GNU MultiPrecision includes a "MPF" sublibrary", which I'll summarize today.
This operates upon an `mpf_t` structure with the following utilities:
* Accessors (possibly involving validation or bittwiddling/adds/subtracts/absolutes)
* Wrappers around math operators adding data conversions, some optionally reimplementing the operator.
* Including a few for random-numbers
* Accessors for globals
* Subtract-then-divide with fastpath for 0, a "relative diff"
* Optimized repeated multiplication
1/?