Mailing-List: contact ooo-dev-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: ooo-dev@incubator.apache.org
Received-SPF: pass (athena.apache.org: local policy)
Mime-Version: 1.0 (Apple Message framework v752.3)
In-Reply-To: <20110906185329.GP11827@kulungile.erack.de>
References: <4E5F433E.6010908@sb-software.de>
 <4E5F5428.8020807@sb-software.de> <4E5F5911.9060808@sb-software.de>
 <4E5F7CCC.20306@sb-software.de> <20110904155555.GC22507@kulungile.erack.de>
 <4E646DB0.4020404@sb-software.de>
 <CAO2PO-hG_syvsJHCa=KaZQYKJeJhX1PHA845nZJcjNOyABcutw@mail.gmail.com>
 <CAP-ksoisb5Doc+X1RJtWrwW4ZPJ_q-_1CkPENZmZQ8HP-pAjzA@mail.gmail.com>
 <CAO2PO-i7az1abLQx9=1mXcF=aBdso+1KBxWYHOyHTGz6OWizGw@mail.gmail.com>
 <9675CC47-3B9E-411C-8C47-7B7CB39189A0@free.fr>
 <20110906185329.GP11827@kulungile.erack.de>
Content-Type: multipart/alternative; boundary=Apple-Mail-3--124013150
Message-Id: <7FA8D640-CD2A-437F-9387-EB3E2A59F51F@free.fr>
From: eric b <eric.bachard@free.fr>
Subject: Re: Who wants to build OpenOffice?
Date: Tue, 6 Sep 2011 21:41:59 +0200
To: ooo-dev@incubator.apache.org

--Apple-Mail-3--124013150
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain;
	charset=UTF-8;
	delsp=yes;
	format=flowed


Le 6 sept. 11 =C3=A0 20:53, Eike Rathke a =C3=A9crit :

> Hi eric,
>

Hi Eike,


>> cd instsetoo_native
>> build --all -P6 -- -P6 -- -P6 -- -P6
>
> Umm.. the first -P6 for build means build 6 directories found in */=20
> prj/build.lst in parallel, the second -P6 instructs dmake to spawn
> 6 processes for targets found in each makefile.mk, so already build =20=

> -P6 -- -P6
>

I agree, that's a lot, but the machine is powerfull. What I =20
observed:   -- separates every dual core, and the 6 processes are =20
balanced on every core. The result is around 100% for every, using =20
6.  5 gives something (average value) less than 100%.

It works well, but maybe I could optimize -> I'll test other values =20
and I'll tell you (I remember, some times ago, Pavel wrote a very =20
interesting blog entry about parallel builds)


> could create up to 36 processes, if for each directory triggered by =20=

> build there are at least 6 independent makefile.mk targets.
>

Sometimes,  s/6/5/ gives better results (I had some breakages with 4x6)


>> Explanation : -- to separate every core, 6 tasks per core. Don't
>> forget to use upper case (else it won't work)
>
> Sure that repeating  -- -P6  has any effect? Never tried that. If so,

I used the gnome tool, showing the 8 cores at around 100% most of the =20=

time.  If this can help, I observed that helpcontent2 is a =20
bottleneck, because some jobs seems to block the other.

On the memory side, I see a top of used ram around 5GB when building =20
sc / sw and sd, but never the full use of the 6GB, so.

> that would probably add to the 36 ...
>


No, it works. FYI, the machine is a i7 920 +  6GB  of ram / 256 GB =20
SSD hard disk, really powerfull. I do build OOo4Kids for 8 locales in =20=

50 min (cold ccache), and 28 min (hot ccache) this way.

Extremely easy for testing a bad commit breaking the build or when =20
unsure .. and so on.


> With a quad core I'm using  build -P3 -- -P2  in the background to =20
> keep my machine responsive for other tasks, or  build -P4 -- -P3  =20
> for a bit
> heavier load.
>


Exact : the load is not virtual, I confirm, but since you told me, =20
I'll try other values.


Regards,
Eric

--=20
q=C9=94=E1=B4=89=C9=B9=C9=99
Education Project:
http://wiki.services.openoffice.org/wiki/Education_Project
Projet OOo4Kids : http://wiki.ooo4kids.org/index.php/Main_Page
L'association EducOOo : http://www.educoo.org
Blog : http://eric.bachard.org/news


--Apple-Mail-3--124013150--