Index of /fc4bench

[ICO]NameLast modifiedSizeDescription

[DIR]Parent Directory  -  
[TXT]README.html14-Dec-2009 14:22 2.0K 
[   ]ab.out18-Nov-2007 23:09 1.3K 
[   ]dh4.tar.gz18-Nov-2007 23:09 43K 

I have a 3 step approach to benching FC4. I run 32 Bit on AMD64. This seems like heresy, but my tests 
show that there is no loss of speed.

1) DH4.tar.gz ( is used to workout the beefed up Perl install by creating large DH keys. 
Without a well configured environment this code will take many minutes to run. Done right, it takes 
only a few seconds. On the latest AMD Linux builds, the duration is approaching 1 second. On my slow 
Intel laptop it takes 20 seconds.

To run, you will need to configure the Perl install:

calling syntax:
#perl > myout.out
#tail -f myout.out //(to get the elapsed time)

Install all GMP related before going to CPAN and installing Crypt-DH
#yum install gmp-devel

To run this code you will need these modules
#yum install perl-Math-BigInt-GMP  (dries repo)

#perl -MCPAN -e 'install Crypt-DH'
repeat the above for both ...
	Digest-SHA (by Mark Shelor)

On my Intel boxes this takes from 8-13 seconds (with or w/o raid). On this AMD it takes 1.9 seconds.
When run on X86_64, I got 2.6 seconds.

This perl code can easily be run on Win32, where
its at least 3x slower. There you will need the libgmp3.dll (Sisyphus).

2) To feel the muscle of the LAMP install, I use ApacheBench (ab). In SuSE this is called (ab2).
See the file [ab.out]
I run this PHP script:


without including, the result is 482 Requests/s.

3) I honor the Chudnovsky brothers by building and running GMP-Chudnovsky.c to generate some
1 millions digits of PI.

use YUM to get GMP-devel. Download the latest gmp-chudnovsky.c
	#include "gmp.h"
	#include ‹gmp.h›

To build,
#gcc gmp-chudnovsky.c -o chudpi -lm -lgmp

Then to calculate lots of digits, do
#chudpi 1000000 329 > digits

On this AMD FC4 box, the elapsed cpu time is 3.8 seconds. The total running 
time is around 20 seconds. The code only reports the used cpu clock.

October 27, 2005