Instead of splitting mp_limb by bit operations (using HIGH/LOW macros),
now we use mp_limb2 (which is bigger integer size, e.g uint64_t). It
makes operations (especially mulitiplication) a lot faster. As a side
effect, it also reduces memory consumption (16,452,033 -> 15,545,853 on
my Linux machine).
Other changes:
- trailing zeros are removed after operations.
- division algorithm is simplified.
Users can now switch to their own implementation of GEM.
However, we do not guarantee that this will not be a problem in the current situation.
This may need to be improved in the future.