octave-maintainers
[Top][All Lists]
Advanced

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

Re: GSOC 16, Improvements to sqrtm,logm and funm


From: Mudit Sharma
Subject: Re: GSOC 16, Improvements to sqrtm,logm and funm
Date: Mon, 21 Mar 2016 00:54:35 +0530

The libpcre3-dev package worked fine.
I am successfully able to build octave now .

I have begun with my proposal now. can you send me the link for the
application format or anything else related to GSoC application?

I will be primarily focussing on the work which was supposed to be done last year.

Also I would like to point out that some scripting work has already been done
for this project under prof Marco. The latest scripts for the functions expm,logm and funm
As pointed out that JIT compiler is unavailable for Octave, so funm
needs to be written in C++.
And sqrtm requires it's 3rd output argument to be updated with that
of https://github.com/RickOne16/matrix/blob/master/sqrtm2.m

I have sound background with C++STL or Boost if required.

Here's a brief summary of what I will be including in the proposal :

* Work on the 3 toolboxes provided by Higham. This is the first half of the
project. This task primarily includes writing tests.

* Write funm in C++, rewrite sqrtm in C++ and since scripts are slow
  rewrite expm and logm in C++ ( scripts supplied earlier in this mail)

* If time allows include function for Pth roots and trigonometric functions.

I need a review about expm and logm, do both of them need a C++ code as well?

-Mudit


On Fri, Mar 18, 2016 at 5:01 AM, Jordi Gutiérrez Hermoso <address@hidden> wrote:
On Fri, 2016-03-11 at 11:26 +0530, Mudit Sharma wrote:
> Well, I have tried the mft toolbox, one can test it by
> the command "mft_test " , it is clear by seeing the code
> of mft_test.m that certain functions are missing from Octave
> which are called by the "mft_test" command/file and thus
> results in errors like
> error: 'superiorfloat' undefined near line 201 column 45
> I have the funm code for Octave which I pasted in the mft folder after
> which errors related to funm disappeared.
> In short octave is missing some more functionality to run mft toolbox.
>
> Will a GSoC project be possible in this area?

Yes, but I don't think focussing on all of the details like
"superiorfloat" are necessary, which is an undocumented Matlab
function. If a slight rewriting of the tests is necessary without
altering their spirit, that's ok.

It would also be interesting to test the performance. The funm code
that Higham et al wrote is for Matlab, so it assumes that a JIT
compiler is available. We may have to do some rewriting, perhaps in
C++, in order to account for the lack of a JIT compiler in Octave.

Seeing how you have already been spending time over the past year
working on this, I encourage you to propose a formal application.

- Jordi G. H.




reply via email to

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