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

Re: Memory requirements in Mathematica 6

• To: mathgroup at smc.vnet.net
• Subject: [mg83024] Re: Memory requirements in Mathematica 6
• From: Szabolcs Horvát <szhorvat at gmail.com>
• Date: Thu, 8 Nov 2007 06:03:54 -0500 (EST)
• References: <fgs8h1$48d$1@smc.vnet.net>

Michael Weyrauch wrote:
> Hello,
> 
>    some years back we solved a problem using Mathematica 3 using
> essentially a simple one line algorithm, which however used up nearly all the
> 4 GB of memory on our Unix workstation. But it worked...
> 
> Now with Mathematica 6 this very same program on a Linux 64 bit
> workstation with 32 GB memory does not run any more. The memory the
> Mathematica kernel demands just exceeds what is available, and the kernel quits (or worse, it
> shuts down the entire machine and we are in for a reboot. The latter effect just
> may be poor process  management on the side of our Linux kernel, which is not the issue here.)
> 
> I just wonder if it is really possible that the internal storage requirements of Mathematica 6
> in order to store expressions have so drastically increased (relative to version 3) that
> the effect described above can be explained. Does anyone have similar observations
> with  Mathematica 6?  Any other ideas or suggestions?  (Of course, tricks like Share[] have already been tried.)
> 

What is that algorithm doing and what kind of data structures is it 
using?  If it is only a single line, could you post it (preferably with 
comments)?  Mathematica 6 *does* fill up the memory more quickly than 
previous versions, but I have thought that this was mainly because of 
the plotting commands.  In some cases (arrays of real numbers), memory 
usage should have /de/creased significantly since version 3.

Szabolcs