ABSTRACT
Neuromorphic computing is recently gaining significant attention as a promising candidate to conquer the well-known von Neumann bottleneck. In this work, we propose RENO -- a efficient reconfigurable neuromorphic computing accelerator. RENO leverages the extremely efficient mixed-signal computation capability of memristor-based crossbar (MBC) arrays to speedup the executions of artificial neural networks (ANNs). The hierarchically arranged MBC arrays can be configured to a variety of ANN topologies through a mixed-signal interconnection network (M-Net). Simulation results on seven ANN applications show that compared to the baseline general-purpose processor, RENO can achieve on average 178.4x (27.06x) performance speedup and 184.2x (25.23x) energy savings in high-efficient multilayer perception (high-accurate auto-associative memory) implementation. Moreover, in the comparison to a pure digital neural processing unit (D-NPU) and a design with MBC arrays co-operating through a digital interconnection network, RENO still achieves the fastest execution time and the lowest energy consumption with similar computation accuracy.
- "Macsim," http://code.google.com/p/macsim/.Google Scholar
- "The mnist database," http://yann.lecun.com/exdb/mnist/.Google Scholar
- "Uci machine learning," http://archive.ics.uci.edu/ml/.Google Scholar
- F. Alibart et al., "High precision tuning of state for memristive devices by adaptable variation-tolerant algorithm," Nanotechnology, vol. 23, no. 7, 2012.Google Scholar
- B. Belhadj et al., "Continuous real-world inputs can open up alternative accelerator designs," in ISCA, 2013, pp. 1--12. Google ScholarDigital Library
- L. O. Chua, "Memristor-the missing circuit element," Circuit Theory, vol. 18, no. 5, pp. 507--519, 1971.Google ScholarCross Ref
- L. Dai and R. Harjani, "Cmos switched-op-amp-based sample-and-hold circuit," in IEEE Transactions on Solid-state circuits, 2000.Google Scholar
- H. Esmaeilzadeh et al., "Neural acceleration for general-purpose approximate programs," in MICRO, 2012, pp. 449--460. Google ScholarDigital Library
- J. Gu et al., "Implementation and evaluation of deep neural networks (dnn) on mainstream heterogeneous systems," in APSys, 2014, p. 12. Google ScholarDigital Library
- M. Gustavsson, J. J. Wikner, and N. Tan, CMOS data converters for communications, 2000. Google ScholarDigital Library
- S. O. Haykin, Neural Networks and Learning Machines. London: Prentice Hall, 2008.Google Scholar
- M. Hu et al., "Hardware realization of bsb recall function using memristor crossbar arrays," in DAC, 2012, pp. 498--503. Google ScholarDigital Library
- N. Jian et al., "A detailed and flexible cycle-accurate network-on-chip simulator," in ISPASS, 2013, pp. 86--96.Google Scholar
- K.-H. Kim et al., "A functional hybrid memristor crossbar-array/cmos system for data storage and neuromorphic applications," Nano letters, vol. 12, no. 1, pp. 389--395, 2011.Google ScholarCross Ref
- S. Li et al., "Mcpat: an integrated power, area, and timing modeling framework for multicore and manycore architectures," in MICRO, 2009, pp. 469--480. Google ScholarDigital Library
- B. Liu et al., "Digital assisted noise eliminating training for memristor crossbar based analog neuromorphic computing engine," in DAC, 2013, pp. 1--6. Google ScholarDigital Library
- X. Liu et al., "A heterogeneous computing system with memristor-based neuromorphic accelerators," in HPEC, 2014, pp. 1--6.Google Scholar
- L. Prechelt, "Proben1-a set of neural network benchmark problems and benchmarking rules," University of Karlsruhe, Tech. Rep., 1994.Google Scholar
- O. Temam, "A defect-tolerant accelerator for emerging high-performance applications," in ISCA, 2012, pp. 356--367. Google ScholarDigital Library
Index Terms
- RENO: a high-efficient reconfigurable neuromorphic computing accelerator design
Recommendations
RENO: A Rename-Based Instruction Optimizer
ISCA 2005RENO is a modified MIPS R10000 register renamer that uses map-table "short-circuiting" to implement dynamic versions of several well-known static optimizations: move elimination, common subexpression elimination, register allocation, and constant ...
RENO: A Rename-Based Instruction Optimizer
ISCA '05: Proceedings of the 32nd annual international symposium on Computer ArchitectureRENO is a modified MIPS R10000 register renamer that uses map-table "short-circuiting" to implement dynamic versions of several well-known static optimizations: move elimination, common subexpression elimination, register allocation, and constant ...
Comments