This library provides the data types APInt and APFloat which model integer and floating point types of arbitrary width.
APInt is a 2's complement integer type of runtime specified bit width. APFloat at this point is not properly implemented and only supports 32 bit and 64 bit modes, implemented using the builtin float and double types.
Many arithmetic operations are provided. The library does not provide operator overloads, partly because APInt models both signed and unsigned integers and the sign interpretation is encoded into the operations. For example sdiv performs signed and udiv performs unsigned division.
The main purpose of the library is to emulate target arithmetic in a compiler. Performance should be reasonable but is not the main focus.