Performance analysis of real-time embedded systems usually is based on the worst-case execution times (WCET) of the executed tasks. Nevertheless, in real-world applications the running time of tasks varies due to different events that trigger the system, imposing a different workload on the system resources. This paper presents an analytic method to characterize tasks with variable execution requirements by finding the operation modes of every task code and estimating the WCET and BCET of each given mode. Then, we build a workload curve for every task for analyzing its performance by means of Real Calculus. A multimedia decoder was used to test this method. © 2012 IEEE.