Re: Performance again

[Joseph Curtis <jc wasatch ps uci edu>]

Another set of data points (note the DEC compiler _key point_):

using only DEC cc: on 333MHz Alpha

(limantour)home/jc% time perftest i
mixed int
0.182u 0.004s 0:00.19 94.7%     0+0k 0+0io 0pf+0w
(limantour)home/jc% time perftest f
float
0.649u 0.003s 0:00.65 98.4%     0+0k 0+0io 0pf+0w
(limantour)home/jc% time perftest d
double
0.391u 0.005s 0:00.39 100.0%    0+0k 0+1io 0pf+0w

using gcc: on 500MHz PC164

(wasatch)/home/jc/% gcc preftest.c -o perftest
(wasatch)/home/jc/% time perftest i
mixed int
2.939u 0.003s 0:06.13 47.7%     0+0k 0+0io 67pf+0w
(wasatch)/home/jc/% time perftest f
float
0.906u 0.003s 0:02.07 43.4%     0+0k 0+0io 67pf+0w
(wasatch)/home/jc/% time perftest d
double
1.279u 0.004s 0:02.50 50.8%     0+0k 0+0io 67pf+0w

using DEC cc (-non_shared) on 500MHz PC164

(wasatch)/home/jc/% time perftest_ns i
mixed int
0.122u 0.000s 0:00.51 23.5%     0+0k 0+0io 15pf+0w
(wasatch)/home/jc/% time perftest_ns f
float
0.432u 0.001s 0:00.99 43.4%     0+0k 0+0io 12pf+0w
(wasatch)/home/jc/% time perftest_ns d
double
0.261u 0.001s 0:00.46 56.5%     0+0k 0+0io 12pf+0w




*******************************************************************************

Joseph E. Curtis                      Office: 949-824-7473  
Department of Chemistry               Pager:  714-238-0557  
590 Physical Sciences I               Fax:    949-824-8571  
University of California              E-mail: curtisj@uci.edu  
Irvine, California 92697-2025         http://CX290227-A.msnv1.occa.home.com

*******************************************************************************





On Sun, 3 Jan 1999, Herwin Jan Steehouwer wrote:

> On Sun, 3 Jan 1999, geerten kuiper wrote:
> 
> > Herwin Jan,
> > 
> > Some ideas:
> > 
> > - as always, much depends on what your program is trying to do.
> > - make shure you use the improved math libraries (I think
> > <http://people.frankfurt.netsurf.de/Joachim.Wesner/libffm.0.28.tar.gz>
> > has the latest.)
> > - David Mosberger once wrote an article on optimizing code for Alpha. You
> > may be able to find it somewhere on the web. Cache optimization is the main
> > thing to realize the Alphas full potential.
> 
> i cannot find him ;-( also his e-mail adres i found, was not working ;(
> 
> > - using 64 bit floats will probably be almost as fast as 32 bits (contrary
> > to your PentiumII, where it will be much slower.)
> 
> ok, will try that !
> 
> > - on some older Alpha boxes, memory bandwith limits performance. On your XL
> > this should be less of an issue (is it 128 or 256 bits wide ?)
> 
> 366Xl is 128 bits wide !
> 
> a test programm ( see attachment ) will give this output ! how can i speed
> this up ?? ( gcc -o preftest preftest.c ) NO optimizing for this test ! 
> 
> SGI 100 Mhz R4k:
>         time perftest i
>         7.359
>         time perftest f
>         3.534
>         time perftest d
>         4.942
> 
> Alpha 233 Mhz:
>         time perftest i
>         9.140
>         time perftest f
>         2.290u
>         time perftest d
>         3.050u
> 
> Intel P2 300Mhz:
>         time perftest i
>         1.955
>         time perftest f
>         1.555
>         time perftest d
>         2.614
> 
> 
> tanx
> HJ
> 
> Herwin Jan Steehouwer
>   herwinjs@palet.nl                 http://www.caiw.nl/~herwinjs
>   steehouwer@kde.org              KDE developer http://www.kde.org 
> 
> 
> -- 
> To unsubscribe: send e-mail to axp-list-request@redhat.com with
> 'unsubscribe' as the subject.  Do not send it to axp-list@redhat.com
>