Release 0.2.1

This patch release

• fixes compilation problems in CUDA 10.1;
• makes the build system a little more friendly; and
• fixes a problem with the bench program which caused it to fail when using the default parameter value.

Release 0.2.0

This release includes

• Improvements to the selection of window parameters in the modular exponentiation algorithms.
• Much faster full-width multiplication and squaring algorithms.
• Introduction of a MODNUM type, based on FIXNUM.
• Implementation of Montgomery REDC-based MODNUM (using the aforementioned new multiplication and squaring algorithms), for comparison with the Montgomery CIOS-based MODNUM.

Release 0.1.0

Initial release of cuda-fixnum. Includes complete support for modular arithmetic, including modular exponentiation, on fixnums up to 2048 bits.