Accelerating Hash Computations Through Efficient Instruction-Set Customisation
Mayuran Sivanesan, Anupam Chattopadhyay, Bajaj Ronak · 2018
Security has become a part of our everyday life due to the breakthrough technologies of wireless communication, e-commerce and digital content delivery. For ensuring authenticity of the transactions, cryptographic hash algorithms, such as SHA-1/2/3 are standardised in different protocols. In this paper, we undertake a detailed instruction-set customization study for accelerating SHA algorithms on a commercially available customisable processor, Tensilica Xtensa core. We have explored the bottlenecks of implementations on Tensilica Xtensa and designed efficient custom instructions, resulting in 30% improvement in performance compared to implementations using generic instructions. Furthermore, we highlight the usage of SHA-256 as a widely used cryptographic hash function for blockchain and cryptocurrency mining.