干旱区生态环境知识资源中心

The goal of this project is to improve the performances of the parallel computer SYMPATI2. This SIMD processor based system performs with a good efficiency the low level image processing operations but this efficiency is drastically cut when considering intermediate level class of algorithms. A study emphasis the drawbacks encountered to perform such operations. The main one is the interconnection between processors. So, a new interconnection network, called the open intelligent network, is proposed and added to SYMPATI2 to form SYMPATIX. This network detailed below allows asynchronous transfers of data between the different processing elements of the new systeM. Furthermore this network allows the efficient interconnection of specific modules. The architecture is now evaluated on representative algorithms of image processing. To achieve this study, a behavioural model of SYMPATIX have been described using a hardware description language, the VHDL Our SIMD computer efficiency has been considerably upgraded for the low arid intermediate levels of image processing. Furthermore, its application area extended. The last part of the paper describes the performances obtained with simulations.