2011-01-26 Dataflow - Parallel Programming with Streams
When: 2011-01-26 09.00
Where: LTH, E-huset, E:2116
Speaker: Jörn W. Janneck, Department of Computer Science
After decades of unabated performance improvements, sequential computing machines seem to have reachd their terminal velocity. Instead of architecting ever more sophisticated sequential processors, therefore, future computers will become increasingly more parallel. Future applications, in order to be portable, will need to be written in a way that is amenable to efficient implementation on platforms with different architectures and different degrees of parallelism.
Dataflow is a parallel programming model based on the notion of streams and computational kernels processing them. In addition to its simple and intuitive model of concurrency, it is a good match to many of the application areas that show particular demand for processing speed, such as video coding, image and video processing and analysis, digital signal processing, and networking and packet processing.
In this talk I will present some of the features of dataflow as a parallel programming model, and discuss some of the research challenges on the way to making it a practical reality.