doubledouble.py is a library for computing with unevaluated sums of two double precision floating-point numbers. The so-called "double-double arithmetic" enables operations with at least 106-bit significand (2x53-bit) and 11-bit exponent, compared to 113-bit and 15-bit, respectively, of the IEEE 754-2008 quadruple precision floating-point format.

Install with pip install doubledouble.