Scalable Multi-Precision Simulation of Spiking Neural Networks on GPU with OpenCL

Transcription

1 Scalable Multi-Precision Simulation of Spiking Neural Networks on GPU with OpenCL Dmitri Yudanov (Advanced Micro Devices, USA) Leon Reznik (Rochester Institute of Technology, USA) WCCI 2012, IJCNN, June 12

2 Agenda Motivation OpenCL. SNN Simulation Platform GPU Device Architecture SNN Simulation Architecture Results: Verification and Performance Next Simulator Architecture Conclusion Q&A

3 Motivation SNN simulation scalability domains: Network size Connection count SNN component models (neuron, synapse, gap junction etc) Simulation methods (event-driven, time-driven, numerical methods) Precision Simulation flexibility and programmability for heterogeneous environment. OpenCL. Configuration: GPU Radeon HD 7970 (code-named Tahiti). OpenCL Izhikevich neuron model Parker-Sochacki simulation method

4 OpenCL. Simulation Platform Open Computing Language. Open standard maintained by Khronos Group Four models: Platform model Memory model Programming model Execution model Based on B Gaster et al. Heterogeneous Computing with OpenCL.: Morgan Kaufmann Pub, 2011.

5 Tahiti GPU Architecture: High Level View Based on AFDS11 presentation: M Houston and M Mantor. (2011, June) Fusion Developer Summit: AMD Graphics Core Next.

6 Tahiti GPU Architecture: Compute Unit Based on AFDS11 presentation: M Houston and M Mantor. (2011, June) Fusion Developer Summit: AMD Graphics Core Next.

7 Simulation: Computation Flow

8 Simulation: Update PS solver is based on sequential implementation of R Stewart and W Bair, "Spiking neural network simulation: numerical integration with the Parker-Sochacki method," Journal of Computational Neuroscience, vol. 27, no. 1, pp , August 2009.

9 Simulation: Expand

10 Simulation: Sort Radix sort example: 1 bit radix. LSD sort. Modified from T Harada and L Howes. (2011, Dec.) Heterogeneous Compute.[Online].

11 Simulation: Address

12 Results: Verification and Testbench A unit test for each kernel A unified integration test with complete host-device verification A variety of compilation modes C++ preprocessor-driven optimizations XML-driven search script for the best performing variant. User Interface: Perl script + XML Microsoft VS

13 Results: Performance Size-connection scalability in multi-precision networks with per-wf precision allocation iterations, 250 us step Randomly-connected SNN with only AMPA synapses. GPU: Radeon HD 7970, CPU: AMD Phenom II, 3.2 GHz (single thread) Network Size (neurons) Average Synapses per Neuron Average Events per Step Average Spikes per Step Total Synapse Count (millions) GPU Time per Step, (ms) CPU Time per Step, (ms) Time Ratio 2,100, ,000 2, ,000 1, , ,000 11, ,

14 Simulator: Next Architecture Out-of-order flow with event-based synchronization Target-oriented synaptic matrix partitioning Mixed hybrid and time-driven simulation flows Variety of neuron models STDP Just-in-time spike-toevent expansion

15 Conclusion Multi-precision scalable (neurons, connections, precision) SNN parallel simulator. OpenCL, Tahiti architecture. Fully verified with CPU original implementation. Up to 90x faster compared to a single thread on CPU. Future Work (in the order of importance) Object-oriented design Out-of-order execution flows STDP feature Linux support Application examples User interface (possibly a library with extensions to PyNN) APU support Other: root-cause Newton-Raphson divergence, just-in-time spike-to-event expansion, sort radix scalability.

16 Q&A Selected Bibliography R. Stewart and W. Bair, "Spiking neural network simulation: numerical integration with the Parker-Sochacki method," Journal of Computational Neuroscience, vol. 27, no. 1, pp , Aug E. M. Izhikevich, "Simple model of spiking neurons," Neural Networks, IEEE Transactions on, vol. 14, pp , B Gaster, D R Kaeli, L Howes, and P Mistry, Heterogeneous Computing with OpenCL.: Morgan Kaufmann Pub, T Harada and L Howes. (2011, Dec.) Introduction to GPU Radix Sort. Heterogeneous Compute. [Online]. M Houston and M Mantor. (2011, June) Fusion Developer Summit: AMD Graphics Core Next. [Online]. D Yudanov, M Shaaban, R Melton, and L Reznik, "GPU-based simulation of spiking neural networks with real-time performance & high accuracy," in The 2010 International Joint Conference on Neural Networks (IJCNN), 2010, pp Code: Thanks to Lee Howes, Dr. Wu-Chun Feng