autoconf
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: autoconf - threads [paraller] - multicore CPUs


From: Ralf Wildenhues
Subject: Re: autoconf - threads [paraller] - multicore CPUs
Date: Tue, 22 Jul 2008 20:05:58 +0200
User-agent: Mutt/1.5.18 (2008-05-17)

Hello little linux girl,

* litlle girl wrote on Tue, Jul 22, 2008 at 10:25:19AM CEST:
> Is there any way to make autoconf more paraller?

In the following I assume that you mean that configure scripts
should run their tests in parallel.

There have been ideas on how Autoconf could evolve that way,
but it's not simple to implement.  Realistically, it will
take some time yet, and also it will not be possible to achieve
a good speedup without some new test semantics which require
users to adjust their configure.ac files.  For example,
AC_CHECK_HEADERS_ONCE macros could all be run in parallel, but
AC_CHECK_HEADERS macros cannot, as the user may be relying on
the ordering.  The same holds for some other macros.

> Most people now, have multicore boxes,
> and autoconf is using only one core.

That's actually not completely true.  configure scripts benefit
a bit from a second core already, because they are I/O and fork-
intensive, and some kernel-side latency can be waited for in
parallel with other stuff that needs to be done.

I noted this effect when working on the parallel Autotest patches,
<http://thread.gmane.org/gmane.comp.sysutils.autoconf.patches/5563/focus=5570>.

> When I recompiling my system autoconf runs more than 100 times, it
> takes a lot of time.

Wait; do you mean that "autoconf" runs very often, or that "configure"
takes a long time to complete?  If the former, then it seems there is
something wrong with your setup; it should run at most once per package.

> It will be very good improvement make autoconf use multicore (at least
> dualcore).

Yes, that statement is undoubtedly true.  :-)

Cheers,
Ralf




reply via email to

[Prev in Thread] Current Thread [Next in Thread]