Implementation of hardware software partitioning in embedded system
Loading...
Date
2018-06
Authors
Masyirah Mohd Nor
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Hardware/Software partitioning is a crucial problem in embedded system. It resides on
deciding which process of embedded system should be executed on specific hardware
and which one on software. To find an optimal partition is hard due to large number and
different characteristics of the system to be consider. In this paper, the heuristic method
is used to evaluate the performance of PSO and GA algorithm in term of to achieve a
best cost, total area and total execution time. PSO and GA algorithm are chosen in this
project to perform hardware software partitioning using Python 2.7.14. The constraints
of these algorithms applied to the program are targeted for total area (2500KB) and total
execution time (2500µs). The parameter setting is used for both algorithms is 15
number of task/node, 500 maximum number of iteration, 100 no of particles/population
size and the pre-defined hardware and software area and execution time for each task as
the input of the algorithms. Both of the algorithms are achieve the same best cost which
is 169.6022 and the optimum solution proposed by both algorithms are nine tasks
should be run in hardware and six tasks should be run in software. The result of total
area and total execution time for PSO is 2495KB and 2363µs. While for GA, the total
area is 1605KB and the total execution time is 2147µs. As a conclusion, the
hardware/software partitioning which is PSO and GA have been successfully develop in
this project. The result shows that, the GA algorithm has better total area and total
execution time but same cost compared to the PSO algorithm. However, in order to
improve the performance, the hybridization of these algorithms is suggested.