An embedded system should react to events in its environment before the end of the determined time. The system is notified about the event occurrence usually by the interrupt request signal. The system software can be designed in such a way that only the event flag is set in the interrupt service routine, whereas the event service is carried out in the routine invoked in the main loop. If the time requirements are difficult or impossible to meet by the software of such structure, the multitasking preemptive kernel is used [1]. The system software is divided into independent tasks of various degrees of importance. The kernel ensures execution of more important tasks before less important ones. However, the kernel code occupies some fixed portion of the program and data memory and its execution takes the extra CPU time. This paper presents results of the research work whose purpose was to determine the size of such overheads introduced by Xilkernel, the multitasking kernel developed by Xilinx [2]. The Virtex-II Pro Development System was used to conduct experiments (Fig. 1). In the FPGA chip, contained in this board, a microprocessor system based on the PowerPC 405 core (one of two in the chip) was implemented (Fig. 2). There were measured the interrupt latency, the task latency and the execution time of kernel services. The results are given in Tables 1 and 2. The results should help embedded systems developer to assess whether a system designed can fulfill the timing requirements.
展开▼
