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

Re: Performance of student version of Mathematica?

• To: mathgroup at christensen.cybernetics.net
• Subject: [mg692] Re: Performance of student version of Mathematica?
• From: gwesp at cosy.sbg.ac.at (Gerhard Wesp)
• Date: 10 Apr 1995 13:54:58 GMT

In article <3m55ub$gm7 at news0.cybernetics.net>,
Kale Beckwitt <kaleb at uclink2.berkeley.edu> wrote:
>I am going to purchase the Windows version of Mathematica and I can get a 
>VERY good deal on the student version of the software through the store at
>the student store at my university. Unfortunately, this version does not
>utilize a math-co-processor. I was wondering if anyone has any experience
>with this version expecially regarding how this limitation effects software
>performance with large data sets.
>
[...]

I wouldn't recommend using Mma for *really* large data sets, 
unless I really had too. Mathematica is quite powerful in doing symbolic
math, but your problem is likely to be of numerical nature. Mma's data
structures are highly dynamic, this gives it a lot of flexibility, but
imposes also some performance limitations.
 However, if you do numerical calculations, you might experience 
a performance degradation of up to 90%, especially if it "doesn't utilize
a math-co-processor", as you say. 
 If you're doing linear algebra, have a look at octave, which is a frontend
to LAPACK. It will be much faster than *any* symbolic math
program. However, if you're doing symbmath only, the presence of a FPU
shouldn't make any difference.

Hope this helps!
-Gerhard