BitMAC: Bit-Serial Computation-Based Efficient Multiply-Accumulate Unit for DNN Accelerator

Author(s):  
Harsh Chhajed ◽  
Gopal Raut ◽  
Narendra Dhakad ◽  
Sudheer Vishwakarma ◽  
Santosh Kumar Vishvakarma
Keyword(s):  
1992 ◽  
Vol 139 (3) ◽  
pp. 230 ◽  
Author(s):  
M.A. Hasan ◽  
V.K. Bhargava
Keyword(s):  

Author(s):  
Sergio Roldán Lombardía ◽  
Fatih Balli ◽  
Subhadeep Banik

AbstractRecently, cryptographic literature has seen new block cipher designs such as , or that aim to be more lightweight than the current standard, i.e., . Even though family of block ciphers were designed two decades ago, they still remain as the de facto encryption standard, with being the most widely deployed variant. In this work, we revisit the combined one-in-all implementation of the family, namely both encryption and decryption of each as a single ASIC circuit. A preliminary version appeared in Africacrypt 2019 by Balli and Banik, where the authors design a byte-serial circuit with such functionality. We improve on their work by reducing the size of the compact circuit to 2268 GE through 1-bit-serial implementation, which achieves 38% reduction in area. We also report stand-alone bit-serial versions of the circuit, targeting only a subset of modes and versions, e.g., and . Our results imply that, in terms of area, and can easily compete with the larger members of recently designed family, e.g., , . Thus, our implementations can be used interchangeably inside authenticated encryption candidates such as , or in place of .


2010 ◽  
Vol 19 (08) ◽  
pp. 1665-1687 ◽  
Author(s):  
MOHAMMAD REZA HOSSEINY FATEMI ◽  
HASAN F. ATES ◽  
ROSLI SALLEH

The sub-pixel motion estimation (SME), together with the interpolation of reference frames, is a computationally extensive part of the H.264 encoder that increases the memory requirement 16-times for each reference frame. Due to the huge computational complexity and memory requirement of the H.264 SME, its hardware architecture design is an important issue especially in high resolution or low power applications. To solve the above difficulties, we propose several optimization techniques in both algorithm and architecture levels. In the algorithm level, we propose a parabolic based algorithm for SME with quarter-pixel accuracy which reduces the computational budget by 94.35% and the memory access requirement by 98.5% in comparison to the standard interpolate and search method. In addition, a fast version of the proposed algorithm is presented that reduces the computational budget 46.28% further while maintaining the video quality. In the architecture level, we propose a novel bit-serial architecture for our algorithm. Due to advantages of the bit-serial architecture, it has a low gate count, high speed operation frequency, low density interconnection, and a reduced number of I/O pins. Also, several optimization techniques including the sum of absolute differences truncation, source sharing exploiting and power saving techniques are applied to the proposed architecture which reduce power consumption and area. Our design can save between 57.71–90.01% of area cost and improves the macroblock (MB) processing speed between 1.7–8.44 times when compared to previous designs. Implementation results show that our design can support real time HD1080 format with 20.3 k gate counts at the operation frequency of 144.9 MHz.


Sign in / Sign up

Export Citation Format

Share Document