[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Code integration into Octave core of ilu and ichol functions.

From: Kai Torben Ohlhus
Subject: Re: Code integration into Octave core of ilu and ichol functions.
Date: Wed, 13 Aug 2014 00:09:14 +0200

On Tue, Aug 12, 2014 at 6:45 PM, Eduardo <address@hidden> wrote:
Hi Kai,

I have moved the code from my personal repository to a new Octave subrepo I have set on bitbucket. I made the changes necessary for adapting the code to compile within Octave development branch. The locations of the files in the source code tree  are:

(1) libinterp/dldfcn for ilu0.cc iluc.cc ilutp.cc ichol0.cc and icholt.cc files (dld functions)
(2) scripts/sparse for ilu.m and iluc.m

I chose those places because I saw (looking at your repository) you did last year this way. The new subrepo can be accessed from here [1].

I also have some questions about the way you managed the visibility of dld functions in terms of being able or not to call them from the interpreter. I noticed that your approach last year was to let the functions in location (1) be callable from the interpreter, maybe because they offered some extra functionality that was not used by the wrapper functions (ilu.m and ichol.m). In my case they do not.  So maybe a good idea would be to make them in some way private so they can be accessed by the wrappers ilu.m and ichol.m but not callable from the interpreter. Is this the way to go? If so, how can this be done?

[1] https://address@hidden/edu159/octave-edu159



Hi Eduardo,

Regarding your question, you're right. I chose to let the individual functions to be callable to someone who has a deeper knowledge about the factorizations and wants to directly apply them. However, when I take a look in the folder dldfcn, the current convection seems to be creating a __filename__.cc file for internal functions that are not intended to be used. But they are not hidden then that is for sure. I don't know if a "private" subdirectory makes sense in this case, like it is done for m-Files.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]