While parallel processing technology has rapidly advanced, with the transputer and other devices, the development of programming methods lags behind the hardware. Most of the programming languages that are currently used are basically sequential languages with the addition of a few parallel constructs. A language which provides intrinsic parallelism in order to free the programmer from the task of closely specifying the parallel parts of the program, combined with formal program verification, would allow the full use of the new multiprocessor's computational powers. CTDNet, which is described in this book, provides an executional mechanism that has been developed to contribute to the realization of real parallel programming languages. This book describes the implementation of a functional execution mechanism called CTDNet that provides the methodology and notation for true parallel processing. Application programs are described as a lambda graph, which contains the structure of the tasks to be computed. Duplication, distribution and reduction of these processes are locally controlled by an operating kernel, CTDOS, which is embedded in each processing element of the system.
Program writing is therefore liberated from consideration of network size as well as from the needs of specific parallel execution directives. An implementation structure for CTDOS on a transputer-based machine has been described. The book also presents results of a CTDOS kernel simulation that provides the tools to verify the CTDNet specifications and enables a lambda graph to be reduced in the context of simulating a 64 node CTDNet machine. Application examples are provided to evaluate the performance of the CTD operating system in relation to an equivalent mono-processor system."synopsis" may belong to another edition of this title.
(No Available Copies)
Search Books: Create a WantCan't find the book you're looking for? We'll keep searching for you. If one of our booksellers adds it to AbeBooks, we'll let you know!
Create a Want