----- Original Message ----- 
From: "Matt Garman" <[EMAIL PROTECTED]>
To: "gentoo-user" <[EMAIL PROTECTED]>
Sent: Friday, November 21, 2003 5:41 PM
Subject: [gentoo-user] update: c++ performance: gentoo, debian, windows


>
> A while back I posted a message talking about the performance
> differences of a c++ program I wrote on gentoo and debian.  In the end,
> I chalked up the performance degrade on gentoo to my having built gcc
> with too many optimizations.  When I recompiled gcc with more
> "conservative" settings, my program ran as fast or faster on gentoo.
>
> If you recall, the processor/memory intensive part of my program is
> reading a 50,000+ line CSV file into memory.  On my gentoo box, it takes
> about 1.6 seconds to load this into memory.  On Windows, using MS Visual
> Studio 6.0, it takes five or six seconds to load into memory!
>
> I don't know how big a role computer speed plays, though.  My home
> computer (gentoo) is an Athlon XP 2500 with 1 GB of ram.  My work
> computer (windows) is a dual Xeon 1.5 GHz with 2 GB of ram.  Judging by
> Windows Task Manager, only one CPU is used to load the data.  So,
> loosely speaking, my home PC is 1.6 times faster, but the same process
> is 3.1 times slower (on my work PC).
>
> Again, this testing isn't scientific by any means, but I thought some
> folks might be interested.
>
> Looks good for gcc or AMD, in my opinion :)
> Matt
>


It probably has more too do with machine load, disk I/O speed (in particular
transfer speeds), memory speed and latency also will have a large impact, as
well as the size of disk cache (some IDE drives have 8MB buffers in them
which will likely do wonders on performance).  Also, it depends upon the
efficiency of your code and how it does the parsing.  A good parser for CSV
files should not be extremely CPU intensive as the bottleneck really should
be I/O.

Tom Veldhouse



--
[EMAIL PROTECTED] mailing list

Reply via email to