Approximate Computing
The end of Dennard scaling and the threat of Dark Silicon are pushing the limits of existing technologies - we are fast approaching the power wall. This has motivated the development of new technologies for both logic and memory. Computational needs are increasing due to the arrival of the Internet of Things that results in big data problems. As a result, energy efficiency has become a key requirement
Today’s computing systems are designed to deliver only exact solutions at high energy cost, while many of the algorithms that are run on data are at their heart statistical, and thus do not require exact answers. We propose to design and implement approximate computing solution that provide an end to end framework which is able to optimally and simultaneously trade-off accuracy and efficiency across software and hardware stacks of IoT applications. We propose a novel architecture with software and hardware support for approximate computing. Hardware components are enhanced with ability to dynamically adapt approximation at a quantifiable and controllable cost in terms of accuracy. Software services complement hardware to ensure user’s perception is not compromised, while maximizing the energy savings due to approximations. The changes to hardware design include approximation enabled CPU, GPU, accelerators (DSPs) and storage. Our work is a first promising technique which can dynamically adapt approximation to quality requirements. In this way, we also effectively address the problem of saving energy whenever it is not required to meet users’ expectations. To do this, we rely on automatic estimation of required quality based on user profile, user feedbacks and system performance metrics. Our platform considers the use of new technology-based hardware in embedded devices, such as non- volatile memories. The proposed design, which integrates all optimized software components with improved hardware designs, can result about 2x energy efficiency and 4x speedup while ensuring that the loss of accuracy is kept small enough to not be perceivable. |
Related Publications:
[ESL'17] M. Imani, D. Peroni, T. Rosing “NVALT: Approximate Lookup Table for GPU Acceleration”, IEEE Embedded System Letter (ESL), 2017.
[DAC'17] M. Imani, S. Gupta, T. Rosing “Ultra-Efficient Processing In-Memory for Data Intensive Applications”, IEEE/ACM Design Automation Conference (DAC), 2017 [PDF].
[DAC'17] M. Imani, D. Peroni, T. Rosing “CFPU: Configurable Floating Point Multiplier for Energy-Efficient Computing”, IEEE/ACM Design Automation Conference (DAC), 2017 [PDF].
[DATE'17] M. Imani, D. Peroni, Y. Kim, A. Rahimi and T. Rosing, "Efficient Neural Network Acceleration on GPGPU using Content Addressable Memory," IEEE/ACM Design Automation and Test in Europe Conference (DATE), 2017 [PDF].
[DATE'17] M. Samragh, M. Imani, F. Koushanfar and T. Rosing, "LookNN: Neural Network with No Multiplication," IEEE/ACM Design Automation and Test in Europe Conference (DATE), 2017 [PDF].
[TMSCS'17] M. Imani, A. Rahimi, P. Mercati, T. Rosing, “Multi-stage Tunable Approximate Search in Resistive Associative Memory” IEEE Transactions on Multi-Scale Computing Systems (TMSCS), 2017 [PDF].
[ISQED'17] M. Imani, T. Rosing, "CAP: Configurable Resistive Associative Processor for Near-Data Computing," IEEE International Symposium on Quality Electronic Design (ISQED), 2017 [PDF].
[TETC'17] M. Imani, D. Peroni, A. Rahimi, T. Rosing, “Resistive CAM Acceleration for Tunable Approximate Computing” IEEE Transactions on Emerging Topics in Computing (TETC), 2017 [PDF].
[NVMW'17] M. Imani, D. Peroni, A. Rahimi, T. Rosing, “Non-volatile Content Addressable Memory for Computing Acceleration” Non-Volatile Memory Workshop (NVMW), 2017.
[ICCD'16] M. Imani, D. Peroni, A. Rahimi, T. Rosing, “Resistive CAM Acceleration for Tunable Approximate Computing” in IEEE International Conference on Computer Design (ICCD), 2016 (selected as top ranked paper for publishing in IEEE TETC).
[ISLPED'16] M. Imani, Y. Kim, A. Rahimi, T. Rosing, "ACAM: Approximate Computing Based on Adaptive Associative Memory with Online Learning" International Symposium on Low Power Electronics and Design (ISLPED), 2016 [PDF].
[TETC'16] M. Imani, Shruti Patil, T. Rosing, "Approximate Computing using Multiple-Access Single-Charge Associative Memory" IEEE Transaction on Emerging Topics in Computing (TETC), 2016 [PDF].
[DATE'16] M. Imani, A. Rahimi, T. Rosing, "Resistive Configurable Associative Memory for Approximate Computing" IEEE/ACM Design Automation and Test in Europe Conference (DATE), 2016 [PDF].
