Abstract
Nonuniform random numbers are key for many technical applications, and designing efficient hardware implementations of non-uniform random number generators is a very active research field. However, most state-of-the-art architectures are either tailored to specific distributions or use up a lot of hardware resources. At ReConFig 2010, we have presented a new design that saves up to 48% of area compared to state-of-the-art inversion-based implementation, usable for arbitrary distributions and precision. In this paper, we introduce a more flexible version together with a refined segmentation scheme that allows to further reduce the approximation error significantly. We provide a free software tool allowing users to implement their own distributions easily, and we have tested our random number generator thoroughly by statistic analysis and two application tests.
Highlights
The fast generation of random numbers is essential for many tasks
Many simulations are still performed on highperformance CPU or general-purpose graphics processing unit (GPGPU) clusters, using reconfigurable hardware accelerators based on field programmable gate arrays (FPGAs) can save up to at least one order of magnitude of power consumption if the random number generator (RNG) is located on the accelerator
Since one single instance of our proposed hardware design consumes less than 1% of the area on the used Xilinx Virtex5 FPGA, we have introduced a line with the extrapolated values for 100 instances to highlight the enormous potential of hardware accelerators with respect to the achievable throughput per energy
Summary
The fast generation of random numbers is essential for many tasks. One of the major fields of application are Monte Carlo simulation, for example widely used in the areas of financial mathematics and communication technology. Since one single instance of our proposed hardware design (together with a uniform random number generator) consumes less than 1% of the area on the used Xilinx Virtex FPGA, we have introduced a line with the extrapolated values for 100 instances to highlight the enormous potential of hardware accelerators with respect to the achievable throughput per energy. We present a refined version of the floating point-based nonuniform random number generator already shown at ReConFig 2010 [1]. The modifications allow a higher precision while having an even lower area consumption compared to the previous results This is due to a refined synthesis.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
