Research On SoC Hardware/Software Partitioning Methodology Based On Cat Swarm Algorithm

With the rapid development of various embedded devices, FPGA technology matures, the on-chip system becomes increasingly large scale. In order for a limited time to design more powerful So C embedded devices, while achieving the desired performance, coordination of the various functional modules of system equipment and the concrete implementation of the different modules put forward higher requirements. Traditional design methods have been unable to meet the growing demand systems, hardware and software co-design has become a major design approach to solve this problem. Hardware and software co-design of hardware and software division is a crucial step. So C hardware and software modules for the partitioning problem with multiple performance indices, this paper presents a software partitioning algorithm based cat swarm optimization with respect to the traditional division algorithm, which can meet the system constraints, accelerate the convergence rate and improve convergence accuracy.In this paper, the basic principle of the hardware and software partitioning algorithm, the model system description, and system structure and division goals are analyzed in detail, the classical division algorithm genetic algorithms, simulated annealing algorithm and particle swarm algorithm were analyzed and compared. Because each of the three algorithms existing problems, the paper proposes the use of software and hardware division cat swarm algorithm software partitioning solutions, and implement the basic principles and methods cat swarm optimization depth study, the specific contents and main results are as follows:(1)In order to obtain better solutions, on the inside track cat swarm algorithm compares the position of the cat population and to analyze the disturbance range study to determine the appropriate range of parameters, and to achieve a continuous cat swarm algorithm binned.(2)To the cat population So C software partitioning algorithm is applied, the structure and division of model systems target selection. And system cost, execution time, power consumption, system performance and area constraints introduced into the objectivefunction. The actual division of the problem, create the target function penalized multiple constraints, the function as a result of the division of evaluation criteria. According cat swarm optimization processes, design and implementation of hardware and software partitioning algorithm based So C cat swarm optimization. First group of randomly generated initial cats and cat swarm algorithm using a unique tracking mode disturbance patterns and simultaneous optimization, two different modes can be simultaneously global and local search. When the cat group of individuals tend to some local optimum results when using the method of random perturbations search criteria and cats individual positions to avoid falling into local optimum and ultimately find the global optimal solution.(3)Simulation and verification of division algorithm. Use TGFF simulation tool to generate multiple sets of test data on the performance and optimization results of genetic algorithms, particle swarm optimization and validation of the algorithm is tested and compared. Simulation results show that compared with traditional genetic algorithm and particle swarm algorithm, which is easier and faster to find the global optimal solution of hardware and software division, and its optimization ability and execution time are better than two or more algorithms. To test the practicality of the algorithm, we use the telephone system as well as audio and video playback system with two instances of the system, and for more than two algorithms were compared to verify the algorithm, comparing the results also proved that the algorithm to optimize the efficiency and optimize the results were excellent traditional partitioning algorithm.This thesis is useful for the future study of hardware/software partitioning optimization of So C system.