Re: [Gretl-devel] Octave support
Allin Cottrell schrieb:
This is now working OK on Linux and Windows (and probably OS X),
but is not yet documented. You can stick Octave script into a
gretl "foreign" block with "language=octave", and can exchange
matrices (as with Ox) via the pre-defined Octave functions
gretl_loadmat(name1)
gretl_export(matrix, name2)
The first of these matches up with mwrite(matrix, name1) in gretl,
and the second pairs with matrix = mread(name2) in gretl.
Now we need a nice example of the use of Octave to illustrate the
yet-to-be-written entry for the User's Guide. Any suggestions?
Well as Jack mentioned, maybe (cross-) spectral stuff would be a good
area because of complex numbers. For example the "cohesion" measure by
Croux/Forni/Reichlin?
And slightly OT: As a pythonista I would of course be interested in
adding numerical Python (NumPy) to the list of supported foreign
languages. I'm ready to help with the import/export functions. What
exactly would be needed?
BTW, part of the background is that I'm currently experimenting with
GPGPU = general programming on graphics processors. GPUs nowadays are
cheap parallel number crunchers. The technology keywords are CUDA
(nvidia only) and OpenCL (open standard). There exist nice wrappers for
NumPy called pycuda and pyopencl, and once everything is installed
correctly, it's quite easy to use. Following the motto "glue everything
together with Python" I thought that for some tasks it may actually pay
off to outsource econometric calculations to pyopencl. Specifically, I'm
thinking about bootstrapping impulse responses in SVARs, because that
typically takes so long that the overhead of calling python is
negligible. (However, a problem is that the SVAR estimation may have to
be re-implemented, which I would like to avoid and instead use Jack's
code, but let's see.)
well, so much for that...
cheers,
sven