Implementing and Testing New Versions of a Good, 48-Bit, Pseudo-Random Number Generator

Abstract

In this paper we describe the design, testing, and use of drand48—a good, pseudo‐random number generator based upon the linear congruential algorithm and 48‐bit integer arithmetic. The drand48 subroutine is callable from C‐language programs and is available in the subroutine library of the UNIX* operating system. Versions coded in assembly language now exist for both the PDP‐11 and VAX‐11 computers; a version coded in a “portable” dialect of C language has been produced by Rosier for the Western Electric 3B20 and other machines. Given the same initialization value, all these versions produce the identical sequence of pseudo‐random numbers. Versions of drand48 in the assembly language of other computers or for other programming languages clearly could be implemented, and some output results have been tabulated to aid in testing and debugging such newly coded subroutines. Timing results for drand48 on the PDP‐11/45, the PDP‐11/70, the VAX‐11/750, and the VAX‐11/780 are also presented and compared.

REACTIONS OF CAPTIVE HARBOR PORPOISES (PHOCOENA PHOCOENA) TO PINGER‐LIKE SOUNDS

“…White, broadband random noise was generated with a 48‐bit random number generation routine and converted with a 12‐bit D/A‐converter operating at a rate of 400 kHz. The noise signal had a flat power spectrum up to half the sampling frequency (Roberts 1982). The signal was bandpass filtered through an electronic filter (FV‐electronics, FV‐665, 48 dB/octave) and gated into 200‐ms rectangular pulses as described below. Ten cycles of sinusoidal frequency modulation from 100 kHz to 140 kHz lasting 200 ms (Fig.…”

SENR/NRPy+ : Numerical relativity in singular curvilinear coordinate systems

“…BMCCM includes many features beyond the scope of this paper [69][70][71][72][73]; here we focus on initial data that represent a version of the robust stability testbed [33][34][35][36], which involves a strong random perturbation about flat spacetime. At each grid point, the random number generator drand48 [74] is seeded with a unique (but constant) integer tied to the grid index. Then, each of the gridfunctions are populated, in turn, with Minkowski initial data (e φ = 1, ε ij = 0,Ā ij = 0, K = 0, α = 1, β i = 0, and B i = 0) plus a random value picked from the uniform distribution [−0.02, 0.02].…”