A Hybrid Approximate Computing Approach for Associative In-Memory Processors
A Hybrid Approximate Computing Approach for Associative In-Memory Processors
复制标题
关联内存处理器的混合近似计算方法
DOI：
--
复制
发表时间：
2018
影响因子：
4.6
通讯作者：
F. Kurdahi
中科院分区：
文献类型：
--
作者：
Hasan Erdem Yantır;A. Eltawil;F. Kurdahi
计算问题的复杂性比计算平台的能力上升得更快，计算平台的能力也由于其对能量的需求增加而变得越来越昂贵。这迫使研究人员寻找替代的范式和方法，以提高计算效率。一个很有前途的范例是使用内存计算加速器来加速计算密集型内核，其中数据移动显著减少。另一种越来越流行的提高能源效率的方法是近似计算。在本文中，我们提出了一种有效的近似内存计算的方法。为了最大限度地节省能源为一个给定的近似约束，一个混合的方法，提出了电压和精度缩放相结合。这可以应用于基于关联存储器的架构，该架构今天可以使用CMOS存储器（SRAM）实现，但可以无缝地扩展到新兴的基于ReRAM的存储器技术。为了评估所提出的方法，涵盖了不同的领域，如图像处理，机器学习，机器视觉和数字信号处理。与全精度、未缩放的实现相比，基于SRAM和基于ReRAM的架构分别报告<inline-formula><tex-math notation="LaTeX">了5.17美元</tex-math></inline-formula>和<inline-formula><tex-math notation="LaTeX">59.11美元的</tex-math></inline-formula>平均节能<inline-formula><tex-math notation="LaTeX">以及2.1美元</tex-math></inline-formula>和<inline-formula><tex-math notation="LaTeX">3.24美元</tex-math></inline-formula>的加速比。
The complexity of the computational problems is rising faster than the computational platforms’ capabilities which are also becoming increasingly costly to operate due to their increased need for energy. This forces researchers to find alternative paradigms and methods for efficient computing. One promising paradigm is accelerating compute-intensive kernels using in-memory computing accelerators, where data movements are significantly reduced. Another increasingly popular method for improving energy efficiency is approximate computing. In this paper, we propose a methodology for efficient approximate in-memory computing. To maximize energy savings for a given approximation constraints, a hybrid approach is presented combining both voltage and precision scaling. This can be applied to an associative memory-based architecture that can be implemented today using CMOS memories (SRAM) but can be seamlessly scaled to emerging ReRAM-based memory technology later with minimal effort. For the evaluation of the proposed methodology, a diverse set of domains is covered, such as image processing, machine learning, machine vision, and digital signal processing. When compared to full-precision, unscaled implementations, average energy savings of <inline-formula> <tex-math notation="LaTeX">$5.17{\times}$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$59.11{\times}$ </tex-math></inline-formula>, and speedups of <inline-formula> <tex-math notation="LaTeX">$2.1{\times}$ </tex-math></inline-formula> and <inline-formula> <tex-math notation="LaTeX">$3.24{\times}$ </tex-math></inline-formula> in SRAM-based and ReRAM-based architectures, respectively, are reported.
DOI：
10.1109/iccad.2017.8203756
发表时间：
2017-11
期刊：
2017 IEEE/ACM International Conference on Computer-Aided Design (ICCAD)
影响因子：
--
作者：
Yeseong Kim;M. Imani;Tajana Simunic
通讯作者：
Yeseong Kim;M. Imani;Tajana Simunic