Performance estimation of embedded applications on microcontrollers

This paper extends our previous work on the source-code level energy consumption estimation with emphasis on more precise performance estimation. In this work we present results for three mainstream microcontrollers with different architectures. Our proposed methodology lies in measuring each C operation's execution time therefore creating a measurement accurate performance models for the microcontrollers. The higher precision for the estimates in this work is achieved by introducing a model for loops and hence improving the results both for nested loops and also for operations. As a result we show that the performance estimation method is applicable on most cases with estimation difference between -1% to 7.5%.