Re: [Help-gsl] Robustness/reliability of GSL

[Brian Gough]

At Sun, 6 Aug 2006 10:10:41 +0300,
David Klein wrote:
> I have used slatec and Lapack in the past, and I am thiniking of
> switching  to GSL for the thread safety. A major concern of mine is
> the robustness of the code. Numerical software is notoriously tricky
> to get right (e.g.I found a bug in cephes gamma routine where it was
> assumed that  x<1 implies that  x+1<2. As far as I know, there is
> exactly one double precision number for which this is wrong.) A big
> advantage of slatec and lapack is that they have been extensively
> tested and used. Does anyone have any feedback to provide on GSL?

Hello,

Some general testing was done at CERN, see the "Reports and Papers"
section at http://seal.web.cern.ch/seal/work-packages/mathlibs/index.html

There is a GSL testsuite, which is fairly extensive, use "make check
GSL_TEST_VERBOSE=1" to see the full output.  Extensions are welcome.

If you want to see what bugs have been reported, the bug mailing list
archives are at http://lists.gnu.org/archive/html/bug-gsl/

-- 
best regards, 

Brian Gough
(GSL Maintainer)

Network Theory Ltd,
Commercial support for GSL --- http://www.network-theory.com/gsl/