Cnvlutin
Authors:
Top Cited Papers
Open Access
- 18 June 2016
- journal article
- conference paper
- Published by Association for Computing Machinery (ACM) in ACM SIGARCH Computer Architecture News
- Vol. 44 (3) , 1-13
Abstract
This work observes that a large fraction of the computations performed by Deep Neural Networks (DNNs) are intrinsically ineffectual as they involve a multiplication where one of the inputs is zero. This observation motivates Cnvlutin ( CNV ), a value-based approach to hardware acceleration that eliminates most of these ineffectual operations, improving performance and energy over a state-of-the-art accelerator with no accuracy loss. CNV uses hierarchical data-parallel units, allowing groups of lanes to proceed mostly independently enabling them to skip over the ineffectual computations. A co-designed data storage format encodes the computation elimination decisions taking them off the critical path while avoiding control divergence in the data parallel units. Combined, the units and the data storage format result in a data-parallel architecture that maintains wide, aligned accesses to its memory hierarchy and that keeps its data lanes busy. By loosening the ineffectual computation identification criterion, CNV enables further performance and energy efficiency improvements, and more so if a loss in accuracy is acceptable. Experimental measurements over a set of state-of-the-art DNNs for image classification show that CNV improves performance over a state-of-the-art accelerator from 1.24× to 1.55× and by 1.37× on average without any loss in accuracy by removing zero-valued operand multiplications alone. While CNV incurs an area overhead of 4.49%, it improves overall EDP (Energy Delay Product) and ED 2 P (Energy Delay Squared Product) on average by 1.47× and 2.01×, respectively. The average performance improvements increase to 1.52× without any loss in accuracy with a broader ineffectual identification policy. Further improvements are demonstrated with a loss in accuracy.Keywords
This publication has 33 references indexed in Scilit:
- Deep learningNature, 2015
- ImageNet Large Scale Visual Recognition ChallengeInternational Journal of Computer Vision, 2015
- DESTINY: A Tool for Modeling Emerging 3D NVM and eDRAM caches2015
- A hardware–software co-design approach for implementing sparse matrix vector multiplication on FPGAsMicroprocessors and Microsystems, 2014
- Return of the Devil in the Details: Delving Deep into Convolutional Nets2014
- Efficient sparse matrix-vector multiplication on x86-based many-core processors2013
- Improving GPU performance via large warps and two-level warp scheduling2011
- Et2: A Metric for Time and Energy Efficiency of Computation2002
- Energy dissipation in general purpose microprocessorsIEEE Journal of Solid-State Circuits, 1996
- Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in positionBiological Cybernetics, 1980