[Numpy-discussion] Apparently non-deterministic behaviour of complex array multiplication

Joe Kington jkington at wisc.edu
Thu Dec 1 17:19:51 EST 2011


On Thu, Dec 1, 2011 at 2:47 PM, kneil <magnetotellurics at gmail.com> wrote:

>
> Hi Pierre,
> I was thinking about uploading some examples but strangely, when I store
> the
> array using for example: np.save('Y',Y)
> and then reload it in a new workspace, I find that the problem does not
> reproduce.  It would seem somehow to be
> associated with the 'overhead' of the workspace I am in...
>

> The context here is that I read in 24 files, totaling about 7GB, and then
> forming data matrices of size 24 x N, where N varies.  I tried for example
> this morning to run the same code, but working with only 12 of the files -
> just to see if NaNs appeared.  No NaN appeared however when the machine was
> being less 'taxed'.
>

Are you using non-ECC RAM, by chance?  (Though if you have >4GB of ram, I
can't imagine that you wouldn't be using ECC...)

Alternately, have you run memtest lately?  That sound suspiciously like bad
ram...


>
> Strangely enough, I also seterr(all='raise') in the workspace before
> executing this (in the case where I read all 24 files and do net NaN) and I
> do not get any messages about the NaN while the calculation is taking
> place.
>
> If you want to play with this I would be willing to put the data on a file
> sharing site (its around 7.1G of data) together with the code and you could
> play with it from there.  The code is not too many lines - under 100 lines,
> and I am sure I could trim it down from there.
>
> Let me know if you are interested.
> cheers,
> K
>
>
> Pierre Haessig-2 wrote:
> >
> > Le 01/12/2011 02:44, Karl Kappler a écrit :
> >> Also note that I have had a similar problem with much smaller arrays,
> >> say 24 x 3076
> > Hi Karl,
> > Could you post a self-contained code with such a "small" array (or even
> > smaller. the smaller, the better...) so that we can run it and play with
> > it ?
> > --
> > Pierre
> >
> > _______________________________________________
> > NumPy-Discussion mailing list
> > NumPy-Discussion at scipy.org
> > http://mail.scipy.org/mailman/listinfo/numpy-discussion
> >
> >
>
> --
> View this message in context:
> http://old.nabble.com/Apparently-non-deterministic-behaviour-of-complex-array-multiplication-tp32893004p32898383.html
> Sent from the Numpy-discussion mailing list archive at Nabble.com.
>
> _______________________________________________
> NumPy-Discussion mailing list
> NumPy-Discussion at scipy.org
> http://mail.scipy.org/mailman/listinfo/numpy-discussion
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/numpy-discussion/attachments/20111201/42325a21/attachment.html>


More information about the NumPy-Discussion mailing list