Quantitative research on variability in the execution-time of digital signal processing applications

Abstract

The task execution-time is unpredictability on the modern complex digital signal processing platform. This paper proposed the mathematical model and the quantitative research methods on the variability in the task execution-time. On the modern complex digital signal processing platform, the task execution-time probability density can be approximated as normal distribution. The mean-variance of execution-time, which can be an indicator to measure the degree of variability in the task execution-time, indicates the dispersion of all execution-time samples from the mean execution-time. This paper analyzed the variability in the task execution-time of the Godson 2E processor-based digital demodulation system. Experiments show that the interaction between different tasks will be reduced by task scheduling. Task scheduling is an effective method to significantly reduce the mean-variance of execution-time, and is an effective method to reduce the degree of variability in the task execution-time.