Implementation efficiency of BLAKE and other contemporary hash algorithms in popular FPGA devices

Abstract

BLAKE is a cryptographic hash function which was one of the 5 finalists in the SHA-3 competition. Although it ultimately lost to KECCAK the cipher was repeatedly acclaimed during evaluation for its both good cryptographic strength and great performance especially in software realizations, and it is still selected as a hash function of choice in contemporary computer systems. The aim of this paper is to investigate how the elaborated particularities of BLAKE internal processing and data multiplexing are handled by automatic implementations tools when two low-cost FPGA platforms are targeted—the standard, well established Spartan-3 and the newer Spartan-6 devices from Xilinx, Inc. The cipher was implemented in high-speed architectures built from the standard iterative one with loop unrolling and (optionally) pipelining. Total of 5 different organizations were created and results of their implementations are compared with analogous results obtained for another two contemporary hash algorithms: Salsa20, which inspired BLAKE core processing, and KECCAK. Presented data illustrate how the fundamental mechanisms of loop unrolling with or without pipelining work in case of this particular cipher.