romberg



>>>>> "Daniel" == Daniel Lakeland <dlakelan at street-artists.org> writes:

    Daniel> completely robust, having failed some of his tests. If a bigfloat
    Daniel> version of quadpack were available it would probably be wise to

I have given this idea a little thought.  I think that quad_qags is
basically independent of the float type except for the routine qk21,
which is the Gauss-Kronrod integrator.  For each bfloat precision,
you'd have to recompute the coefficients of the Gauss-Kronrod
integrator.  I don't know how to do that, but if requires solving for
the roots of a polynomial, you have very good starting points since
you have the values for the double-float version.

An interesting project, I think.

Ray