[PATCH] [Reassocate] Add support for vector instructions.

[Chad Rosier]

Hi spatel, chandlerc, resistor, hfinkel,

As the title suggests, this patch adds support for vector instructions in the reassociation pass.  This includes both integer and floating-point (with unsafe algebra) vector instructions.  This patch is a WIP, but I thought it was in a stable enough state to begin the review process.

If anyone has vector workloads, I would really like to know if they benefit from this patch.  I didn't see any real benefit on the EEMBC or SPEC suites.

 Regards,
  Chad

http://reviews.llvm.org/D5222

Files:
  lib/Transforms/Scalar/Reassociate.cpp
  test/Transforms/Reassociate/commute.ll
  test/Transforms/Reassociate/fast-AgressiveSubMove.ll
  test/Transforms/Reassociate/fast-ReassociateVector.ll
  test/Transforms/Reassociate/fast-fp-commute.ll
  test/Transforms/Reassociate/fp-commute.ll
  test/Transforms/Reassociate/multistep.ll
  test/Transforms/Reassociate/vector-basictests.ll
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D5222.13353.patch
Type: text/x-patch
Size: 39560 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140906/dee61b25/attachment.bin>