As a typical artificial intelligence algorithm, the convolutional neural network (CNN) is widely used in the Internet of Things (IoT) system. In order to improve the computing ability of an IoT CPU, this paper designs a reconfigurable CNN-accelerated coprocessor based on the RISC-V instruction set. The interconnection structure of the acceleration chain designed by the predecessors is optimized, and the accelerator is connected to the RISC-V CPU core in the form of a coprocessor. The corresponding instruction of the coprocessor is designed and the instruction compiling environment is established. Through the inline assembly in the C language, the coprocessor instructions are called, coprocessor acceleration library functions are established, and common algorithms in the IoT system are implemented on the coprocessor. Finally, resource consumption evaluation and performance analysis of the coprocessor are completed on a Xilinx FPGA. The evaluation results show that the reconfigurable CNN-accelerated coprocessor only consumes 8534 LUTS, accounting for 47.6% of the total SoC system. The number of instruction cycles required to implement functions such as convolution and pooling based on the designed coprocessor instructions is better than using the standard instruction set, and the acceleration ratio of convolution is 6.27 times that of the standard instruction set.
As a classical artificial intelligence algorithm, the convolutional neural network (CNN) algorithm plays an important role in image recognition and classification and is gradually being applied in the Internet of Things (IoT) system. A compact CNN accelerator for the IoT endpoint System-on-Chip (SoC) is proposed in this paper to meet the needs of CNN computations. Based on analysis of the CNN structure, basic functional modules of CNN such as convolution circuit and pooling circuit with a low data bandwidth and a smaller area are designed, and an accelerator is constructed in the form of four acceleration chains. After the acceleration unit design is completed, the Cortex-M3 is used to construct a verification SoC and the designed verification platform is implemented on the FPGA to evaluate the resource consumption and performance analysis of the CNN accelerator. The CNN accelerator achieved a throughput of 6.54 GOPS (giga operations per second) by consuming 4901 LUTs without using any hardware multipliers. The comparison shows that the compact accelerator proposed in this paper makes the CNN computational power of the SoC based on the Cortex-M3 kernel two times higher than the quad-core Cortex-A7 SoC and 67% of the computational power of eight-core Cortex-A53 SoC.
One of the well‐known physical attacks, i.e. differential fault analysis (DFA), can break the secret key of cryptographic device by using differential information between faulty and correct ciphertexts. Here, the authors propose a random 2‐byte fault model, present a novel DFA on AES key schedule, and show how an entire AES‐128 key can be cracked by using two pairs of faulty and correct ciphertexts. By inducing a random 2‐byte fault in the first column of 9th round key with discontiguous rows, the authors can obtain 64 bits of AES‐128 key using one pair of faulty and correct ciphertexts, two pairs of them can retrieve the entire 128‐bit key without exhaustive search. The authors implement the proposed attack on HP Intel(R) Core i5‐7300HQ Quad‐Core 2.5 GHz CPU, 8G RAM. It takes <2 min on average to break the key. Considering the number of faulty ciphertexts, fault‐induced depth, and fault model, authors’ attack is the most efficient DFA as compared to existing schemes on AES‐128 key schedule.
S-Box based on Composite field arithmetic (CFA) technology is optimized by Genetic algorithm (GA) and Cartesian genetic programming (CGP) model for reducing the hardware complexity. After using the CFA technique to map Multiplicative inverse (MI) over GF(28) into composite field GF((24)2), the compact MI circuit over GF(24) is selected from 100 evolved circuits, and same design method is applied to the compact multiplication circuit over GF(22). Compared with the direct implementations, the areas of optimized circuits of MI over GF(24) and multiplication over GF((22)2) are reduced by 66% and 57.69%, respectively. The area reductions for MI over GF(28) and the whole of S-Box are up to 59.23% and 56.14%, separately. In 180nm 1.8V COMS technology, compared to previous works, the S-Box proposed in this paper has the minimum area and minimum power, which are 11.27% and 6.65% smaller than that of the smallest area S-Box, respectively.
Optical networks on chip (ONoC) delivers a promising alternative to meet growing needs of higher bandwidth and low power consumption in manycore processors. Optical routers are the key element in ONoCs that significantly affect the performance of overall network. In this letter, we propose a rearrangeable non-blocking 6x6 router (RoR) constructed with 2x2 hybrid photonic-plasmonic switching (HPPS) elements. Router architectures with 15 HPPS elements and an optimized design using reduced HPPS elements are presented and analyzed. In optimized form, proposed design consumes only 12 HPPS elements which results in low insertion loss and crosstalk noise in comparison to the un-optimized architecture. We observe up to 50% reduction in switching elements count in comparison to other router architecture of same radix.
13 July 2019 in Journal of Electrical Engineering & Technology