[15049] in Athena Bugs
Re: sun4 8.0J: X
daemon@ATHENA.MIT.EDU (Greg Hudson)
Wed Apr 2 03:53:38 1997
Date: Wed, 2 Apr 1997 03:53:30 -0500
From: Greg Hudson <ghudson@MIT.EDU>
To: danw@MIT.EDU
Cc: bugs@MIT.EDU
In-Reply-To: "[14894] in Athena Bugs"
> PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
> 28909 root 34 0 24328K 17484K sleep 0:04 0.89% 0.89% Xsun
After a call to Sun and some investigation, it appears that Xsun just
has some really big device memory mapping, and is only consuming about
2.5MB of core (where by "core" I mean physical memory plus swap space
on disk). Xsun uses about 1.8MB of core on a Sparc 5 at startup time,
but I don't think the 700K would explain any performance degradation
we might be seeing on Sparc 4s.
If you want to prove to yourself that ps's size figure includes
mappings which do not translate into actual allocated core, compile a
trivial program (say, one that does just "while (1);") with gcc, run
it, and look at the ps output. Then compile it again with -nostdlib,
and again run it and look at the ps output. Notice that there is a
difference of 476K, and that "size" run on libc and libdl totals 468K.
Sparc 4s do seem rather slow to me without a memory upgrade to 64MB,
but I haven't done any careful comparisons or come with any good
theories as to why.
