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`.