[Scilab-users] Update: Maximum allocation with Scilab 6.0 beta

[Antoine Monmayrant]

Le 02/26/2016 01:41 PM, Samuel Gougeon a écrit :
> Le 26/02/2016 13:24, antoine.monmayrant at laas.fr a écrit :
>> Le 02/26/2016 01:17 PM, antoine.monmayrant at laas.fr a écrit :
>>> Hi all,
>>>
>>> For testing purposes, I am trying to allocate big chunks of memory 
>>> with scilab 6.0beta.
>>> I have ~256GB of free ram on my machine, however, Scilab fails 
>>> creating variables that are way smaller than that:
>>>
>>> --> a=rand(100000,100000);
> .
> With Scilab <6, the maximal number of components in a matrix is 
> 2^31-1, since their linearized index is encoded with 4-byte-signed 
> integers.
> Now, 1e10 is bigger than that.
It seems that you are right.
But things can get a bit funny:

--> a=[1:2^31-2];
  //no problemo
--> a=[1:2^31-1];
  //no problemo
--> a=[1:2^31];
// no error, just never ending

Antoine


> I do not think that this limit is removed in Scilab 6, but a 
> confirmation from the dev team would be useful.
>
> So here, the misleading message could come from this "index limit", 
> without respect to the available or allocatable memory.


>
> HTH
> Samuel
>
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users
>