marc.glisse at inria.fr
Sat Jun 23 18:51:55 CEST 2012
On Tue, 19 Jun 2012, Torbjorn Granlund wrote:
> I don't think we require C99 for GMP, only C89.
> The ldexpl function was introduced with C99.
Would mpz_get_ld become acceptable with an appropriate autoconf test for
ldexpl and a silly loop as a fallback?
Paul, what kind of guarantee would you document for the function? On
platforms where long double is binary128 (and double is binary64), we
would need a third round for a chance at maximal precision, and again it
would be strange to have such a function without full precision...
More information about the gmp-devel