This is the mail archive of the gsl-discuss@sources.redhat.com mailing list for the GSL project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Some Speed Data of GSL Win32 Version


Hi:
After a whole day's work, I test GSL on two machines.
Machine A:	Red Linux 7.1
Machine B:	Windows 2000
		Windows 98
My test program is a matrix multiplication whose size is 500 x 500.
I cannot guarantee the two machine's Pentium CPU is the same, but at least
when I ran the same problem with SAS IML, both of them need around 3.1
seconds.
At Machine A in Red Linux, my program needs 0.8 seconds.
At Machine B in Windows 98, it needs more than 5 seconds with GSL VC
library.
If complied with GSL GCC Win32 library, it needs only 1.8 seconds in Dos
prompt.
Interestingly, if it ran in the CYGWin simulator, only 1.5 seconds are
needed.
At Machine B in Windows 2000, it needs 3.3 seconds with GSL VC library and
only 0.8 seconds with GCC Win32 library.
It seems VC and windows 98 are not suitable for numeric computing.
I hope my result is helpful for you.

Thank you again.

Liu, Li

-----Original Message-----
From: jochen@box.home.de [mailto:jochen@box.home.de]On Behalf Of Jochen
Küpper
Sent: Saturday, March 16, 2002 12:18 PM
To: Liu, Li
Cc: gsl-discuss@sources.redhat.com
Subject: Re: Speed Problem with GSL VC Version


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sat, 16 Mar 2002 10:49:38 -0600 Li Liu wrote:

Li> At Linux, it takes only 0.8 seconds, and SAS takes 3.1 seconds.
Li> At PC, SAS still takes 3.1 seconds, but my program takes 3.7
Li> seconds!  I already modified the result with the different CPU
Li> speeds of two machines.  Is there so huge difference for GSL
Li> between Unix and Win32?

Did you compile GSL on Windows with VC [1], too?  Their compilers suck
with respect to speed! [2] You can use the mingw toolchain on Windoze
and see what you get.

You also have to consider that MHz is not all you have to correct for,
cache size, memory throughput, and such might be more important.

Greetings,
Jochen

Footnotes:
[1]  I assume that is M$'s compiler.
[2]  Although one wouldn't expect that much of a difference. More in
     the 15--30% range in general, IIRC.
- --
Einigkeit und Recht und Freiheit                http://www.Jochen-Kuepper.de
    Liberté, Égalité, Fraternité                GnuPG key: 44BCCD8E
        Sex, drugs and rock-n-roll
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Processed by Mailcrypt and GnuPG <http://www.gnupg.org/>

iD8DBQE8k4xziJ/aUUS8zY4RAlQ1AJ9PRXGZmKra2UFmxCi6hPdmBUMaSACfbfRz
TP0i75Pg4yxcHhcbNsTlOcQ=
=CiGz
-----END PGP SIGNATURE-----


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]