Arb is a C library for arbitrary-precision floating-point ball arithmetic. It supports efficient high-precision computation with polynomials, power series, matrices and special functions over the real and complex numbers, with automatic, rigorous error control.