Abstract: This article is about twofold arithmetic. Here I introduce algorithms and experimental code for twofold variant of C/C++ standard functions exp(x) and log(y), and expm1(x) and log1p(y). Twofold function z0+z1≈f(x0+x1) is nearly 2x-precise so can assess accuracy of standard one. Performance allows assessing on-fly: twofold texp(x) over double is ~10x times faster than expq(x) by GNU quadmath.
2015 (C) Evgeny Latkin, Free for academic and non-commercial
Download:
Text also uploaded to arXiv, link: http://arxiv.org/abs/1502.05216