[patch] enabling vectorization by default at -O3

Dorit Nuzman DORIT@il.ibm.com
Sat Sep 15 02:24:00 GMT 2007


> Richard Guenther wrote:
>
> >>>  Should we turn on vectorization at -O3, or are we not quite there
yet?
>
> >> IMO, the biggest gain is the stability it will bring to the vectorizer
> >> and the incentive to keep tuning the cost model.
>
> Turning on a feature that will affect users should be about the users --
> not providing us incentives to do things. :-)
>
> However:
>
> >> I agree.  We should enable vectorization with -O3.  I also like the
idea
> >> of enabling vectorization at -O2 for a subset of cases as suggested.
>
> We all agree that the option is at least a wash, and that, really, we
> want it on at -O3.  So, it's OK with me to give it a try.  If, for some
> reason, that doesn't work out, we can always back down.  So, Dorit,
> patch approved.
>

I will wait for PR26449 to be fixed (as requested by Uros), and also I have
a fix to commit for a couple ICEs that came up today (one reported under
PR33373, and one that I discovered in my last Spec run).

thanks,

dorit

> Thanks,
>
> --
> Mark Mitchell
> CodeSourcery
> mark@codesourcery.com
> (650) 331-3385 x713



More information about the Gcc-patches mailing list