I've fixed issue #58 by porting the Pade code to multiple precision floats from GMP. The number of binary digits is fixed to 256 (hardcoded as a static const member of Pade_approximant class). With this precision discrepancies in the test are limited to 1e-15.
I've fixed issue #58 by porting the Pade code to multiple precision floats from GMP. The number of binary digits is fixed to 256 (hardcoded as a static const member of Pade_approximant class). With this precision discrepancies in the test are limited to 1e-15.