Multiprocessor Programming, 7,5 credits

Parallelism has been an approach to achieve high performance in computer systems for many years. Today, computer systems incorporate parallelism at several levels: regular arrays (matrices) of SIMD processor cores as in GPUs, a smaller number of general CPU cores as in multicore processors, or as connected processing nodes as in a distributed system. To utilize the performance potential of the hardware, we need to write efficient parallel programs. In other words, parallel computer systems and parallel programming are fundamental building blocks for contemporary as well as future computer systems. The aim of this course is to teach the students how to program computer systems consisting of many processors or processor cores to achieve high performance. We will look at programming of shared memory, both MIMD and SIMD, as well as programming of distributed systems. The focus in the course will be on programming of shared-memory multiprocessors.