<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#ffffff">
Le 07/10/2010 22:20, Jean-Pierre Dussault a écrit :
<blockquote
cite="mid:835048110.1187875.1286482848931.JavaMail.root@zmbs3.inria.fr"
type="cite">
<meta http-equiv="content-type"
content="text/html; charset=ISO-8859-1">
Hi!<br>
<br>
the toolbox_skeleton as well as the instructions at <br>
<blockquote><a moz-do-not-send="true" class="moz-txt-link-freetext"
href="http://wiki.scilab.org/howto/Create_a_toolbox">http://wiki.scilab.org/howto/Create_a_toolbox</a>
<br>
</blockquote>
wrap the example fortran fsum into a C-wrapper. However, there exists
a "fort" command in scilab which allows to interface directly a fortran
function. Why then use the complicated fortran --> C --> scilab
path instead of fortran --> fort(...)? <br>
<br>
Further complications arise when dealing with non-scalar variables
which will have to be properly converted in C before being properly
transmitted to scilab while the instruction "fort" takes care of this
easily while hiding cumbersome conversion details.<br>
<br>
Will the "fort" command remain in scilab? If so, we will stick to using
it!<br>
<br>
Thx,<br>
<br>
JPD<br>
<br>
P.S. In the toolbox_skeleton (scilab binary distribution), csum.h
and csub.h are missing.<br>
</blockquote>
Using gateways instead of fort is generally more efficient, because the
argument compatibility tests,... are made in C instead of in Scilab
code.<br>
Moreover fort as limitations: it is not possible to pass array of
strings, short integers, function as arguments.<br>
<br>
Serge Steer<br>
</body>
</html>