On the Simulation of Software-Driven Stochastic Computing for Emerging Applications


Aygün S. , Güneş E. O.

Design, Automation and Test in Europe Conference (DATE), SCONA Workshop, Grenoble, France, 13 March 2020, pp.1-2

  • Publication Type: Conference Paper / Full Text
  • City: Grenoble
  • Country: France
  • Page Numbers: pp.1-2

Abstract

Stochastic computing (SC) is a re-emerging computation methodology that uses the hardware-based approaches in the presence of bit-streams. In recent years, stochastic computing is being used from communication systems to the neural networks including hardware design issues, spintronic device & memory modeling, image processing and finally the deep learning. SC is basically advantageous for the area and power consumption criterion on the hardware, however, especially for applications such as computer vision and neural systems, the deep investigation and a flexible environment for algorithm modeling is required. Therefore, this study presents the platform-depended implementation notes which are so far obtained during the current efforts on SC. In the literature, there are examples of very large scale integrated (VLSI) circuit design efforts to implement stochastic computing systems, but, for the more flexible environment, there are fewer efforts to define any simulation related solutions. There is an increasing trend in the learning systems which have recently growing parameter size of millions. Therefore, the computation and the simulation environments become so crucial especially if the hardware perspective like SC is in question. In the following, after the introduction and literature review, hands-on remarks of the hardware-software co-operated SC simulation environment can be found.