Implementation of multiple-precision floating-point arithmetic library for GPU computing