high performance computing on graphics processing units: hgpu.org

The main issue discussed in this book is concerned with solving job scheduling problems in parallel calculating environments, such as multiprocessor computers, clusters or distributed calculation nodes in networks, by applying algorithms which use various parallelization technologies starting from multiple calculation threads (multithread technique) up to distributed calculation processes. Strongly sequential character of the scheduling algorithms is considered to be the main obstacle in designing sufficiently effective parallel algorithms. On the one hand, up till now sequential algorithms exhausted the possibilities of significant growth in the power of solution methods. On the other hand, parallel computations offer essential advantages of solving difficult problems of combinatorial optimization, pushing towards theory, methodology and engineering of solution algorithms.