[SciPy-user] FEM

David Cournapeau david@ar.media.kyoto-u.ac...
Mon Mar 19 05:47:42 CDT 2007


Robert Cimrman wrote:
>
> Before I coded a lot in matlab (a heart simulation FE code, with inverse
> problem solution capability), writing critical parts as C mex-files. It
> was rather painful comparing to ease of SWIG or f2py or ctypes or (plug
> your favourite wrapper generator here). The pass-by-value (all function
> arguments are immutable) concept of matlab was one of the main obstacles
> to get a decent speed - we had to use dirty tricks to override this.
To be exact, matlab uses COW to pass arguments efficiently. Otherwise, I 
agree entirely, interfacing matlab with C is really a painful 
experience: there is no way to do it efficiently and nicely at the same 
time, the C-API of matlab being far from complete for anything non 
trivial. That was one of my main reason to try (and stay with 
afterwards) python.

David


More information about the SciPy-user mailing list