Massiv parallele Programmierung mit dem Parallaxis-Modell

From MaRDI portal





The author presents first the basic idea of the Parallaxis-model: he introduces a new layer between the physical and logical system structure of a parallel architecture. This intermediate layer virtualizes the processing capacity of the system in an analogous manner as a virtual storage management module the memory capacity. The main components of the Parallaxis-model are: the interconnection topology, the parallel processing elements and a central control processing unit (host). Possible concrete hardware realizations of the model are placed in the proximity of SIMD-architectures. After the more introductory chapters 1 and 2, the author discusses classification schemes for parallel hardware (SIMD, MIMD,..) and parallel operations on different levels (e.g. vector operations), and language constructs. He introduces then the basic feature of Parallaxis which is the possibility of not only specifying parallel algorithms but also the interconnection structures (topologies) on which they are based. As target hardware, he has reconfigurable SIMD systems with a control unit in mind. As topologies he considers primarily regular topologies (grids, rings, cubes etc.). The chapters 6,7,9 describe in more detail the constructs of the Parallaxis language (statements, declarations and data types), primitives for E/A and the communication with the host. Chapter 8 presents a brief excursion about parallel semantics and gives a specification of denotational semantics for the parallel Parallaxis constructs. Chapter 10 gives a survey of the implementation of Parallaxis which consists of a compiler, simulator (of a target architecture) and debugging tools. The chapters 11 and 12 present some examples of applying the Parallaxis model. The author discusses not only systolic applications but also examples from the areas of computer graphics (e.g. ray tracing) and neuronal networks. Chapter 13 discusses again some hardware aspects; it gives in particular a list of existing systems which can be taken as target machines for Parallaxis. In the remaining chapters, the author refers to related work and mentions certain extensions of his concept. This book seems particularly interesting for readers wanting to deepen a first, superficial knowledge of parallel architectures. It leads the reader all the way from the basic idea through the stages of building an abstract model to the concrete implementation. A rich list of references supports this process.












This page was built for publication: Massiv parallele Programmierung mit dem Parallaxis-Modell

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3998631)