Realistic analysis of some randomized algorithms

Many problems such as primality testing can be solved efficiently using a source of independent, identically distributed random numbers. It is therefore customary in the theory of algorithms to assume the availability of such a source. However, probabilistic algorithms often work well in practice with pseudo-random numbers; the point of this paper is to offer a justification for this fact. The results below apply to sequences generated by iteratively applying functions of the form ƒ (&khgr;) = &agr;&khgr; + &bgr; (mod p) to a randomly chosen seed x, and estimate the probability that a predetermined number of trials of an algorithm will fail. In particular, the following bounds hold: For finding square roots modulo a prime p, a failure probability of &Ogr; (log p/√p). For testing p for primality, a failure probability of &Ogr; (p-1/4+&egr;), for any &egr;>0. (In both cases, the number of trials is about 1/2 log p.) The analysis uses results of André Weil concerning the number of points on algebraic varieties over finite fields.