Parallelism in large software packages (why is it difficult to create efficient software)

In the article, types of parallelism used in architectures of modern computer systems are considered, and the ways of their manifestation in programs are described. Six paradigms of parallel programming are analyzed, and the relationship of paradigms to generations of high-performance computing systems is shown. Different methods of description and representation of parallelism based on various kinds of program models are considered. The reasons that determine challenges of developing efficient software packages for parallel computing systems are discussed. The connection between the material under discussion and the actively developed Internet encyclopedia of properties and features of AlgoWiki parallel algorithms is noted.