DhrystonePack

Short:        UN*X port (SAS adapted), incl. binaries
Author:       Torsten Hiddessen / math@sun.rz.tu-clausthal.de
Uploader:     Torsten Hiddessen / math sun rz tu-clausthal de
Type:         util/moni
Architecture: m68k-amigaos

                        
COPYRIGHT

 This  package is public-domain.  This means you can copy it freely as
long  as  you don't ask any money for it, except perhaps a nominal fee
for copying.


CONTENTS

 This  package  contains  the  well known 'Dhrystone' benchmark.  It's
available  everwhere  on  the  net  - but if you wish to compile it on
AMIGA,  you  need  to  alter  the  source a little bit.  The source is
adapted  to  take  advantage  of the 'timer.device' - NOTHING else has
been altered.


WHAT'S NEW

 I  discovered  even  more optimization-switches at the compiler - now
the executables are around 15% faster on my machine.
Release date: 05-Jul-1995


USAGE
 Just  run  your  favourite  version  (1.0,  1.1  or  2.1)  from  your
'processors  directory'  (68000+,  68020+, 68040+).  For each CPU, the
benchmark  runs  at  a  useful  (IMHO) loop count.  The excutables are
compiled under SAS/C 6.55.


INFO

 I  wished  to  have  the  standard  dhrystone  around - just from the
original  sources  and optimized for my processor.  So I could compare
our  workstations and a some PC's (under LINUX) to my A4000/040.  Here
are the results from DHRYSTONE 1.1:
                               ^^^

Machine:             Clock:   Result:  Compiler:
----------------------------------------------------------------------
DEC ALPHA AXP        150MHz   427000   CC (GNU/C rates 377000, only)
SUN SPARC 10          33MHz    83000   GNU/C
SUN SPARC 20          ? MHZ   135000   GNU/C
PC 486/DX33           33MHz    28000   GNU/C (under LINUX)
PC 486/DX66           66MHz    53000   ? (under OS/2)
PC 486/DX4/100       100MHz    83000   GNU/C (under LINUX)
A4000/40              30MHz    44000   SAS/C
A4000/60/Cyberstorm   60MHZ    90000   SAS/C

Notice,  that  I NEVER compiled the benchmark 'register-optimized',
because   this   benchmark   should   rate   the   speed   of   the
processor-memory interface (while performing lots of 'string-copy')
and not the capability of the CPU (especially at RISC-design) to
store runtime variables at dozens of registers.



GREETINGS

 to Steffi, Jerome, Mark, Stefan, Michael, Thomas, as usual..



If  this stuff makes you happy or if you want to make any comments (or
just mail your RESULTS - especially from 68060) then E-mail me at

   math@sun.rz.tu-clausthal.de
  
 or send snail-mail to

   Torsten Hiddessen
   Walther-Nernst-Str. 1
   38678 Clausthal-Zellerfeld
   Germany

 and have a look at my pages at the WWW

   http://www.rz.tu-clausthal.de/~math



--
o~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~o
{ Name:    Torsten Hiddessen            Author of the 40 Magic-Backgrounds  }
{ Address: Walther-Nernst-Str, 1        Riding on: A4000/40 (30MHz), 10MB,  }
{          37678 Clausthal.Zellerfeld      \|/     420MB HD, Mitsumi FX001D }
{          Germany                         @ @     and AKF 50    __         }
{~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~oOO-(_)-OOo~~\           / /         }
{ E-Mail:   math@sun.rz.tu-clausthal.de (preferred)  \     __  / /  AMIGA   }
{           hiddessen@vax.in.tu-clausthal.de          \    \ \/ /  POWER    }
{           thiddess@in.tu-clausthal.de               /     \_\/  rUlEz     }
{ WWW-Link: http://www.rz.tu-clausthal.de/~math/     /                      }
o~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~/~~~~~~~~~~~~~~~~~~~~~~~o