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

Re: parallel and remote computing

• To: mathgroup at smc.vnet.net
• Subject: [mg32462] Re: parallel and remote computing
• From: Mike Yukish <may106 at psu.edu>
• Date: Tue, 22 Jan 2002 03:19:30 -0500 (EST)
• Organization: Penn State University, Center for Academic Computing
• References: <a2ghs4$ke$1@smc.vnet.net>
• Sender: owner-wri-mathgroup at wolfram.com

Steve Gray wrote:

>     I have a calculation that takes so much time that my PC
> would have no chance of finishing it in a reasonable time
> (e.g one month running full tilt).
>     I have access to other PC's but they don't have Mathematica.
> Is there any way to compile a Mathematica program on my main PC and
> export the code so generated to the other PC's? They would
> need no interaction, but just write their results to a disk file.
>     Also other suggestions for speedup would help. I have
> already looked into algorithmic speedups and so far have
> acheived, through about five different means, a speedup of about
> 100x but am about out of ideas along that line.

The product Mathcode C++ supposedly can generate stand alone C++ code
from your Mathematica code, for a subset of Mathematica functions. (If
you call a function it doesn't support, it needs to call back to the
kernel.) If your routine fit within the bounds of what they can
translate, you'd have a means to run your code faster, and to spread it
around from machine to machine. Wolfram sells it on their site.