Some experimental results on Kilo/Mega Core Architectures

Inhalt:
Our community joined to Cellular Neural//Nonlinear Network (CNN) research in a very early phase. As the first steps some CNN simulators were developed to analyze the capabilities of the CNN architectures. One of the most attractive feature of this computing paradigm is the parallel processing and thank to this fact the very high computing power in analog VLSI implementations. An industrial camera computer SPS02 have been implemented by Toshiba based on the research chip designed by Li˜nan et. al. To overcome some limits of analogue technology in silicon the emulated digital solution came into the picture, allowing nonlinear, space variant templates, higher accuracy etc. But the computing power were lower, only a small array of CNN processors could be implemented in a reasonable size of chip.