A Data Flow Program Executor Study Based on the Petri Nets

• Main Authors: Wei-Min Lin, 林偉民
• Other Authors: Chi-Pan Hwang
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/78966346655667110444

碩士 === 國立彰化師範大學 === 電子工程學系 === 97 === A Petri net executor has been devised to own the performance of adaptive computing system that is implemented on the FPGA devices based on the data flow computing model. It is expected to break through the von Neumann bottleneck of the traditional computers based on control flow computing model. There a C compiler has been developed to generate an intermediate code represented by the Petri net based on the data flow model. It is downloaded to the Petri net executor to complete the program execution. The Petri net executor consists of triggering, scheduling, and executing hardware circuits that form an execution cycle. The realization of triggering and scheduling circuits is referred to the CAM technique. The functions of triggering circuit are token consuming, marking updating, triggerable-transitions matching, and exclusive transitions determination. The scheduling circuit generates a set of firable transitions refereeing the outputs from the triggering circuit and delivers to the executing circuit. The executing circuit is a group of precasted hardware computing circuits corresponding to the preseted operations of transitions to complete the program execution, specifically. Furthermore, the executing circuit reports the fired transitions to the triggering circuit to renew the execution cycle. There are fixed number of FPGA gates to realize the hardware circuits of Petri net executor. However, the adaptive computing system creates a disappointment which hardware gate counts depending on the program size are not easy to control. Owing to the limited computing circuits and cyclic operations of Petri net executor, its performance is inferior to adaptive computing system. But the performance of Petri net executor is far to surpass the traditional computing system based on the control flow operations that have been evaluated by the benchmark programs.