Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 25 Apr 2009 02:30:00 +0400
From:      Stanislav Sedov <stas@FreeBSD.org>
To:        rea-fbsd@codelabs.ru
Cc:        x11@freebsd.org, obrien@freebsd.org, bug-followup@freebsd.org, miwi@FreeBSD.org
Subject:   Re: ports/132615: [patch] x11/libX11: work aroung parsing bug in Bash 4.0
Message-ID:  <20090425023000.7466392f.stas@FreeBSD.org>
In-Reply-To: <3vHRxK8zyisyJx9ZrK67n180LIY@veJXoSoJtqjaE%2Bn2qYrh3k99%2BAQ>
References:  <200903140956.n2E9uJ9G049103@freefall.freebsd.org> <CNRvQuQsWjPceEprCAhye6Gtlm4@gaXtT4o8WV/C0Q%2BqaYA3hARX3do> <3vHRxK8zyisyJx9ZrK67n180LIY@veJXoSoJtqjaE%2Bn2qYrh3k99%2BAQ>

next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 25 Apr 2009 00:03:20 +0400
Eygene Ryabinkin <rea-fbsd@codelabs.ru> mentioned:

> Gentlemen, good day.
> 
> Sun, Mar 15, 2009 at 10:20:17AM +0300, Eygene Ryabinkin wrote:
> > I had researched on the DOLT a little bit more and found that
> > doltcompile greatly speeds up the compilation process, because it was
> > written precisely for this purpose:
> >   http://lists.debian.org/debian-devel/2008/04/msg00286.html
> > 
> > So, I would say that it's better to patch the doltcompile script itself
> > rather than to drop support for it.
> 
> I had made some measurements for non-DOLT and DOLT case for the
> current libX11 case.  Here are the results of ministat spawned
> for data from 10 compilation cycles with DOLT and without DOLT:
> (fully warmed disk cache, almost zero activity on the machine,
> all resources are dedicated to the port building):
> -----
> x noDOLT.real
> + withDOLT.real
> +------------------------------------------------------------------------------+
> |       ++                                                                     |
> |       ++                                                          x          |
> |    +++++                           +                        x     xxx xx  xxx|
> ||______M_A_________|                                              |___A____|  |
> +------------------------------------------------------------------------------+
>     N           Min           Max        Median           Avg        Stddev
> x  10        177.48        189.25       184.365       184.366     3.5870173
> +  10        135.01        159.06       137.345       139.211     7.0399976
> Difference at 95.0% confidence
> 	-45.155 +/- 5.24948
> 	-24.492% +/- 2.84732%
> 	(Student's t, pooled s = 5.58696)
> -----
> I'd say that 45 seconds (24.5%) speedup worth some patching.
> 
> Any thoughts 'bout this?

I think we could try to replace the libtool by our own stripped-down
implementation that will be even simpler than dolt. If dolt is able to
reduce the compilations speed by about one forth, then with a FreeBSD
specific libtool replacement the imporvement may be even better.

As a benefit we can also stop installing .la and .a library files.

-- 
Stanislav Sedov
ST4096-RIPE

!DSPAM:49f23d69967001517610879!





Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20090425023000.7466392f.stas>