Re: 2.2.STABLE5 Possible Memory Leak?

From: Jens-S. Voeckler <voeckler@dont-contact.us>
Date: Wed, 10 Nov 1999 17:39:25 +0100

On Wed, 10 Nov 1999, Jim Richey wrote:

]I do have priority_paging=1 set in /etc/system. What does the -mv8
]option due? Do you see much difference in performance between -02 and
]-03?

"-mv8" compiles SparcLite code AKA SuperSPARC. As todays' machines are
usually using UltraSPARC processors, e.g. "uname -m" says "sun4u", using
anything better than the default (which is -mv7) should improve on
pipeline stalls, register handling, etc, and all the nifty actions a RISC
CPU expects from its compiler. You might want to look into the gcc info
page -> invocation -> submodel options -> SPARC options.

BTW, the gcc 2.8 series seems to compile things o.k. when using
-mcpu=ultrasparc, but the 2.95 series (used on an E450/Sol7/32&64 bit
mode) seems to break the code, when using -mcpu=ultrasparc. Anybody some
lead on this?

]That acutally makes a lot of sense. Other OS's do this too. I bevieve
]top actually shows this on Linux. Unfortunately is does not on Solaris.
]I assume it's because it's not available in /proc. Anyone know if there
]is a utility available to show buffer cache in Solaris?

It should be available with the libkvm interface, but you might want to
have a look into "netstat -k" instead. Please don't ask me what all the
stuff in there is for...

Le deagh dhùrachd,
Dipl.-Ing. Jens-S. Vöckler (voeckler@rvs.uni-hannover.de)
Institute for Computer Networks and Distributed Systems
University of Hanover, Germany; +49 511 762 4726
Received on Wed Nov 10 1999 - 09:50:59 MST

This archive was generated by hypermail pre-2.1.9 : Wed Apr 09 2008 - 11:57:32 MDT