Comparison of power consumption in pipelined implementations of the BLAKE3 cipher in FPGA devices

This article analyzes the dynamic power losses generated by various hardware implementations of the BLAKE3 hash function. Estimations of the parameters were based on the results of post-route simulations of designs implemented in Xilinx Spartan-7 FPGAs. The algorithm was tested in various hardware organizations: based on a standard iterative architecture with one round instance in the programmable array, various derived versions with pipeline processing were elaborated, which ultimately led to a set of 6 architectural variants of the cipher, from the iterative case (without pipeline) to one with maximum of 6 pipeline stages. Moreover, the results obtained for the iterative architecture were compared with analogous implementations of the BLAKE2 (direct predecessor) and KECCAK (the foundation of the current SHA-3 standard) algorithms. This case study illustrates the differences (or lack thereof) in the power requirements of these three hash functions when they are implemented on an FPGA platform, and illustrate the significant savings that can be achieved by introducing pipeline to the processing of the BLAKE round.