MathGroup Archive 2000

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: can Mathematica use dual processors?

  • To: mathgroup at smc.vnet.net
  • Subject: [mg24579] Re: can Mathematica use dual processors?
  • From: Harald Giese <giese at ifm.uni-hamburg.de>
  • Date: Tue, 25 Jul 2000 00:56:11 -0400 (EDT)
  • Sender: owner-wri-mathgroup at wolfram.com

rob wrote:
> 
> I looked thru the postings and didn't find this subject.  Sorry if it
> has been discussed a lot and I missed it.
> 
> I'm wondering if having dual processor sysem will speed up mathematica
> v. 4.0?  Seems like several years ago this was discussed and Mathematica
> couldn't make use of the dual scheme.  That may have changed.
> 


Hi Rob,

There is no speed up on dual processor system. Even worse, some
computations may take longer! This happens with the benchmark test of
Stefan Steinhaus (www.scientificweb.de/testreport/mathtest3.nb):

Pentium II, 512MB RAM, Win. NT 4.0 (SP5), Mma. Version: 4.0.0

SINGLE CPU:

Creation,trans.,resh. of a matrix       24.359
Sorting of values in asc. order	        11.766
450001!                                 28.5, 
Computation of prime numbers            15.938
Computation of Fibonacci  numbers       33.281
"Creation of a Toeplitz matrix          27.094
Computation of Pi on 200000  decimals   18.469
Comp. of the inverse of a matrix        35.875
Comp. of Eigenvalues of matrix          58.781
Numerical comp. of a triple  integral   12.469
FFT over 4194304 random values          13.235
Polynomial regression                   67.25,
Comp. of a Laplace  transformation      26.141
Solving of the Euler-McLaurin sum.      1.64  
Comp. of 5 Koch Snowflakes              20.078
High resolution Klein  Bottle           63.219
Parametric plot of 3 pipes              34.782
Animation of a 3D Mountain in  heatcol. 22.328
                                              
Total timing:                           515.205


DUAL CPU:

Creation,trans.,resh. of a matrix       24.219 
Sorting of values in asc. order	        11.703 
450001!                                 28.453 
Computation of prime numbers            15.703 
Computation of Fibonacci  numbers       32.594 
"Creation of a Toeplitz matrix          26.563 
Computation of Pi on 200000  decimals   18.406 
Comp. of the inverse of a matrix        35.594 
Comp. of Eigenvalues of matrix          58.156 
Numerical comp. of a triple  integral   38.39  (!)
FFT over 4194304 random values          13.313 
Polynomial regression                   65.547 
Comp. of a Laplace  transformation      27.421 
Solving of the Euler-McLaurin sum.      1.563  
Comp. of 5 Koch Snowflakes              19.719 
High resolution Klein  Bottle           145.5  (!!)
Parametric plot of 3 pipes              34.234 
Animation of a 3D Mountain in  heatcol. 22.265 
                                               
Total timing:                           619.343



There is the "Parallel Computing Toolkit" written by R. Maeder
(www.mathdirect.com/products/par) available, but it costs some money
(USD 845/USD 445 Educational) and you have to call explicitly the
parallel functions (e.g. ParallelMap instead of Map).

Regards,
Harald

-- 

Harald Giese
Email: giese at ifm.uni-hamburg.de
Phone: +49 (0)40 42838 5796; Fax: +49 (0)40 5605724
Institut fuer Meereskunde der Universitaet Hamburg
(Institute of Oceanography of the University of Hamburg)
Troplowitzstrasse 7, D-22529 Hamburg


  • Prev by Date: Re: Speeding up Replacement Rules
  • Next by Date: interruptedWhile (useful for monitoring a computation non-interactively)
  • Previous by thread: Re: can Mathematica use dual processors?
  • Next by thread: Re: can Mathematica use dual processors?