Message Boards Message Boards


Why does the WolframKernel exit on OS X without a trace during Integrate?

Posted 1 year ago
5 Replies
2 Total Likes

I am trying to calculate a quantity from its density in 3D over a rectangular 2D window frame the following way:

mflxx = Integrate[
  mi3ret[[1, 1]], {y, -w/2, 
   w/2}, {z, -(c/2 + zf + d/2), -(c/2 + zf + d/2 + h)}]

where mi3ret[[1,1]] is a fairly long expression containing only elementary functions of the x,y and z variables, and it is the x-component of a 3D vector function. It also contains some constant parameters, like a,b,c. zf,d,h are additional constants for the definite integral. After a long calculation the WolframKernel is killed by the OS without any message or CrashReport. So, after rebooting the machine and running just Mathematica and the Activity Monitor, I made some screenshuts to see, how the memory is used and put the memory related value into lists. See notebook attached, where the 0 minute is really after about one and half hours into the run. The machine is an early 2011 Early MacPro with 64GB memory and a 500GB Flashdrive partitioned into two 250GB partition. So, the questions are: 1. Why the WolframKernel is killed without a trace? 2. Can the OS tuned in such a way that Integrate can finish its run? 3 Is there a command or function in Mathematica that would alert to the coming doom and would allow the examination of the status of the machine, including the status of the WolframKernel? 4. Is it possible to configure Mathematica to use, for disk based intermediate storage, the partition that have more free space, /the boot drive itself has 150GB, the other partition has 230GB free space/ ? Thanks ahead, János

5 Replies

Might be a kernel problem involving a failure to properly deal with the result of an internal interrupt. Integrate uses TimeConstrained in many places to avoid hanging when e.g. trying to sort out provisos (conditions for which one has convergence, or a result might hold), attempting to take limits, etc.

Posted 1 year ago

Should not be an option for the user to disable all TimeConstrained, especially if the user expecting an Integrate that lasts for hours or even overnight ?

Variations of that have been discussed in past but never implemented.Perhaps just as well, since many integrals with parameters, and even some without, would just wander off to Never-never Land. I should add that this is more or less incidental to the crash issue, which is just a bug (at least if the cause is what I suspect and described).

Posted 1 year ago

Thanks Mr. Lichtblau. Next Monday I can upload an OS X panic file I sent to Apple if that helps catching the bug.

A notebook or .m file with the input would be more useful in terms of diagnosing the issue.

Reply to this discussion
Community posts can be styled and formatted using the Markdown syntax.
Reply Preview
or Discard

Group Abstract Group Abstract