MathGroup Archive 2009

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

Search the Archive

Re: Interpolation -> segfault?

  • To: mathgroup at smc.vnet.net
  • Subject: [mg98270] Re: Interpolation -> segfault?
  • From: oshaughn at gravity.psu.edu
  • Date: Fri, 3 Apr 2009 20:12:21 -0500 (EST)
  • References: <gqsn0r$3mf$1@smc.vnet.net> <gr0gil$95b$1@smc.vnet.net>

On Apr 1, 3:50 pm, Jens-Peer Kuska <ku... at informatik.uni-leipzig.de>
wrote:
> Hi,
>
> your InterpolationFunction[] cause the kernel crash and it would helpful
> to find out where it come from ..
>
> The input seems to be nonsense and it is no wonder that the kernel
> crashes.
>

Nonsensical?  It's a standard InterpolatingFunction object.  I don't
know what you mean.

The function is made by (1) using FunctionInterpolation on a very
complicated, slow-to-evaluate expression to create an object f1; (2)
wrapping it with
If[#1<... || #1> ..., 0, fn]; and then (3) using 'Save'.  That'sit.

The original function, before saving, does evaluate fine: I've used it
to create other data products (e.g., with NIntegrate).  Just hangs
when restoring it.

Also, I recently discovered that the function as above works fine if
you manually change only a few of the very small numbers (10^(-316) !)
to 0.   And, going back to the original code, adding 'Chop' before I
do the FunctionInterpolation yields an object that saves and restores
without trouble.

I therefore *strongly suspect* this is an issue pertaining to the
precision with which the object was saved.


  • Prev by Date: Re: release memory
  • Next by Date: Re: help in constructing a binomial consequence
  • Previous by thread: Re: Interpolation -> segfault?
  • Next by thread: Re: Interpolation -> segfault?