Re: Odeset - Odeget

[c.]

Roberto,

When you write to the list, please do not top post, i.e. put your answer AFTER, 
not BEFORE the message you are responding to.
We follow this guideline as it makes the discussions more readable for users 
who find a post by searching on the web archives.

On 2 Sep 2013, at 04:06, Roberto Porcù <address@hidden> wrote:
>> 
>> On 1 September 2013 18:47, Roberto Porcù <address@hidden> wrote:
>>> Il giorno 01/set/2013, alle ore 13:43, "c." <address@hidden> ha scritto:
>>>> 
>>>> On 1 Sep 2013, at 17:40, Roberto Porcù <address@hidden> wrote:
>>>> 
>>>>> Ok, thank you very much Carlo and Olaf.
>>>>> So for now i will start working on
>>>>> https://savannah.gnu.org/patch/?8047
>>>>> cleaning the function according to the discussion.
>>>>> 
>>>>> Roberto
>>>> 
>>>> Roberto,
>>>> Do you already have write access to OctaveForge?
>>>> If not you should ask carnë to enable your account.
>>>> c.
>>> I think i have access to Octave Forge because i was able to download the 
>>> source code from the page you linked to me, clicking on the links at the 
>>> bottom of the page. Is it ok?
>>> 
>>> roberto
>> 
>> I'll need your username to give you commit access. Also, please read
>> the commit message guidelines before making commits [1].
>> 
>> Carnë
>> 
>> [1] http://wiki.octave.org/Commit_message_guidelines
> 
> Ok Carnë, my username is   robbb88
> thank you
> 
> I have a question about the patch: 
> i read the files  odebwenk.m  and  inexact_newton.m
> and i read the comments about that code, but i've not understood
> very well what modifies do i have to do. 
> Do i have to merge odebwenk and odebwe? Or do i have to do what Carlo said in
> his last comment?
> Thanks

I don't see the difference between the two,
the idea is to have only one odebwe function that 
can use either Newton Krylov (like odebwenk) or plain Newton (like odebwe)
and can use an exact Jacobian if given (like odebwe) or an approximate 
Jacobian otherwise (like odebwenk).

What else would you do to unify the two functions?

As for timestep selection you can use the simple analytic formula used in 
odebwenk as the extrapolation approach in odebwe is slitely more 
computationally expensive.

> Regards
> Roberto

c.