Group Abstract Group Abstract

Message Boards Message Boards

0
|
9.2K Views
|
4 Replies
|
0 Total Likes
View groups...
Share
Share this post:

Why does my MathKernel crash after long evaluations (10.3 on Linux)?

Posted 10 years ago

Hey everybody,

I have a huge problem with a project involving a numerical calculation in Mathematica 10.3 (but this problem is reproducible in all versions from 9.0 up to 10.3 that I could find on my machine). For the calculation I solve differential equations numerically for randomly chosen starting values. I do this for 10^5 samples and the estimated total running time of the simulation is ~2 days. The whole thing happens in a screen terminal, using "math -script <filename>". My problem is now that after a few hours (sometimes even after minutes) the kernel dies, without any error message. When I use subkernels to parallelize the whole thing, I only get messages that one of the kernel "appears to be dead". When I use a single thread it will just close the math kernel without further notice. Does anybody know where I might be able to find at least a reason why the kernels die? Some hidden log file maybe that I haven't found yet? Any help is appreciated, especially since Wolfram support is not answering my emails for some reason and this is affecting an important project.

Kindly, Matthias

POSTED BY: Matthias König
4 Replies
Posted 10 years ago

I'm sorry but usually in such situations it is impossible to help without a minimal code reproducing the problem. I understand that you probably have a very large code but please try to distill a minimal code snippet reproducing your problem.

POSTED BY: Alexey Popkov
Posted 10 years ago
POSTED BY: Matthias König

Do you monitor your RAM usage as the program runs? Programs as large as Mathematica frequently have memory leaks.

POSTED BY: Frank Kampas
Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
Attachments
Remove
or Discard