User: Password:
|
|
Subscribe / Log in / New account

ELC: How much memory are applications really using?

ELC: How much memory are applications really using?

Posted May 31, 2007 18:31 UTC (Thu) by czr (guest, #13701)
Parent article: ELC: How much memory are applications really using?

It is somewhat funny how nowadays many people working with Linux seem to
reinvent the same wheels, albeit of different color and size :-).

Some time ago I wrote a tool that does reporting based on URES (unique
residest size, RRES - SHARED in this case, which is the same as USS). For
this to work, no kernel patching is required. And of course the number by
itself has the same limitations as the USS.

The tool (written in Python) can be found here:
http://koltsoff.com/pub/meminfo . And another page explaining URES is
mentioned there as well.

The idea of URES/USS has been implemented (at least in the VCS-version)
of KDE ksysguard and based on short email communication will not be
implemented in gnome-system-monitor.

When dealing with systems that are running X server(s), it also is useful
to know that GUI programs often use bitmaps (and other resources) that
are actually accounted in X server at some point, and this will lead to
somewhat skewed memory usage results. This will also apply to any
client-server system in which part of client resources are allocated (and
present in memory of) the server.


(Log in to post comments)

ELC: How much memory are applications really using?

Posted Jan 5, 2009 8:56 UTC (Mon) by Blaisorblade (guest, #25465) [Link]

> When dealing with systems that are running X server(s), it also is useful
to know that GUI programs often use bitmaps (and other resources) that
are actually accounted in X server at some point, and this will lead to
somewhat skewed memory usage results.

xrestop can give some stats about those resources. However, I used to have the gut feeling that the increase in memory usage of the X server caused by client programs is bigger than the one reported by xrestop.

For instance, right now ~130M (mostly for pixmaps, with 40M for Firefox with ~60 tabs in 4 windows) are allocated by client programs on my system, out of 300M of X's RSS, and 390M of anonymous data memory (i.e. mostly heap).

Time ago, I read on some blog about optimizations for OLPC that the funny thing is that those cached bitmaps are _uncompressed_ (yes, even for JPEGs), and that Firefox caches all bitmaps from all pages. That's a real pity, since decompressing them on the fly would be surely faster than swapping, and maybe even faster than loading them from memory (not sure on this, unless the graphic card supports loading JPEGs in memory).


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds