ΔΙΠΛΩΜΑΤΙΚΕΣ ΕΡΓΑΣΙΕΣ 2015-2016

Synchronization algorithms for concurrent data structures on manycore embedded systems Brief Discription: Modern multicore embedded systems integrate a large number of CPU cores, which require efficient synchronization methods for accessing shared data. Algorithms that have been proposed in the High Performance Computing domain cannot be applied directly to the embedded systems, due to the limitations they impose, such as the energy consumption and limited support for atomic operations. On the other hand, the traditional synchronization methods, like mutexes and spinlocks provide poor scalability and high energy consumption and therefore are inefficient for the modern low power embedded systems. The scope of this thesis is to develop new synchronization algorithms for modern multicore embedded systems that run applications based on shared data structures. The algorithms should focus on high performance along with low energy consumption. The embedded chips that will be used for the development of the algorithms are the Freescale imx6 4-core and the Myriad 8-core platform. Prerequisites: Good knowledge of C and data structures; concurrent development and embedded systems architecture principles. Keywords: Concurrent data structures, lock-based and lock-free programming, multicore embedded systems. Contact Information