BSRNG: A High Throughput Parallel BitSliced Approach for Random Number Generators

In this work, a high throughput method for generating high-quality Pseudo-Random Numbers using the bitslicing technique is proposed. In such a technique, instead of the conventional row-major data representation, column-major data representation is employed, which allows the bitslicing implementation to take full advantage of all the available datapath of the hardware platform. By employing this data representation as building blocks of algorithms, we showcase the capability and scalability of our proposed method in various PRNG methods in the category of block and stream ciphers. The LFSR-based (Linear Feedback Shift Register) nature of the PRNG in our implementation perfectly suits the GPU’s many-core structure due to its register oriented architecture. In the proposed SIMD vectorized GPU implementation, each GPU thread can generate several 32 pseudo-random bits in each LFSR clock cycle. We then compare our implementation with some of the most significant PRNGs that display a satisfactory performance throughput and randomness criteria. The proposed implementation successfully passes the NIST test for statistical randomness and bit-wise correlation criteria. For computer-based PRNG and the optical solutions in terms of performance and performance per cost, this technique is efficient while maintaining an acceptable randomness measure. Our highest performance among all of the implemented CPRNGs with the proposed method is achieved by the MICKEY 2.0 algorithm, which shows 40% improvement over state of the art NVIDIA’s proprietary high-performance PRNG, cuRAND library, achieving 2.72 Tb/s of throughput on the affordable NVIDIA GTX 2080 Ti.