Title :
Language and visualization support for large-scale concurrency
Author :
Roman, Gruia-Catalin
Author_Institution :
Dept. of Comput. Sci., Washington Univ., St. Louis, MO, USA
Abstract :
SDL (shared dataspace language) is a language for writing and visualizing programs consisting of thousands of processes executing on a highly parallel multiprocessor. SDL is based on a model in which processes use powerful transactions to manipulate abstract views of a virtual content-addressable data structure called the dataspace. The process society is dynamic and supports varying degrees of process anonymity. The transactions are executed over abstract views of the dataspace. This facilitates elegant conceptualization of dataspace transformations and compact program representation. Processes and transactions enable SDL to combine elements of both large- and fine-grained concurrency. The view is a novel abstraction mechanism whose significance is derived from the fact that it allows processes to interrogate the dataspace at a level of abstraction convenient for the task they are pursuing
Keywords :
data structures; parallel programming; software engineering; specification languages; SDL; abstraction mechanism; compact program representation; fine-grained concurrency; large-scale concurrency; parallel multiprocessor; shared dataspace language; software engineering; transactions; virtual content-addressable data structure; visualization support; Computer languages; Computer science; Concurrent computing; Cultural differences; Hypercubes; Large-scale systems; Power system reliability; Programming profession; Visualization; Writing;
Conference_Titel :
Software Engineering, 1988., Proceedings of the 10th International Conference on
Print_ISBN :
0-89791-258-6
DOI :
10.1109/ICSE.1988.93710