Network on Chip simulations
Fernandez Alguacil, Carlos (2016)
2016
Master's Degree Programme in Information Technology
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2016-12-07
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201611284811
https://urn.fi/URN:NBN:fi:tty-201611284811
Tiivistelmä
From the beginning of time, people are concerned to compare different things. On this thesis, it will be analyzed the performance of a NoC architecture. To complete this purpose, it is used a program called Transaction Generator (TG). TG is a program developed in TUT, and it is used to benchmark various parameters of NoC architecture. Different traffic models will be executed using TG, to know how the different characteristics (number of tasks, existence of periodical events or number of PEs) of these models will influence on the performance of the NoC architecture.
To get a better knowledge of the behaviour of the NoC, traffic models are executed several times using TG while some of the attributes of the NoC are modified. With these changes (different traffic models and different attributes), the thesis presents a wider view of the NoC’s behaviour.
Results and conclusions are presented on the final chapter of the thesis. The difference on the performance of the NoC is more reduced than expected when different traffic models are executed. This is because the traffic models selected do not present a huge difference between them. Besides, the impact of the attributes modified on the NoC is easily appreciated, and basically do not depend on the traffic model selected. However, this not implies that all the attributes measured have the same impact. For example, the activation of the DMA presents a high difference on the performance, while different NoC frequencies do not have any impact.
TG allows the user to monitor the execution of the traffic model in real time, but that stays out of the scope of this thesis. This tool of TG is called Execution Monitor and it could be a good choice to use it in futures thesis. With Execution Monitor the user could do a deeper research on the performance of a NoC architecture.
To get a better knowledge of the behaviour of the NoC, traffic models are executed several times using TG while some of the attributes of the NoC are modified. With these changes (different traffic models and different attributes), the thesis presents a wider view of the NoC’s behaviour.
Results and conclusions are presented on the final chapter of the thesis. The difference on the performance of the NoC is more reduced than expected when different traffic models are executed. This is because the traffic models selected do not present a huge difference between them. Besides, the impact of the attributes modified on the NoC is easily appreciated, and basically do not depend on the traffic model selected. However, this not implies that all the attributes measured have the same impact. For example, the activation of the DMA presents a high difference on the performance, while different NoC frequencies do not have any impact.
TG allows the user to monitor the execution of the traffic model in real time, but that stays out of the scope of this thesis. This tool of TG is called Execution Monitor and it could be a good choice to use it in futures thesis. With Execution Monitor the user could do a deeper research on the performance of a NoC architecture.