Abstract

Nowadays, it is very important that integrating parallel processors on a chip offers high performance and low interactive response time on applications with fine-grained parallelism and high degree of data sharing. We propose a novel real-shared cache module with new multiport ring-bus architecture to overcome the bus bottleneck problem of the existing parallel processors chip on shared cache level. A testbench of solving a large scale of simultaneous linear equation is also designed to evaluate such architecture. The evaluation results show that it can offer immediate data sharing without conflicts or delay, and the performance of parallel processors chips with such novel real-shared cache module improves in proportion to the number of processor elements.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call