Problems with GAMMA functions?

[Angelo Graziosi]

On Tue, 9 Oct 2007, Thomas Koenig wrote:

>
> We are now using the gamma functions provided by the respective
> C libraries.  Looking at the gcc-testresults failures, and at
> your report, it seems as if the quality of existing gamma functions
> is rather low.

For the sake of completeness, the test case I cited
(gfc_gamma_test_c302m.F) is completed successfully when built with G95, so
it looks that G95 has better GAMMAs then GFortran. Why not use them? Or
the GAMMAs of CERNLIB?


>
> At the least, we should
>
> - relax the tolerances in gamma_5.f90 to something that doesn't fail
>   on as many systems

This does not seem a good solution: it is too simple to relax the
tollerance so that the test passes. Too simple!


   Angelo.