[John Doty <jpd@space.mit.edu>] Re: problems compiling in ground/cmd/plan/orbit

[Mark Galassi <rosalia at galassi dot org>]

Amigos, here is a very nice endorsement of GSL by a collaborator of
mine at MIT who had used Numerical Recipes to solve a problem (thus
causing all sorts of problems).  This is for mission-ops software for
the HETE-2 satellite which launches this Saturday.

Does anyone have access to a linux/ppc machine so we can fix the
configure problem he talks about before the 0.7 release?

------- Start of forwarded message -------
Date: Tue, 3 Oct 2000 21:00:42 -0400 (EDT)
From: John Doty <jpd@space.mit.edu>
To: hete-soft@space.mit.edu
Subject: Re: problems compiling in ground/cmd/plan/orbit
Message-ID: <Pine.GSO.4.21.0010032059030.6695-100000@benz.mit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

OK, I wound up using the GSL: much more satisfactory than NR. Thanks,
Carlo, for the NIST pointer (I have it bookmarked), but it was easier not
to have to deal with mixed FORTRAN/C (been there, done that, it can be
done but it has its pitfalls).

The GSL multidimensional root finder is much better developed than I
expected from the disclaimer in the documentation. The only problem I had
in installation was the usual misconception by certain configure script
writers that "linux" is a synonym for "intel" (seen that one too often).

Aside from the usual mathematical difficulties (the desired root lies in
the vicinity of two coordinate singularities as well as a large
ill-defined region in which the function is very unstable), I had no
trouble using it. Once I found a well behaved parameterization for
inclination the difficulties evaporated.

Thanks to the GSL team for a very useful package!

It still won't compile if you haven't installed the GSL (I'll try to get
to that on the MIT SPARCs tomorrow or the next day).

John Doty		"You can't confuse me, that's my job."
Home: jpd@w-d.org
Work: jpd@space.mit.edu


------- End of forwarded message -------