reference implementation of BLAS

[Andrew Pinski]

On Sun, 2006-09-24 at 21:50 -0400, Jack Howarth wrote:
> Steve,
>    Have you considered creating a testcase for the situations
> where the gcc backend isn't properly propagating the
> Inf and NaN with complex arithmetic so a PR can be filed?

For -ffast-math, this is not a bug as -ffast-math enables
-ffinite-math-only.

-- Pinski