This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH x86] Increase PARAM_MAX_COMPLETELY_PEELED_INSNS when branch is costly
- From: Evgeny Stupachenko <evstupac at gmail dot com>
- To: Jan Hubicka <hubicka at ucw dot cz>
- Cc: Uros Bizjak <ubizjak at gmail dot com>, Richard Biener <richard dot guenther at gmail dot com>, GCC Patches <gcc-patches at gcc dot gnu dot org>
- Date: Thu, 30 Oct 2014 20:53:58 +0300
- Subject: Re: [PATCH x86] Increase PARAM_MAX_COMPLETELY_PEELED_INSNS when branch is costly
- Authentication-results: sourceware.org; auth=none
- References: <CAOvf_xwBqY++PGRC_+1=rzHO18jvc+TP5mmR=PjYvGgm=63NuA at mail dot gmail dot com> <CAFiYyc1YjAYk8M8QMp=2Uo11SoQmZtcj-Eb=L-gxLyQ=6GTXmA at mail dot gmail dot com> <20141013122332 dot GA9404 at atrey dot karlin dot mff dot cuni dot cz> <CAOvf_xyG3f9uCTLr3D_Kxo_4i+FHcACzZvcUtojT=EbLxXBRPQ at mail dot gmail dot com> <CAOvf_xxiB3UNBU98YPVHxe3YEkxG7Tiyy9_RPU9synU+pSpcoA at mail dot gmail dot com> <CAFULd4b=3xyVyCP3YejCOtiEaKaiJY+spmGA7wPOwfhAYDZ-yw at mail dot gmail dot com> <20141030172731 dot GF17795 at kam dot mff dot cuni dot cz>
Yes the speed up is the same. However I'm testing only x86
performance. Potentially we can somehow hurt ARM or others
GCC already has the tuning enabled for rs6000,s390, spu.
On Thu, Oct 30, 2014 at 8:27 PM, Jan Hubicka <firstname.lastname@example.org> wrote:
>> On Tue, Oct 28, 2014 at 1:07 PM, Evgeny Stupachenko <email@example.com> wrote:
>> > make check for gcc passed
>> > On Mon, Oct 27, 2014 at 11:10 AM, Evgeny Stupachenko <firstname.lastname@example.org> wrote:
>> >> The results are the same for Silvermont.
>> >> There are no significant changes on Haswell.
>> >> So I agree with Richard, let's enable this x86 wide.
>> >> Bootstrap/ passed.
>> >> Make check in progress.
>> >> Is it ok?
>> >> 2014-10-25 Evgeny Stupachenko <email@example.com>
>> >> * config/i386/i386.c (ix86_option_override_internal): Increase
>> >> PARAM_MAX_COMPLETELY_PEELED_INSNS.
>> Let's wait for Honza's approval ...
> Looking through the emails, it is not clear to me if you re-tested that this still
> makes the intended speedup with the tree-level loop peeling? (comitted 2014-10-14).
> If it still works as intended, I do not think we have any reason to not change the
> default in params.def given that even ARM folks are calling for peeling by default.