Dataflow actor network partitioning for multiple FPGAS
Dataflow actor network is used to display the relation between different actors in a directed graph. It is suitable for modelling signal and video processing in software applications. In this paper, the use of dataflow actor network is extended to the hardware implementation of streaming application...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2016
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/80933/1/ChinYongHuanMFKE2016.pdf http://eprints.utm.my/id/eprint/80933/ http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:121153 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Dataflow actor network is used to display the relation between different actors in a directed graph. It is suitable for modelling signal and video processing in software applications. In this paper, the use of dataflow actor network is extended to the hardware implementation of streaming applications via dataflow actor network partitioning for multiple FPGAs based on the number of cuts, connection workload, resource utilization ratio and latency. Multiple FPGAs partitioning is often required for implementing design with large logic count, for cost reduction, multi clock and multi power domains design implementation. The motivation of using the dataflow actor network is due to the nature of the network which closely resembles the structural view and the inter-connections of a design at the architecture level. This representation in the form of a dataflow actor network is suitable for implementing graph partitioning algorithms. The KL algorithm, GA, PSO, SA and WOA are used for single objective partitioning while the MOPSO, MOSA and MOWOA have been used for multi objective partitioning. The objective of this study is to develop partitioning algorithm suitable for use in dataflow actor network and to determine the appropriate partitioning criteria. Results showed that SA has better performance as compared to other partitioning algorithm for single objective partitioning. On the other hand, for multi objective partitioning the MOPSO has better performance for small design while MOSA has better performance for larger design. |
|---|