Improving performance and efficiency of PRNG

[visitorckw]

We've recently integrated the MT19937 algorithm into our project for pseudo-random number generation. However, it's crucial to assess its performance and efficiency.

Quoting from @jserv :
With full support from processor hardware, achieving around 60 Gbits/s with AES, supplemented by intermittent reseeds from RDRAND, is achievable. When implemented purely in software, we can approach similar performance levels for algorithms such as SplitMix64, Xoroshiro128+, and PCG64. Additionally, with the utilization of vector processing capabilities, ChaCha20 offers competitive speeds and outperforms algorithms like the Mersenne Twister.

Action items:

• Compare the performance and efficiency of MT19937 against alternative algorithms like SplitMix64, Xoroshiro128+, PCG64, and ChaCha20.
• Investigate potential optimizations or enhancements to MT19937's implementation to improve its performance and efficiency further.
• Consider the implications of adopting MT19937 versus alternative algorithms in terms of speed, efficiency, and suitability for the project's requirements.

Refs:

• The fastest conventional random number generator that can pass Big Crush?
• Comparison of hashing functions with small code size

See: #21