[PATCH] [SLPVectorization] Vectorize Reduction Chain feeding into a 'return' statement

[James Molloy]

Hi,

Could you also please check the code generated when a smaller-than-32bit
type is used? Historically we've had really poor codegen for these
reductions and this is the thing most likely to me to cause regressions if
there are any.

Cheers,

James
On Wed, 12 Nov 2014 at 17:14 Renato Golin <renato.golin at linaro.org> wrote:

> On 12 November 2014 16:41, Arnold Schwaighofer <aschwaighofer at apple.com>
> wrote:
> > I see this as an intermediate step that will allow us to catch certain
> > patterns. If performance tests show it is beneficial why not take it?
> >
> > Enabling horizontal reductions is a second step that might subsume this
> > logic.
>
> Would this code be re-used in that case? Or catch patterns that this
> one doesn't?
>
> If not, than I suggest we at least put a FIXME to explain the situation...
>
> cheers,
> -renato
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20141112/88d2ae8c/attachment.html>