[PATCH] [SLPVectorizer] Reorder operands of shufflevector if it can result in a vectorized code.

Karthik Bhat kv.bhat at samsung.com
Wed Jan 14 23:35:06 PST 2015


Hi Michael,
Thanks for the reveiw. Addressed all review comments.
I'm ok with the current way we are handling reorderInputsAccordingToOpcode. If we were to move the load matching before checking AllSameOpcodeRight  we would need to maintain a flag so that we do not reorder operands that were swapped during consecutive load matching.(This would end up similar to the initial version of the patch that was uploaded with  needsReordering flag). I do not see much difference in the 2 approaches. I hope we can keep this as it is?

Please let me know if i can go ahead and commit this into mainline or if you have any other comments.
Thanks a lot.

Regards
Karthik Bhat


REPOSITORY
  rL LLVM

http://reviews.llvm.org/D6677

Files:
  lib/Transforms/Vectorize/SLPVectorizer.cpp
  test/Transforms/SLPVectorizer/X86/addsub.ll
  test/Transforms/SLPVectorizer/X86/operandorder.ll

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D6677.18214.patch
Type: text/x-patch
Size: 23948 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150115/77c37c8f/attachment.bin>


More information about the llvm-commits mailing list