[PATCH]: middle-end: Use MPFR for Bessel functions y0, y1, yn

[Richard Guenther]

On 4/25/07, Kaveh R. GHAZI <ghazi@caip.rutgers.edu> wrote:
> This patch uses MPFR to evaluate at compile-time the y* Bessel functions.
> This patch relies on the previous j* Bessel patch posted here:
>
> http://gcc.gnu.org/ml/gcc-patches/2007-04/msg01624.html
>
> Tested on sparc-sun-solaris2.10 with the mpfr-2.3.0 trunk, no regressions.
>
> I also tested this against the famous FP accuracy suite I've been using
> and it reported there were supposedly some problems with the results.
> When I discussed this with the author of the MPFR Bessel routine, he
> suspected it was a bug in Maple because Mathematica and GP/Pari both
> agreed with MPFR's numbers.  (The FP accuracy testsuite creates the
> "expected" values using Maple.)  After reporting this problem, I was told
> the folks at Maplesoft acknowledged it was a bug in Maple's BesselY code
> and resolved to fix it in a future Maple release.
>
> So, okay for mainline?

This is ok.

Richard.

:REVIEWMAIL:


>                 Thanks,
>                 --Kaveh
>
>
> 2007-04-24  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
>
>         * builtins.c (fold_builtin_1): Handle y0, y1.
>         (fold_builtin_2): Handle yn.
>
> testsuite:
>         * gcc.dg/torture/builtin-math-2.c: Test y0, y1, yn.
>         * gcc.dg/torture/builtin-math-4.c: Likewise.