MathGroup Archive 1997

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

Search the Archive

swap orgy with v.3

  • To: mathgroup at
  • Subject: [mg7539] swap orgy with v.3
  • From: Buttgereit <Buttgereit at>
  • Date: Fri, 13 Jun 1997 03:08:54 -0400 (EDT)
  • Sender: owner-wri-mathgroup at

Hi there,
I use Mathematica 3.0 for Students on a 64 MB iP 100 Asus I/P55 with a 620
MB Pagefile under Windows NT 4.0 SP 2.
Using the NonlinearFit Package to perform a nonlinear least squares fit on
bivariate data of about .5 MB, the whole thing evaluates fine up to the 10
iteration (ShowProgress->True) which is already very stable.
By then the kernel has occupied about 35 MB Ram and the pagefile has grown
from 30 to 150 MB.
The system is ok - you can do word-processing or something without
Then, the kernel sucks in memory: occupied memory rises up to 45 MB, the
swap file grows up to almost 500 MB. From then on no further output is
produced (traced by ShowProgress->True). The hard disks (I have two EIDE
disks) are working hard, indicating a swap orgy. It is impossible to use
the computer for anything else but waiting: it takes 10 to 30 seconds to
change a windows focus. If one is patient, one can watch that situation for
the next 24 hours (I did so...). There is no sign indicating that this
state will ever change if one waits longer. The same thing happens saving a
large notebook (size about 40 MB because of 40 ListDensityPlot Graphics).
If it is possible to save such a notebook at all, it will take longer than
20 hours!
The system is ok immediately and all memory is freed (pagefile returns to
30 MB) after quitting MathKernel.

The pagefile is on a single (stripe set) partition formatted with FAT.
Referring to "Diskeeper" it is not fragmented at all. Anyway, I have tested
the whole thing with a new pagefile with the same problems.

Suggestions appreciated,
thanks in advance,

  • Prev by Date: Re: Re: Just another bug in MMA 3.0
  • Next by Date: Re: Need a new type of Hold
  • Previous by thread: AutoLoading Packages-License 2079-6702, Sean Ross
  • Next by thread: Re: List plot with + possible?