Mathematica 9 is now available
Services & Resources / Wolfram Forums / MathGroup Archive
-----

MathGroup Archive 2010

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

Search the Archive

Re: Floating point exception when taking a reciprocal

  • To: mathgroup at smc.vnet.net
  • Subject: [mg110515] Re: Floating point exception when taking a reciprocal
  • From: David Reiss <dbreiss at gmail.com>
  • Date: Tue, 22 Jun 2010 07:00:56 -0400 (EDT)
  • References: <hvn01h$q75$1@smc.vnet.net>

For Mathematica version 7 and 7.01 there is a bug in the Intel Core i7
(Nehalem) based CPU processor architecture as it is used on Linux
which requires a special patched version of Mathematica to be
installed.  you should contact Woflram support for the patched
version.

--David


On Jun 21, 2:15 am, Niels Walet <Niels.Wa... at manchester.ac.uk> wrote:
> Leo Alekseyev wrote:
> > I am running into a strange bug where taking reciprocals of a list of
> > floating point numbers crashes the kernel with a floating point
> > exception.  This is on a Linux x86_64 system, Ubuntu 9.10, running
> > Mathematica 7.0.1.0
>
> > Something like
> >    foo=Array[1.0&,1000]; 1/foo
> > or
> >    foo=RandomReal/@Range[100]; 1/foo
> > gives me the crash
> > Replacing floats with integers makes things work fine, also, for small
> > enough lists (order of 10 elements) everything works.
>
> > Same version of Mathematica on a 32-bit linux system or on Windows
> > doesn't exhibit this behavior.
>
> > I am rather mystified by what's going on; my hardware/OS is pretty
> > standard, so if the bug were this obvious it would have long been
> > known by now...  But if anyone else has run into similar behavior
> > and/or knows of some sort of a fix, please let me know.
>
> > Thanks,
>
> > --Leo
>
> I have seen similar problems (with Scientific linux 5 64 bits on an
> intel i5 processor). The same code works on the identical OS with an
> older Core2 processor. I have the feeling it is due to the issue with
> the intel math library for i7
> (http://support.wolfram.com/technotes/IntelCorei7.html), but am not
> sure, since I can't directly talk to wolfram (site license).
>
> --
> Prof. Niels R. Walet                    Phone:  +44=
(0)1613063693
> School of Physics and Astronomy         Fax:    +44(0)1613064=
303
> The University of Manchester            Mobile: +44(0)7905438=
934
> Manchester, M60 1QD,  UK
> Niels.Wa... at manchester.ac.ukhttp://theory.phy.umist.ac.uk/~mccsnrw



  • Prev by Date: Re: Ballistic launch trajectory code
  • Next by Date: Re: Difficulty with saving Package .m files
  • Previous by thread: Re: Floating point exception when taking a reciprocal
  • Next by thread: More about algebraic simplification