We present a new algorithm which computes dot-products of arbitrary le
ngth with minimal rounding errors, independent of the number of addend
s. The algorithm has an O(n) time and O(1) memory complexity and does
not need extensions of the arithmetic kernel, i.e., usual floating-poi
nt operations. A slight modification yields an algorithm which compute
s the dot-product in machine precision. Due to its simplicity, the alg
orithm can easily be implemented in hardware.