[Scilab-Dev] Gmp in scilab revisited.

[Jonathan Blanchard]

I see what you mean I think. But what happen when it get destroyed.
Let say for example the following code.

a = gmpinit(123);
b = gmpinit(234);

a = gmpmul(a,b);

In this code the initial variable 'a' get cleared but if memory was
allocated out of the Scilab stack it won't be freed.

Jonathan Blanchard



On Mon, Sep 29, 2008 at 5:17 PM, Yann COLLETTE <ycollet at freesurf.fr> wrote:
> Hello,
>
> I don't know anybody currently working on this topic today.
> I have some idea for the interface: use the pointer type "p". With this
> type, you can allocate the  memory required to store the gmp number.
> Then, you store the pointer to the memory allocated in a mlist (here, the
> type of the mlist is fannlist and the fann field contains the pointer).
> To retrieve the gmp number, use the GetRhsVar with the "p" type.
> Here are some examples I use in a FANN (Fast Artificle Neural Network)
> interface.
>
> For the basic operations like +, -, etc ..., maybe you can use overloading:
> for the + operator:
> function c = gmplist_a_gmplist(a,b) c = gmp_add(a,b); endfunction
>
> int createScilabFannStructFromCFannStruct(struct fann* ann,unsigned int
> StackPos)
> {
>  int m_list_labels,  n_list_labels;
>  int m_fann_pointer, n_fann_pointer;
>  int m_extra,        n_extra,   l_extra;
>
>  //The struct field names
>  static char *fieldnames[] = {"fannlist","fann"};
>
>  m_fann_pointer = 1; n_fann_pointer = 1;
>  m_extra        = 2; n_extra        = 1;
>
>  CreateVar(StackPos, "m", &m_extra,  &n_extra, &l_extra);
>
>  m_list_labels = 2; n_list_labels = 1;
>  CreateListVarFromPtr(StackPos, POS_LABELS,       "S", &m_list_labels,
>  &n_list_labels,  fieldnames);   // Put the type and the labels
>  CreateListVarFromPtr(StackPos, POS_FANN_POINTER, "p", &m_fann_pointer,
> &n_fann_pointer, ann);
>
>  return 0;
> }
>
> int createScilabFannStructFromCFannStruct(struct fann* ann,unsigned int
> StackPos)
> {
>  int m_list_labels,  n_list_labels;
>  int m_fann_pointer, n_fann_pointer;
>  int m_extra,        n_extra,   l_extra;
>
>  //The struct field names
>  static char *fieldnames[] = {"fannlist","fann"};
>
>  m_fann_pointer = 1; n_fann_pointer = 1;
>  m_extra        = 2; n_extra        = 1;
>
>  CreateVar(StackPos, "m", &m_extra,  &n_extra, &l_extra);
>
>  m_list_labels = 2; n_list_labels = 1;
>  CreateListVarFromPtr(StackPos, POS_LABELS,       "S", &m_list_labels,
>  &n_list_labels,  fieldnames);   // Put the type and the labels
>  CreateListVarFromPtr(StackPos, POS_FANN_POINTER, "p", &m_fann_pointer,
> &n_fann_pointer, ann);
>
>  return 0;
> }
>
> I hope this will help ...
>
> YC
>
> Some ideas for the
> Sylvestre Ledru a écrit :
>>
>> Le lundi 29 septembre 2008 à 16:13 -0300, Jonathan Blanchard a écrit :
>>
>>>
>>> Sorry to bother everyone on the list again about this subject.
>>>
>>> I have a question to Ledru you said yes to my question earlier but I'm
>>> not sure which. Is there any plan for gmp/mpr in scilab in the future
>>> and is there a developer currently looking at this?
>>>
>>
>> No and no ;)
>> At least, as far I know. Yann is the one who added this idea to the
>> Wiki. Maybe he knows someone who is working on it.
>>
>>
>>>
>>> I also didn't listen to your warning and hacked myself a way to use
>>> gmp in scilab.
>>>
>>
>> Seems you did the right choice ;)
>>
>>
>>>
>>> Sadly because of the nature of gmp it is impossible to
>>> create a toolbox for Scilab 5 as I would have loved to create one.
>>>
>>
>> Why is it impossible ? Too tightly linked with Scilab's internal stuff?
>>
>> Sylvestre
>>
>>
>
>