Requirements for embedded systems are very stringent and sometimes conflicting. The Traditional software-only methods (TSOMs) for measuring the execution time of real-time code are easy to use and low in cost, but provide low resolution and great overhead. The two drawbacks degrade accuracy of analyzing real-time performance of embedded systems. Because only the indispensable resources are employed, side effects of the drawbacks on analysis results are even more. So TSOMs don't qualify for accurately analyzing real-time performance, such as identifying whether a task set is schedulable. In order to overcome the two main obstacles, this paper presents a software-only method, which takes the information fusion idea of multi-source data to improve accuracy of the measurements, while it reserves the good features owned by traditional ones. This paper uses the measurements got by a hardware method as criterion to evaluate the softwareonly method experimentally. And the result shows the most deviation score is about 1%.
展开▼
