Multi implements multi-precision arithmetic for integers and real numbers (using both fixed and floating point formats). There is a comprehensive real function library (Sines, Exponentials, Log etc) tested to 10000 decimal places, but probably good for many more.
There are advanced 'fast' algorithms (based on Fourier transforms (FFTs) and Number Theoretic Transforms (NTTs)) that implement basic arithmetic effectively to multi-million decimal places. This is used to implement a variety of fast algorithms to compute Pi to this precision. There are also several fast integer factorisation algorithms implemented (such as the Continued Fraction method), although the most sophisticated methods (Elliptic Curve and Number Field Sieve) are not available.
This subsystem also implements arithmetic for the 32-bit & 64-bit 'Cardinals', ie unsigned integers, and algorithms that use these (eg factorisation).