This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Introduce 4-stages profiledbootstrap to get a better profile.
- From: Jan Hubicka <hubicka at ucw dot cz>
- To: Martin Liška <mliska at suse dot cz>
- Cc: Markus Trippelsdorf <markus at trippelsdorf dot de>, GCC Patches <gcc-patches at gcc dot gnu dot org>, Richard Biener <richard dot guenther at gmail dot com>
- Date: Mon, 19 Jun 2017 12:37:13 +0200
- Subject: Re: [PATCH] Introduce 4-stages profiledbootstrap to get a better profile.
- Authentication-results: sourceware.org; auth=none
- References: <f61115af-55e1-8d59-545f-295f7e5b53cb@suse.cz> <20170529050407.GA15061@x4> <9203f3a8-3aa4-292b-1f2d-6749d3e6d0f7@suse.cz>
> On 05/29/2017 07:04 AM, Markus Trippelsdorf wrote:
> > On 2017.05.25 at 11:55 +0200, Martin Liška wrote:
> >> Hi.
> >>
> >> As I spoke about the PGO with Honza and Richi, current 3-stage is not ideal for following
> >> 2 reasons:
> >>
> >> 1) stageprofile compiler is train just on libraries that are built during stage2
> >> 2) apart from that, as the compiler is also used to build the final compiler, profile
> >> is being updated during the build. So the stage2 compiler is making different decisions.
> >>
> >> Both problems can be resolved by adding another step in between current stage2 and stage3
> >> where we train stage2 compiler by building compiler with default options.
> >
> > Another issue that I've noticed is that LTO doesn't get used in the
> > final stage (stagefeedback) with "bootstrap-O3 bootstrap-lto".
> > It only is used during training. So either move -flto to stagefeedback,
> > or use -flto both during training and during the final stage.
> >
>
> You're right, thus I'm suggesting to use -flto in all stages of PGO if 'make profiledbootstrap'
> is invoked.
Thanks, this makes sense to me. I wonder about quality of training. One perofmrance
critical bit that does not get trained at all is LTO stremaing when you build non-LTO
compiler. I guess we should add someting for that as that seems to be rather common
case. Similarly -O0 path may need bit of care. But your changes seems improvement
over current state for sure.
What are the bootstrap times now?
Honza
>
> Martin
> >From 77fb3302ef98548d37bf0f891ff09bca297f77fa Mon Sep 17 00:00:00 2001
> From: marxin <mliska@suse.cz>
> Date: Tue, 6 Jun 2017 11:23:38 +0200
> Subject: [PATCH] Enable -flto in all PGO stages for
> bootstrap-lto-{,noplugin}.mk.
>
> config/ChangeLog:
>
> 2017-06-06 Martin Liska <mliska@suse.cz>
>
> * bootstrap-lto-noplugin.mk: Enable -flto in all PGO stages.
> * bootstrap-lto.mk: Likewise.
> ---
> config/bootstrap-lto-noplugin.mk | 4 +++-
> config/bootstrap-lto.mk | 4 +++-
> 2 files changed, 6 insertions(+), 2 deletions(-)
>
> diff --git a/config/bootstrap-lto-noplugin.mk b/config/bootstrap-lto-noplugin.mk
> index a5073365b5a..fc980d2bc17 100644
> --- a/config/bootstrap-lto-noplugin.mk
> +++ b/config/bootstrap-lto-noplugin.mk
> @@ -3,4 +3,6 @@
>
> STAGE2_CFLAGS += -flto=jobserver -frandom-seed=1 -ffat-lto-objects
> STAGE3_CFLAGS += -flto=jobserver -frandom-seed=1 -ffat-lto-objects
> -STAGEprofile_CFLAGS += -fno-lto
> +STAGEprofile_CFLAGS += -flto=jobserver -frandom-seed=1 -ffat-lto-objects
> +STAGEtrain_CFLAGS += -flto=jobserver -frandom-seed=1 -ffat-lto-objects
> +STAGEfeedback_CFLAGS += -flto=jobserver -frandom-seed=1 -ffat-lto-objects
> diff --git a/config/bootstrap-lto.mk b/config/bootstrap-lto.mk
> index 9e065e1d85a..50b86ef1c81 100644
> --- a/config/bootstrap-lto.mk
> +++ b/config/bootstrap-lto.mk
> @@ -2,7 +2,9 @@
>
> STAGE2_CFLAGS += -flto=jobserver -frandom-seed=1
> STAGE3_CFLAGS += -flto=jobserver -frandom-seed=1
> -STAGEprofile_CFLAGS += -fno-lto
> +STAGEprofile_CFLAGS += -flto=jobserver -frandom-seed=1
> +STAGEtrain_CFLAGS += -flto=jobserver -frandom-seed=1
> +STAGEfeedback_CFLAGS += -flto=jobserver -frandom-seed=1
>
> # assumes the host supports the linker plugin
> LTO_AR = $$r/$(HOST_SUBDIR)/prev-gcc/gcc-ar$(exeext) -B$$r/$(HOST_SUBDIR)/prev-gcc/
> --
> 2.13.0
>