[CM] s7, undefined reference errors

[bil at ccrma.Stanford.EDU]

> By the way, have you measured any difference in performance
> for non-bignumbers when using WITH_GMP?

The gmp version is always slower, sometimes a lot slower
(more than a factor of 10).  I was using gmp/mpfr for
some (for lack of a less pompous word) research and
didn't care about the speed.  In numerical contexts,
the multiprecision library calls completely dominate
(99% of the compute time), so I didn't try to optimize
the other 1%.  I plead lack of time and energy!
I've reached that point in life that if I try to
work harder, I instantly fall asleep.