Den här sidan är utskriven från Högskolan i Halmstads webbplats (www.hh.se). Texten uppdaterades senast den 2017-06-08. Besök webbplatsen om du vill vara säker på att läsa den senaste versionen.
Big data is just what it sounds like – huge volumes of data, collected by sensors and computers all around the world. When a large amount of information is analysed, conclusions can be used to for example decrease traffic accidents. The challenge is to create computers and programmes with enough capacity to process and evaluate the data. A newly published thesis from Halmstad University suggests heterogeneous processors for higher computational power – because difference matters.
– We want to increase the performance of computers and decrease their energy consumption. This can be done through both software and hardware development – I have studied both in my research, says Süleyman Savas, PhD student in the Embedded Parallel Computing group at the School of Information Technology at Halmstad University.
Süleyman Savas and his fellow researchers aim to increase computer quality and open new paths for future applications. They also try to keep the energy consumption low for minimal environmental harm.
– There are many applications requiring massive amounts of computation power and lots of time to execute, for example self-driving and communicating vehicles. When combining large amounts of data with an increased computational power, it is possible to give the autonomous vehicles improved driving skills. This would decrease the risk of accidents, says Süleyman Savas and continues:
– Additionally, in our everyday life there is always a need for high processing power due to the huge amounts of data being gathered and communicated by sensors and Internet of Things devices all around the world. A better computer capacity can provide enhanced services for the users with more bandwidth and higher speed.
New computer processors have many cores, but they are mostly identical. Süleyman Savas and his research colleagues in Halmstad are trying to develop a better hardware architecture with increased performance by combining different types of processing cores instead. But this hardware development places greater demands on software developers:
– It is more challenging to develop applications which can make use of all these processing cores. Therefore, we are also creating software tools to help the software developers design their applications so that they can easily make the best use of these new processors. I have evaluated some of these tools, helped optimise them and created a design method for developing the new heterogeneous manycore processors.
Süleyman Savas will continue his research at Halmstad University:
– My plan is to improve the design method for developing new processors. Additionally, instead of doing all the design steps manually, I am planning to automate them so that it will be much faster to develop new types of processors which can execute different sets of applications very efficiently.
The licentiate seminar took place June 2 at Halmstad University.
Thesis: ”Utilizing Heterogeneity in Manycore Architectures for Streaming Applications”, Süleyman Savas, the School of Information Technology at Halmstad University.
Supervisors: Tomas Nordström and Zain Ul-Abdin, Halmstad University.
Examinator: Antanas Verikas, Halmstad University.
Opponent: Håkan Grahn, Blekinge Institute of Technology.
The Centre for Research on Embedded Systems, CERES, is a long-term research program on embedded systems established by Halmstad University, supported by funding from the Knowledge Foundation and by research cooperation with Swedish industry. A general principle in this research programme is that functionality and performance are achieved through the intelligent and massive co-operation between a large number of simple devices. ‘The power of cooperation’ is thus a summarising statement for the research. This is applicable on both the ‘micro’ and ‘macro’ levels.
Text : LOUISE WANDEL
Photo: ISTOCK (if nothing else is stated)
When combining large amounts of data with an increased computational power, it is possible to give the autonomous vehicles improved driving skills. This would decrease the risk of accidents.