Abstract
Today, more and more complex tasks are emerging. To finish these tasks within a reasonable time, using the complex embedded system which has multiple processing units is necessary. Hardware/software partitioning is one of the key technologies in designing complex embedded systems, it is usually taken as an optimization problem and be solved with different optimization methods. Among the optimization methods, swarm intelligent (SI) algorithms are easily applied and have the advantages of strong robustness and excellent global search ability. Due to the high complexity of hardware/software partitioning problems, the SI algorithms are ideal methods to solve the problems. In this paper, a new SI algorithm, called brainstorm optimization (BSO), is applied to hardware/software partitioning. In order to improve the performance of the BSO, we analyzed its optimization process when solving the hardware/software partitioning problem and found the disadvantages in terms of the clustering method and the updating strategy. Then we proposed the improved brainstorm optimization (IBSO) which ameliorated the original clustering method by setting the cluster points and improved the updating strategy by decreasing the number of updated individuals in each iteration. Based on the simulation methods which are usually used to evaluate the performance of the hardware/software partitioning algorithms, we generated eight benchmarks which represent tasks with different scales to test the performance of IBSO, BSO, four original heuristic algorithms and two improved BSO. Simulation results show that the IBSO algorithm can achieve the solutions with the highest quality within the shortest running time among these algorithms.
Highlights
With the development of semiconductor technology, the application of embedded system is becoming more widely used and it’s architecture is becoming increasingly complex
This clustering method may lead individuals to gather in the same area and cause the algorithm to fall into a local optimum
In order to further demonstrate the performance of the improved brainstorm optimization (IBSO), we introduce the quality gain α and the running time gain β in this paper
Summary
With the development of semiconductor technology, the application of embedded system is becoming more widely used and it’s architecture is becoming increasingly complex. Because the software processing units (such as CPU, GPU) have the characteristics of short development period, flexibility and easy maintenance, while the hardware processing units (such as ASIC, FPGA) have the characteristics of high efficiency and low power consumption, the development of embedded system needs the hardware and software co-design to satisfy the demands of cost, power, performance and so on. The execution effect of a subtask assigned to the hardware processing unit is different from that assigned to the software processing unit. The goal of hardware/software partitioning is to get an optimal partitioning scheme to make the system achieve the best performance. Hardware/software partitioning is the key of the hardware and software co-design
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
