Abstract :
A hardware architecture for non-uniform random number generation, which allows the generator´s distribution to be modified at run-time without reconfiguration is presented. The architecture is based on a piecewise linear approximation, using just one table lookup, one comparison and one subtract operation to map from a uniform source to an arbitrary non-uniform distribution, resulting in very low area utilisation and high speeds. Customisation of the distribution is fully automatic, requiring less than a second of CPU time to approximate a new distribution, and typically around 1000 cycles to switch distributions at run-time. Comparison with Gaussian-specific generators shows that the new architecture uses less than half the resources, provides a higher sample rate and retains statistical quality for up to 50 billion samples, but can also generate other distributions. When higher statistical quality is required and multiple samples are required per cycle, a two-level piecewise generator can be used, reducing the RAM required per generated sample while retaining the simplicity and speed of the basic technique.
