The Melbourne University heterogeneous multiprocessor
J. Trivellas, K.E. Forward
Abstract
J. Trivellas, K.E. Forward
Abstract
The Melbourne University heterogeneous multiprocessor is an experimental multiple-input multiple-data (MIMD) multiprocessor with different processing elements and is based on a shared memory architecture. The authors discuss the overall structure of the multiprocessor, and concentrate on the bus arbitration and the synchronization techniques used. Results based on a parallel implementation of the Sieve of Eratosthenes are presented. The results indicate that for a three-processor system, the memory contention is fairly small (even though the processors access the common memory repeatedly), and that the heterogeneous multiprocessor achieves a good speedup over the uniprocessor system (approximately 1.5*) for a small increase in cost.>
A significance statement is not available in the OpenAlex record.
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.
The Melbourne University heterogeneous multiprocessor is an experimental multiple-input multiple-data (MIMD) multiprocessor with different processing elements and is based on a shared memory architecture. The authors discuss the overall structure of the multiprocessor, and concentrate on the bus arbitration and the synchronization techniques used. Results based on a parallel implementation of the Sieve of Eratosthenes are presented. The results indicate that for a three-processor system, the memory contention is fairly small (even though the processors access the common memory repeatedly), and that the heterogeneous multiprocessor achieves a good speedup over the uniprocessor system (approximately 1.5*) for a small increase in cost.>
Key concepts: Uniprocessor system, Multiprocessing, Computer science, Symmetric multiprocessor system, Parallel computing, MIMD, Synchronization (alternating current), Speedup