[Scilab-users] scilab2c toolbox query

[Clément David]

Hello Brijesh,

To me, scilab2c for 5.5.2 works in its limited functional scope. For sysdiag or any other complex
function is suggest you to define types / sizes in a dynamic way using a scilab function to compute
them from inputs rather than implementing a mapping table (as it is currently done).

FYI scilab2c does not seems to work on Scilab 6.0.0 (at least last time I checked).

Thanks,

--
Clément

Le dimanche 10 septembre 2017 à 12:17 +0530, brijesh guptacr a écrit :
> Hello,
> 
> Brijesh here. Abhinav and I work together on Scilab2c. 
> 
> 1)Will a completely rewritten scilab2c in 5.5.2 solve the issues we are facing?
> 
> 2) Can you please explain in a little more detail as to how the suggested approach works in Scilab
> 6 and why not in 5.5.2. 
> 
> 3)Would switching from 5.5.2 to 6 help us?
> 
> Sysdiag, orth, pca and svd to name a few functions that are problematic. 
> 
> With Regards,
> Brijesh Gupta C R
> 
> On Wed, Sep 6, 2017 at 7:13 PM, Clément David <Clement.David at esi-group.com> wrote:
> > Hello Abhinav,
> > 
> > Nice to have someone interested in scilab2c. Currently there is no active maintainer, feel free
> > to
> > contribute to it.
> > 
> > > 1) To write a new function, we always have to decide the output sizes according to scalar
> > inputs
> > > or size of input matrix. But, if our output size depends on an element which is a part of
> > matrix,
> > > what can we do to assign the output size accordingly.
> > 
> > The current code generation need a C-like arguments representation for the Scilab function to
> > generate the good corresponding C function call. If you wish to have a more dynamic behavior to
> > define the outputs argument you should probably need to serialize/deserialize them in some way
> > to a
> > raw vector.
> > 
> > > 2) There are few functions whose output size has no relation to input matrix sizes, or they
> > don't
> > > follow any regular pattern. For example, they depend on rank of input matrix or something
> > else.
> > > For such functions, is there any work around to estimate the size first and assign accordingly
> > in
> > > the Tool Initialisation part.
> > 
> > No currently in the Scilab2C code and with my limited understanding, there is no way to compute
> > the
> > number of / size of the arguments. This approach however is fully valid and is partially
> > implemented
> > in the Scilab 6 analysis pass. A completely rewritten Scilab2C could then use the propagated
> > size /
> > types constraints to handle theses dynamic functions. Could you give me a function you wish to
> > add ?
> > 
> > Thanks,
> > 
> > --
> > Clément
> > _______________________________________________
> > users mailing list
> > users at lists.scilab.org
> > http://lists.scilab.org/mailman/listinfo/users
> 
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users