Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 7 Aug 2003 20:07:56 -0600
From:      Tillman <tillman@seekingfire.com>
To:        sparc64@freebsd.org
Subject:   Re: ponderous 'make world' times post GCC 3.3...
Message-ID:  <20030807200756.M16062@seekingfire.com>
In-Reply-To: <p052106c5bb58a760735b@[128.113.24.47]>; from drosih@rpi.edu on Thu, Aug 07, 2003 at 09:08:11PM -0400
References:  <20030807062536.GA68747@dragon.nuxi.com> <p052106c5bb58a760735b@[128.113.24.47]>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Aug 07, 2003 at 09:08:11PM -0400, Garance A Drosihn wrote:
> At 11:25 PM -0700 8/6/03, David O'Brien wrote:
> >Am I the only one that saw 'make world' go from almost 3 hours
> >with GCC 3.2 to:
> >
> >     --------------------------------------------------------------
> >     >>> make world completed on Wed Aug  6 20:49:47 PDT 2003
> >                        (started Wed Aug  6 09:49:30 PDT 2003)
> >     --------------------------------------------------------------
> >     11h17.00s real          9h29m42.85s user    1h15m22.05s sys
> >
> >post GCC 3.3?  This is a 500mhz Blade 100 with a GENERIC minus
> >WITNESS* kernel.
> 
> I haven't rebuilt my sparc in awhile.  I can do that this weekend
> though.  I certainly hope it hasn't become that much slower, because
> my sparc is slower than yours!

It's definitely longer. On an Ultra 5 my buildworld time is now
1068m1.246s and my buildkernel is 442m44.184s. Add those together and
you get more than 25 hours.

Relevent CPU details:
cpu0: Sun Microsystems UltraSparc-IIi Processor (360.00 MHz CPU)

(I beleive the version of the UltraSparc 360 in the Ultra 5 has the 256K
of cache rather than the 2M ... anyone know of a way to probe that from
within FreeBSD? `dmesg` doesn't seem to show it.)

I used to be able to start `make buildkernel && make -j4 buildworld`
before going to work and have it done by the time I got home, so figure
that at under 10 hours. I don't know for sure how long it took - it was
short enough that I didn't see any point in timing it :-)

The recent problems where -jX wouldn't work on buildworld have also had
an additive (though small) impact on the times, roughly two hours in my
case (slow drives, heh). I haven't had a chance to try a -jX buildworld
recently to see if they're working again (note that the times I gave are
from when they were working).

On the other hand, I haven't had /any/ problems with the new gcc (extra
warnings about inlining excepted), which is a good thing.

-T


-- 
Happiness is wanting what you get, NOT getting what you want.
	- Robert Heinlein



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030807200756.M16062>