Research Of Task Mapping For Spike Sorting System On Chip

Studying the neuron functions in the brain cortex requires recording the action potential from many neurons simultaneously. In practical, we could achieve this by transplanting electrons into the brain. However, the signals recorded by the transplanted electrons are mixture of action potentials from all neurons around the electrons. In such circumstances, analytical techniques that can differentiate action potentials from different neurons are needed. Such analytical technique is defined as spike sorting.Currently, researches about spike sorting is mainly concerned about how to improve the spike sorting algorithm to solve the big data probl em brought by the rapid development of spike signal acquisition devices, and at the same time make spike sorting more efficient and precise. But the more complex and precise the spike sorting algorithm is, the more computing ability it need and the more energy it will consume. For such reasons, implement of spike sorting system using bus based system on chip or FPGA will face great challenge.With the development of semiconductor technique, network on chip rises as a new paradigm for designing system on chip. The core idea of network on chip is try to transplant the internet techniques to the system on chip. It actually constructed a network on the chip, and the cores on the chip communicate with each other through the network. The network on chip solved the problems of global synchronous and long delay faced by the traditional bus based system on chip. In the system designed using network on chip methodology, the cores communicate with each other parallelism, the throughput of the system is much higher, and the system can be easily scaled. The network on chip is the most promising solution for multi core system on chip. Based on this trend, we consider designing system on chip for spike sorting using network methodology as a feasible solution to deal with the increasing data of spike sorting.As a part work of designing spike sorting system on chip using network on chip methodology, this article map the tasks of spike sorting onto the system on chip. In this article, we carefully analyzed the spike sorting program, and get a task graph of the spike sorting program and its data accesses during its execution. Then we scheduled the tasks of spike sorting, mapped the tasks and data blocks onto the node of the system on chip using mapping algorithm. And we also proposed an application task graph expand algorithm to expand the task graph of the application such that the tasks in the task graph will be close to the number of nodes in the system on chip. This can avoid the waste of some processors in the system on chip. Keyworks: spike sorting, network on chip, system on chip, task mapping...