Program granularity and performance in multiprocessor systems
Walter H. Burkhardt
Abstract
Walter H. Burkhardt
Abstract
Performance in multiprocessor systems depends on various factors. Granularity of the program exerts a surprising effect on the performance. This relationship is investigated theoretically and experimentally for a transputer system of up to 24 slave modules. The results show a strong dependency of performance on particular values of processor number, granularity, and number of tasks to be computed.>
OpenAlex reports 1 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Performance in multiprocessor systems depends on various factors. Granularity of the program exerts a surprising effect on the performance. This relationship is investigated theoretically and experimentally for a transputer system of up to 24 slave modules. The results show a strong dependency of performance on particular values of processor number, granularity, and number of tasks to be computed.>
Key concepts: Granularity, Multiprocessing, Computer science, Dependency (UML), Transputer, Parallel computing, Computer architecture, Programming language