Publication Details
Architektury a programování paralelních systémů
Parallel architectures, symmetrical multiprocessors, clusters, distributed shared memory, OpenMP, MPI, HPF
Lecture notes cover parallel system architecture as well as programming. Firstly, parallelism is treated in general together with performance figures, overhead of parallel processing and laws restricting the speedup. Further on the most frequently used symmetrical multiprocessors (SMP) with a common bus are discussed, then cache coherence and related MSI and MESI protocols follow. The next topic is programming in OpenMP, synchronization constructs, shared memory communications and parallelization with examples how to use explained directives. The othe part devoted to interconnection networks, to models of communication, routing techniques and algorithms and is a base for understanding popular clusters and distributed memory architectures. Programming in standard user interface MPI is completed by examples of tzpical parallel applications. The closing part is oriented to data parallel systems and programming (HPF).