* Error scaling in ODE-IV
@ 2002-07-23 13:08 Dave Auslander
2002-07-24 15:30 ` Brian Gough
0 siblings, 1 reply; 2+ messages in thread
From: Dave Auslander @ 2002-07-23 13:08 UTC (permalink / raw)
To: gsl-discuss
I am using the ODE-initial value routines from GSL. In looking at section
24.3 or the GSL manual, I note that there does not appear to be any scaling
array for the error calculation in the adaptive algorithms. eps_rel and
eps_abs are scalars. Most ODE systems, however, are not naturally scaled so
all components have the same scale so it is very difficult to optimize
performance.
Am I missing something or is this the intended usage?
Thanks
Dave Auslander
PS - many thanks to the GSL developers. This is a huge, very useful effort.
+++++++++++++++++++++++++++++++++++++++++++++++
David M. Auslander, Prof. of ME and Assoc. Dean
Mechanical Engineering, Univ of Calif, Berkeley, CA 94720-1740
Dean's office: 510-642-7594 ME: 510-642-4930 Fax: 510-643-5599
dma@me.berkeley.edu www.me.berkeley.edu/~dma
+++++++++++++++++++++++++++++++++++++++++++++++
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: Error scaling in ODE-IV
2002-07-23 13:08 Error scaling in ODE-IV Dave Auslander
@ 2002-07-24 15:30 ` Brian Gough
0 siblings, 0 replies; 2+ messages in thread
From: Brian Gough @ 2002-07-24 15:30 UTC (permalink / raw)
To: Dave Auslander; +Cc: gsl-discuss
Dave Auslander writes:
> I am using the ODE-initial value routines from GSL. In looking at
> section 24.3 or the GSL manual, I note that there does not appear
> to be any scaling array for the error calculation in the adaptive
> algorithms. eps_rel and eps_abs are scalars. Most ODE systems,
> however, are not naturally scaled so all components have the same
> scale so it is very difficult to optimize performance.
> Am I missing something or is this the intended usage?
Hi,
Hmmm.. you're right, component-wise scaling doesn't seem to be part of
the existing API. I think this will need to be added to a future
release, with a control object that accepts arrays, rather than single
values. It should be a straightforward generalisation of the existing
code in ode-initval/cstd.c.
regards
Brian Gough
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2002-07-24 22:30 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-07-23 13:08 Error scaling in ODE-IV Dave Auslander
2002-07-24 15:30 ` Brian Gough
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).