PPE-ARX: Area- and power-efficient VLIW programmable processing element for IoT crypto-systems

In this paper, we propose a novel programmable processing element (PPE) for various cryptographic systems that can be used in IoT applications. The design enables the programmability, thus supporting a wide range of bit-widths (such as 16, 32, and 64). It employs a very long instruction word (VLIW) architecture with an instruction set and memory hierarchy specialized for crypto-processing. Both FPGA and ASIC implementations demonstrate that the design utilizes a very tiny area and consumes very low power. For example, it takes only 227 slices for FPGA implementations to include 512-byte instruction and coefficient memory along with the computational unit by achieving a maximum clock frequency of 250 MHz. For ASIC implementation (in 28/32nm technology), the design takes only 0.15mm
²
of silicon area and consumes only 34.5μW of total power while achieving a maximum frequency of 952MHz. To evaluate the effectiveness of the design in a larger system, we implement Blue Midnight Wish (BMW) hash function with the PPE. Compared to the previous BMW-512 implementation which stores the intermediate coefficients of the BMW-512 in 2048 bytes, the proposed design just uses 512 bytes. Meanwhile, we reduce the instruction memory size from 4864 bytes to 1792 bytes.