Phabricator via llvm-commits]

reames created this revision.
reames added a reviewer: craig.topper.
Herald added subscribers: bollu, mcrosier.

If we've proven only a single lane is active in a gather, we can convert it to either a <1 x Y> masked.load, or even a plain load if the load is safe to speculate.  See D57140 <https://reviews.llvm.org/D57140> for broader context.


https://reviews.llvm.org/D59703

Files:
  lib/Transforms/InstCombine/InstCombineCalls.cpp
  test/Transforms/InstCombine/masked_intrinsics.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D59703.191900.patch
Type: text/x-patch
Size: 6584 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190322/96f2d752/attachment.bin>