[DATE'16] M. Imani, S. Patil, T. Rosing, "MASC: Ultra-Low Energy Multiple-Access Single-Charge TCAM for Approximate Computing" IEEE/ACM Design Automation and Test in Europe Conference (DATE), 2016 [PDF].
[MEMSYS'16] M. Imani, Y. Cheng, T. Rosing, "Processing Acceleration with Resistive Memory-based Computation" ACM International Symposium on Memory Systems (MEMSYS), 2016 [PDF].
[ISQED'16] M. Imani, P. Mercati, T. Rosing, "ReMAM: Low Energy Resistive Multi-Stage Associative Memory for Energy Efficient Computing" IEEE International Symposium on Quality Electronic Design (ISQED), 2016 [PDF].
[ESL'17] M. Imani, D. Peroni, T. Rosing “NVALT: Approximate Lookup Table for GPU Acceleration”, IEEE Embedded System Letter (ESL), 2017.
[DAC'17] M. Imani, S. Gupta, T. Rosing “Ultra-Efficient Processing In-Memory for Data Intensive Applications”, IEEE/ACM Design Automation Conference (DAC), 2017 [PDF].
[DAC'17] M. Imani, D. Peroni, T. Rosing “CFPU: Configurable Floating Point Multiplier for Energy-Efficient Computing”, IEEE/ACM Design Automation Conference (DAC), 2017 [PDF].
[DATE'17] M. Imani, D. Peroni, Y. Kim, A. Rahimi and T. Rosing, "Efficient Neural Network Acceleration on GPGPU using Content Addressable Memory," IEEE/ACM Design Automation and Test in Europe Conference (DATE), 2017 [PDF].
[DATE'17] M. Samragh, M. Imani, F. Koushanfar and T. Rosing, "LookNN: Neural Network with No Multiplication," IEEE/ACM Design Automation and Test in Europe Conference (DATE), 2017 [PDF].
[TMSCS'17] M. Imani, A. Rahimi, P. Mercati, T. Rosing, “Multi-stage Tunable Approximate Search in Resistive Associative Memory” IEEE Transactions on Multi-Scale Computing Systems (TMSCS), 2017 [PDF].
[ISQED'17] M. Imani, T. Rosing, "CAP: Configurable Resistive Associative Processor for Near-Data Computing," IEEE International Symposium on Quality Electronic Design (ISQED), 2017 [PDF].
[TETC'17] M. Imani, D. Peroni, A. Rahimi, T. Rosing, “Resistive CAM Acceleration for Tunable Approximate Computing” IEEE Transactions on Emerging Topics in Computing (TETC), 2017 [PDF].
[NVMW'17] M. Imani, D. Peroni, A. Rahimi, T. Rosing, “Non-volatile Content Addressable Memory for Computing Acceleration” Non-Volatile Memory Workshop (NVMW), 2017.
[ICCD'16] M. Imani, D. Peroni, A. Rahimi, T. Rosing, “Resistive CAM Acceleration for Tunable Approximate Computing” in IEEE International Conference on Computer Design (ICCD), 2016 (selected as top ranked paper for publishing in IEEE TETC).
[ISLPED'16] M. Imani, Y. Kim, A. Rahimi, T. Rosing, "ACAM: Approximate Computing Based on Adaptive Associative Memory with Online Learning" International Symposium on Low Power Electronics and Design (ISLPED), 2016 [PDF].
[TETC'16] M. Imani, Shruti Patil, T. Rosing, "Approximate Computing using Multiple-Access Single-Charge Associative Memory" IEEE Transaction on Emerging Topics in Computing (TETC), 2016 [PDF].
[DATE'16] M. Imani, A. Rahimi, T. Rosing, "Resistive Configurable Associative Memory for Approximate Computing" IEEE/ACM Design Automation and Test in Europe Conference (DATE), 2016 [PDF].
[DATE'16] M. Imani, S. Patil, T. Rosing, "MASC: Ultra-Low Energy Multiple-Access Single-Charge TCAM for Approximate Computing" IEEE/ACM Design Automation and Test in Europe Conference (DATE), 2016 [PDF].
[MEMSYS'16] M. Imani, Y. Cheng, T. Rosing, "Processing Acceleration with Resistive Memory-based Computation" ACM International Symposium on Memory Systems (MEMSYS), 2016 [PDF].
[ISQED'16] M. Imani, P. Mercati, T. Rosing, "ReMAM: Low Energy Resistive Multi-Stage Associative Memory for Energy Efficient Computing" IEEE International Symposium on Quality Electronic Design (ISQED), 2016 [PDF].