[PATCH][X86] Add target combine rules for horizontal add/sub.

Andrea Di Biagio andrea.dibiagio at gmail.com
Tue Jun 10 09:57:56 PDT 2014

Hi Nadav,

I just committed revision 210545 to fix a bug introduced by my
original patch committed at revision 210477.
I spotted this bug when running some extra tests to see if the
algorithm was able to identify all the interesting cases.

As described in the commit log message for revision 210545, the old
algorithm wrongly matched the 256-bit versions of horizontal add/sub
(on AVX and AVX2).

Sorry for the mistake.
Since the change was small, I decided to commit it immediately. I hope
it was ok to do it. If not, then please let me know so that I revert


On Mon, Jun 9, 2014 at 6:03 PM, Andrea Di Biagio
<andrea.dibiagio at gmail.com> wrote:
> Thanks for the review!
> Committed revision 210477.
> On Mon, Jun 9, 2014 at 5:20 PM, Nadav Rotem <nrotem at apple.com> wrote:
>> Thanks for working on this Andrea.  The patch LGTM.
>> On Jun 6, 2014, at 12:50 PM, Andrea Di Biagio <andrea.dibiagio at gmail.com> wrote:

More information about the llvm-commits mailing list