Re: [Nautilus-list] Nautilus Performance Analysis
- From: Alan Cox <alan lxorguk ukuu org uk>
- To: alexl redhat com (Alex Larsson)
- Cc: darin bentspoon com (Darin Adler), nautilus-list lists eazel com, alan lxorguk ukuu org uk (Alan Cox)
- Subject: Re: [Nautilus-list] Nautilus Performance Analysis
- Date: Wed, 15 Aug 2001 00:48:49 +0100 (BST)
Ok I've been refining my tools a little here are some more stats
(main thread)
Time spent executing *stat64() 1.93 seconds (64498 *stat64 calls made)
Time spent executing write() syscalls 0.97 seconds (983)
Time spent executing read() syscalls 0.90 seconds (14628)
Time spent executing open() syscalls 0.30 seconds (6506)
Time spent executing mmap() syscalls 0.14 seconds (5994)
Time spent executing munmap() syscalls 0.14 seconds (5848)
Also each font loading cycle takes about 0.001 seconds timed from the
lstat64 to the munmap. We load the font 4800 times, so the cost is about
5 seconds
The main thread spends about 7.5 seconds waiting for things to happen (in
poll or select) adjusting for the time to terminate/kill it
Alex did a new eel with the font load fixed. It seems to be about 4 seconds
faster. I suspect the font problem is actually related to number of icons
so my 1700 made it hurt.
This time it spent 2.14 seconds executing 66753 calls to stat. There seem
to be more complex interactions at work. since I haven't added lots of files
in the meantime
The number of repeats of each icon load have gone from 46 to 48 (and the
corresponding i-regular.xml).
What mechanism governs the way nautilus sees things like thumbnail updates ?
Alan
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]