Re: limit

One fix that was suggested was to use something like this to increase  
space dynamically, rather than quitting:


  if (k + tmp1.nnz() > nzmax) {
       nzmax = std::max((int)((k + tmp1.nnz())*2),(int)(nzmax*1.5));
       c.resize(nzmax);
       jc.resize(nzmax);
       std::cerr << "Not enough work space, increasing the capacity to:  
" <<
nzmax << std::endl;
       //return;
     }


On Saturday, October 4, 2003, at 05:41  PM, Leandro Franco Londoño  
wrote:

>
> Thank you Janusz, your suggestion was really appreciated.
>
> I converted the matrices type to compressed and I fixed some problems
> (the unsigned long int index helped a lot), now I have some problems
> because the rows and columns methods don't work with these matices but  
> I will
> figure out how to solve them later.
>
> Well,  I'm writing again because I found another funny problem. This  
> time
> I get a "Not enough work space, increase capacity of C.", I found a  
> similar
> question in the mail archives where someone suggested "to allocate  
> enough
> storage to C to contain the result of the multiplication".
>
> I think I get the idea.. you need to tell the matrix how big it is  
> going to be
> after the multiplication, but if both matrices are sparse and you  
> don't know
> their sizes what should you do? should you allocate enough space for a  
> full
> matrix, or could you do it dynamically ?
>
> Finally, ( I know I'm really boring but... ) could someone give me  
> some hints
> about the space allocation for the compressed matrices?
>
>
> Thank you very much.
>
>
>
> Leandro Franco Londoño
> Ing. de Sistemas y Computación
> Universidad de los Andes
> Bogota
> Colombia
>
>
>
> Selon Janusz opechowski <[email protected]>:
>
>> Hello!
>>
>> So probably it will be a sparse matrix
>> (having zeroes in much more than a half of all
>> positions) ,
>> like this :
>>
>>   typedef matrix<double,
>>                  rectangle<>,
>>                  compressed<int, external,
>>                                  index_from_one>,
>>                  row_major>::type MatA;
>>
>> but here instead of "int"  you should take
>> as a index type something "safer"  i.e.
>> "unsigned long int", because
>> on some architectures
>> int is 16 bit only so : around + - 32000
>> and you need 20 bits one way so 32 bits
>> would be enough.
>>
>> If those matrices are not sparse ... ,
>>  (10^6)^2 = 10^12,
>>    so ..  terabyte matrices ?
>> Congratulations!
>>
>> Probably :
>> typedef matrix<double>::type MatA,
>> will be sutable if no
>> try to play with arrays :
>>
>>   typedef matrix< double,
>>                   rectangle<>,
>>                   array< dense<> >,
>>                   row_major>::type MatA;
>>
>> Cheers
>> Janusz
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>  --- Leandro Franco Londoño <[email protected]>
>> wrote: > Hello there,
>>>
>>> In the first place I would like to thank you guys
>>> for developing MTL,
>>> it was nice to find just exactly what I was looking
>>> for.
>>>
>>> Well... here is my question : I'am using matrices of
>>> type complex<double>
>>> for some calculations (a quantum computing
>>> simulator) but these matrices are
>>> really big (from 2^10 to 2^20 ) and I having some
>>> problems in the
>>> implementation.
>>>
>>> I would like to know how I should work with that
>>> type of matrices (that is a
>>> little bit general so all kind of ideas are welcome)
>>> and it will be specially
>>> usefull if someone could tell me what are the limits
>>> in the sizes using MTL
>>> ( can I do something like Matrix(2^32,2^32) ? ).
>>>
>>>
>>> Thank you very much for your help
>>>
>>>
>>>
>>> Leandro Franco Londoño
>>> Ing. de Sistemas y Computación
>>> Universidad de los Andes
>>> Bogota
>>> Colombia
>>>
>>> _______________________________________________
>>> This list is archived at
>> http://www.osl.iu.edu/MailArchives/mtl-devel/
>>
>> ______________________________________________________________________ 
>> __
>> Want to chat instantly with your online friends?  Get the FREE Yahoo!
>> Messenger http://mail.messenger.yahoo.co.uk
>> _______________________________________________
>> This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/
>>
>
>
> _______________________________________________
> This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/


_______________________________________________
This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/