[Scilab-users] scilab 6.1 crashes when trying to load big matfile (loadmatfile)

[Stéphane Mottelet]

it's the loadmatfile that crashes (here on Scilab-branch-6.1 under OSX, 
with 16Gb ram)

   "savematfile('a28.mat','a28');"
scilab-cli-bin(3310,0x115f1f5c0) malloc: can't allocate region
*** mach_vm_map(size=18446744071562067968) failed (error code=3)
scilab-cli-bin(3310,0x115f1f5c0) malloc: *** set a breakpoint in 
malloc_error_break to debug

Same crash with Scilab-branch-6.1 under Ubuntu 18.04, with 128Gb ram:

   "a28=[1:2^28];"
ATTENTION : Option -v7 ajoutée.

   "savematfile('a28.mat','a28');"
Segmentation fault (core dumped)

S.

Le 19/06/2020 à 08:57, Antoine Monmayrant a écrit :
> Hello all,
>
>
> Here is a small script that systematically crashes scilab on my machine:
>
>
> ////////////////////////////////
>
> // on my machine with 8Gb or ram and usual workload, n=28 crashes scilab
> n=[24,26,28];
>
> for i=n
>
>     disp('---- '+string(i)+' ----');
>     execstr('a'+string(i)+'=[1:2^'+string(i)+'];')
>     disp('a'+string(i)+'=[1:2^'+string(i)+'];')
> execstr("savematfile(''a"+string(i)+".mat'',''a"+string(i)+"'');")
> disp("savematfile(''a"+string(i)+".mat'',''a"+string(i)+"'');")
>     execstr("loadmatfile(''a"+string(i)+".mat'');")
>     disp("loadmatfile(''a"+string(i)+".mat'');")
>     disp('---- OK ----');
>
> end
>
> ////////////////////////////////
>
> You'll have to adapt the maximum value for n depending on your 
> available ram.
>
> Antoine
>
>
> On 18/06/2020 13:06, Stéphane Mottelet wrote:
>> Hello Antoine,
>>
>> I made a 4Gb file with Matlab:
>>
>> >> a=rand(645,645,645);
>> >> b=rand(645,645,645);
>> >> c={a,b};
>> >> save("c.mat","-7.3","c")
>>
>> and managed to load it successfully in Scilab:
>>
>> --> loadmatfile("c.mat");
>>
>> --> c
>>  c  =
>>
>>   [645x645x645 constant]  [645x645x645 constant]
>>
>> Maybe the structure I tried is too simple (just two hypermatrices in 
>> a cell) so please give us a representative failing example.
>>
>> S.
>>
>> Le 17/06/2020 à 13:22, Antoine Monmayrant a écrit :
>>> Hello All,
>>>
>>> I cannot open large matfile in scilab (~3.4Gb).
>>> Scilab is always dying with an error message that is extremely 
>>> instructive: "Killed".
>>>
>>> It's a bit cumbersome to share this big fie, so do you have any idea 
>>> on how to investigate this issue and try to locate the root cause?
>>>
>>> As a side note, it might be a problem related to either the size of 
>>> the variables or their nature: cells containing hypermatrices.
>>> The file itself is not corrupted (I managed to open it in matlab and 
>>> to save each hypermatrix into individual files that got imported in 
>>> scilab with no problem.
>>>
>>> Cheers,
>>>
>>> Antoine
>>>
>>> _______________________________________________
>>> users mailing list
>>> users at lists.scilab.org
>>> https://antispam.utc.fr/proxy/2/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/antispam.utc.fr/proxy/1/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/lists.scilab.org/mailman/listinfo/users 
>>>
>>
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> https://antispam.utc.fr/proxy/1/c3RlcGhhbmUubW90dGVsZXRAdXRjLmZy/lists.scilab.org/mailman/listinfo/users 
>

-- 
Stéphane Mottelet
Ingénieur de recherche
EA 4297 Transformations Intégrées de la Matière Renouvelable
Département Génie des Procédés Industriels
Sorbonne Universités - Université de Technologie de Compiègne
CS 60319, 60203 Compiègne cedex
Tel : +33(0)344234688
http://www.utc.fr/~mottelet