2016/12/05 by Georgios Smaragdos, Smaragdos, Georgios, Georgios Chatzikonstantis +19
Computer Science · Engineering · Neuroscience · #Advanced Memory and Neural Computing #Distributed #FOS: Computer and information sciences #Neural Networks and Reservoir Computing #Neural and Evolutionary Computing (cs.NE) #Neural dynamics and brain function #Parallel #Parallel Computing and Optimization Techniques #and Cluster Computing (cs.DC)
paper · pdf · doi:10.48550/arxiv.1612.01501
openalex publication_date 2016/12/05 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/28
Objective: The advent of High-Performance Computing (HPC) in recent years has\nled to its increasing use in brain study through computational models. The\nscale and complexity of such models are constantly increasing, leading to\nchallenging computational requirements. Even though modern HPC platforms can\noften deal with such challenges, the vast diversity of the modeling field does\nnot permit for a single acceleration (or homogeneous) platform to effectively\naddress the complete array of modeling requirements. Approach: In this paper we\npropose and build BrainFrame, a heterogeneous acceleration platform,\nincorporating three distinct acceleration technologies, a Dataflow Engine, a\nXeon Phi and a GP-GPU. The PyNN framework is also integrated into the platform.\nAs a challenging proof of concept, we analyze the performance of BrainFrame on\ndifferent instances of a state-of-the-art neuron model, modeling the Inferior-\nOlivary Nucleus using a biophysically-meaningful, extended Hodgkin-Huxley\nrepresentation. The model instances take into account not only the neuronal-\nnetwork dimensions but also different network-connectivity circumstances that\ncan drastically change application workload characteristics. Main results: The\nsynthetic approach of three HPC technologies demonstrated that BrainFrame is\nbetter able to cope with the modeling diversity encountered. Our performance\nanalysis shows clearly that the model directly affect performance and all three\ntechnologies are required to cope with all the model use cases.\n