This is the mail archive of the
gsl-discuss@sourceware.cygnus.com
mailing list for the GSL project.
Re: How to organise Tolerance problem in integration?
- To: Vanroose Wim <vanroose at ruca dot ua dot ac dot be>
- Subject: Re: How to organise Tolerance problem in integration?
- From: Brian Gough <bjg at network-theory dot co dot uk>
- Date: Tue, 2 May 2000 20:09:47 +0100 (BST)
- Cc: gsl-discuss at sourceware dot cygnus dot com
- References: <390EB52D.E9BF4339@ruca.ua.ac.be>
The simplest way to avoid complaints about roundoff error is to choose
an absolute error tolerance (epsabs) which is sufficiently large,
i.e. greater than the level of round-off error.
The alternative is too turn off the error handler (described in the
manual) and check whether the return code is GSL_EROUND.
Vanroose Wim writes:
> Dear GSL-users,
>
> I am using the integration package "gsl_integration_qag" to
> calculate matrix elements of a potential between two oscillator eigen
> functions. So of each n and m I obtain a matrix element <n|V|m>.
>
> Problem is dat <n|V|m> decreases with n and m and from a
> certain point it is a small a 10e-13. For these high matrix
> elements. The integration program start to protest and mentions
> "ERROR roundoff error prevents tolerance from being achieved".
>
> So how do I organise a program that results into a 0 when
> the upperbound is smaller than a certain threshold.
>
>
> Best Wishes
>
> Wim Vanroose
>
>
>