Re: Octave on windows - frequent crashes

[George]

On Wed, Dec 16, 2009 at 8:48 PM, Benjamin Lindner <address@hidden> wrote:
>>
>> Also, gnuplot will always crash when I run "close all".
>>
>
> Yes this I also see sometimes, not always. I don't know what's
> the cause of this, but it is an issue of gnuplot, not octave per se.
> Octave also has a continuously evolving builtin plotting backend, based
> on fltk. You can switch by executing
>  backend("fltk");
>
>> octave-3.2.3.exe:178:~\Desktop\literature review
>>>
>>> cov(noise)
>>
>> error: memory exhausted or requested size too large for range of
>> Octave's index type -- trying to return to prompt
>
> Now this should not make you wonder why.
> Octave/mingw32 is compiled as 32bit software.
> You are trying to create a 40000x40000 matrix which would
> require 4*4*10^8 * 4 bytes = 6.4*10^9 bytes (in single prec)
> And this is simply not possible. Remember, it's a 32bit software.
>
> You can easily generate this error by starting octave afresh
> and try to execute
>  t = zeros(40000,40000,"single");
>
>
>> octave-3.2.3.exe:178:~\Desktop\literature review
>>>
>>> whos
>>
>> Variables in the current scope:
>>
>>  Attr Name        Size                     Bytes  Class
>>  ==== ====        ====                     =====  =====
>>       argn        1x5                          5  char
>>       n           1x1                          8  double
>>   f   x         251x40000               40160000  single
>>
>> Total is 10040006 elements using 40160013 bytes
>>
>
> This now looks like a bug.
> These are the variables local to the scope of the function that
> was executed when the out-of-memory exception was thrown.
> That can be simply reproduced by:
>> foo = 1;
>> a = zeros(251,40000);
>> b = cov(a);
> error: memory exhausted or requested size too large for range of Octave's
> index type -- trying to return to prompt
>> whos
> Variables in the current scope:
>
>  Attr Name        Size                     Bytes  Class
>  ==== ====        ====                     =====  =====
>       argn        1x1                          1  char
>       n           1x1                          8  double
>   f   x         251x40000               80320000  double
>
> Total is 10040002 elements using 80320009 bytes
>
> This should be taken to octave's bug mailing list.
>
> benjamin
>

So is there a bug report for this or should I create one? It seems
consistently reproducible.