[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Oct-file version of the cquad integrator
From: |
Pedro Gonnet |
Subject: |
Re: Oct-file version of the cquad integrator |
Date: |
Thu, 15 Jul 2010 16:14:11 +0100 |
Hi Jaroslav,
It's been a while since I've heard anything regarding the cquad
integrator. Are there any plans to make this a fixed part of octave? If
so, is there anything I can do to help?
Cheers,
Pedro
On Tue, 2010-04-27 at 15:20 +0100, Pedro Gonnet wrote:
> Hi Jaroslav,
>
> Thanks for the comments! Here are the fixes:
>
> 1. Damn, these are standard in C and unfortunately, I'm not much
> of a C++ programmer. I've replaced them with the local macros.
>
> 2. You actually uncovered a quite different bug... The number of
> sings (and thus initial intervals) is actually limited by the
> size of the heap, so I just declared a constant array and added
> a check for overflow.
>
> 3. I have absolutely no idea what you're talking about... Feel
> free to change it however you feel is best!
>
> Cheers,
> Pedro
>
>
> On Tue, 2010-04-27 at 15:29 +0200, Jaroslav Hajek wrote:
> > On Tue, Apr 27, 2010 at 2:01 PM, Pedro Gonnet <address@hidden> wrote:
> > >
> > > Small bug in the handling of string arguments fixed, e.g. it doesn't
> > > handle functions passed as strings, just handles the error correctly
> > > now.
> > >
> > > Cheers, Pedro.
> > >
> >
> >
> > Nice job. I'm on vacation now, but I had a quick look anyway. I have
> > several suggestions:
> >
> > 1. don't use std::isnan, std::isinf, std::isfinite. These are not
> > standard C++, so we shouldn't rely on them being available. Instead,
> > Octave provides xisnan, xisinf, xfinite. These will map onto the std::
> > functions if they are available, so they incur no performance penalty,
> > otherwise they will be emulated. When Octave switches to C++0x
> > standard, we'll probably replace the x-versions.
> >
> > 2. *don't* use alloca, it's non-standard and often buggy. Use
> > OCTAVE_LOCAL_BUFFER instead.
> >
> > 3. Extracting function pointer through function_value works, but it
> > has the drawback that overloads for a named handle are ignored. An
> > alternative is to call directly the octave_value::do_multi_index_op
> > method.
> >
> > I'll probably address these points later unless you do it first. In
> > any case, thanks for this contribution.
> >
> > regards
> >
>
- Re: Oct-file version of the cquad integrator,
Pedro Gonnet <=