Automatic Parallelization for a Class of Regular Computations
This book emerged from the contributions to the development of a high level CAD-tool for parallel code generation. All in all, it has 250 pages on which the following is described in detail: Chapters 2 to 5 consider the general partitioning and mapping problem of code statements and subprocedures onto the so-called computation polyhedron. Chapter 2 gives a survey and analysis of the partitioning and mapping problem. Chapter 3 presents improvements for some existing partitioning techniques. Chapter 4 proposes the authors basic methodology of partitioning and mapping onto fixed sized (N-1)-D prozessor arrays. Chapter 5 deals with the lower-dimensional mapping problem. Chapter 6 considers the problems of determining the structure, the boundaries, and the data flow of parallel programs. In Chapter 7, automatic parallel code generation for given arrays is presented. Experimental results are discussed then in Chapter 8, a summary and a comparison of methods is given in the final Chapter 9. Algorithms and samples of automatically generated parallel codes are to be found in the appendices.
This page was built for publication: Automatic Parallelization for a Class of Regular Computations
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4379558)