[Maxima-discuss] About gentran and cffi

[Michel T. <ta...@lp...>]

Hello,

since recently gentran is becoming fashionable, i have investigated  the 
possibility of using gentran and external programs

to help doing calculations in maxima, using cffi. It took me a lot of 
time because the documentation of cffi is more oriented

to show that the authors are very smart than to explain how to use cffi 
in many different use cases in a direct way.

Moreover i encountered a bug, there is a variable err in the external 
program which confuses a variable *err*

somewhere in lisp, and gives a message "memory corrupted" which is not 
very informative.  Anyways i took as an example

the program intde (integration with double exponential approch to the 
limits) that D. Prodanov advocated here recently,

because it is quite simple and at the same time presents the same type 
of problems encountered in many computational

problems, namely there are functions called via function pointers and 
results returned via pointers.

I have worked with the C version of gentran , that is setting

gentranlang:c;

because i am more used to C than Fortran,  and the declarations of cffi 
are close to the C declarations.  Note that

gentran does not output the necessary declarations so the the program 
compiles, and one has to put them by hand.

An interesting point about intde is that the author provide both C code 
and Fortran code and one can see the great

parallelism between the two (see the package in 
http://www.kurims.kyoto-u.ac.jp/~ooura/intde.html).

All this i have run under Ubuntu 20.04

Maxima 5.43.2 http://maxima.sourceforge.net
using Lisp SBCL 2.0.1.debian

quipped with clang and gcc, i have no idea how it works on other 
platforms and  i don't care at all. I have just tried

loading quicklisp under clisp and it did not work, complaining about the 
asdf version.

You can find attached the integration program intde1.c,  a program 
tot.c  which contains the function f to be integrated

and can also be run directly to show the value of the integral, finally 
a maxima program intdetest.mac which does the

cffi business and computes the same integral under cffi control, so that 
the result can be used elsewhere in maxima.

This is obviously not very useful for intde, but may be very useful for 
compute intensive programs like colnew.

Hoping this may be a useful example for people wanting to pursue the idea.



-- 
Michel Talon