octave-maintainers
[Top][All Lists]
Advanced

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

Re: qrupdate - advice sought


From: Jaroslav Hajek
Subject: Re: qrupdate - advice sought
Date: Fri, 9 Jan 2009 15:58:33 +0100

On Fri, Jan 9, 2009 at 3:30 PM, Thomas Weber
<address@hidden> wrote:
> Am Freitag, den 09.01.2009, 14:17 +0100 schrieb Jaroslav Hajek:
>> On Fri, Jan 9, 2009 at 1:32 PM, Thomas Weber
>> <address@hidden> wrote:
>> > Am Freitag, den 09.01.2009, 13:00 +0100 schrieb Søren Hauberg:
>> >> fre, 09 01 2009 kl. 12:40 +0100, skrev Jaroslav Hajek:
>> >> > I would like Octave to take advantage of these improvements.
>> >> > Subsequently, I want to use them to further improve fsolve, lsqnonneg,
>> >> > and possibly (in less immediate future) implement other optimization
>> >> > routines that can benefit.
>> >> >
>> >> > Now, what is the best way to proceed?
>> >> >
>> >> > 1. Refresh and keep maintaining a copy of qrupdate under libcruft/ in 
>> >> > Octave.
>> >> > 2. Remove libcruft/qrupdate and require qrupdate to be linked.
>> >> >
>> >> > 1 is simpler. If we strive to get rid of some or all of libcruft
>> >> > eventually, then maybe 2 is better, but it will make another (weak)
>> >> > dependency.
>> >>
>> >> The problem with a library such as qrupdate is that no distributions are
>> >> packing it at the moment. So, I'd say put it in libcruft for now, and
>> >> remove it once distributions start packing it. Of course if nobody is
>> >> packing it in a year, then I think we should remove it from libcruft to
>> >> force distributions into packing it.
>> >
>> > The number of distributions that are keen on packaging Fortran libraries
>> > is probably zero. Good luck in forcing them.
>> >
>>
>> Well I would consider using something like OpenSUSE build service and
>> maintaining the package myself, since it's probably going to be little
>> work. But as it seems nobody objects against carrying on the inclusion
>> in libcruft, I'll probably take the least resistance route.
>
> Hmm, if the library is useful for other stuff, I suggest giving
> distributions a chance by releasing a tarball :)
>

OK, I'll do it after I figure out how :)
Maybe there's other free software similar to Octave that can benefit
(R? FreeMat?).

>
> First wishlist bug: please give a one-line summary of passed/failed
> tests. This is probably difficult currently, as you don't know what
> thresholds to chose, but that should become clear over time.
>

Good suggestion, thanks. I guess I'll just multiply my machine's
numbers by a factor of, say, 5 and use that. I could even calculate
the bounds using matrix analysis (just kidding).


>        Thomas
>
>



-- 
RNDr. Jaroslav Hajek
computing expert
Aeronautical Research and Test Institute (VZLU)
Prague, Czech Republic
url: www.highegg.matfyz.cz



reply via email to

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