Design of Hardware Gausian Random Number Generator Considering Side-Effect

摘要

research on hardware accelerator has drawn more and more attention nowadays. Many compute-intensive algorithm could be accelerated here, such as Monte Carlo simulation and image processing. Gaussian Random Number Generators (GRNG) are always at the core of these algorithms. In this paper, we analysize the Side-Effect on the design of Box Muller GRNG and draw the relation between the output range of the generated Gaussian random numbers and the bit-width L of the uniform random numbers. To prove the relation drawn here, we design three different GRNGs with L=8, L=16 and L=32. The implementation results show that our analysis conclusions just match the facts.