Effective use of Multi-Core Architecture through Multi-Threading towards Computation Intensive Signal Processing Applications

Abstract

With the advent of Multicore architecture availability, exploiting parallelism is posing certain trends and tides for application deployment. Earlier approaches to explore parallelism in applications were limited to either instruction level parallelism (ILP) or use of architectural redundant resources. In this paper, we attempted to use multicore processor to demonstrate the speedup in compute intensive tasks such as Convolution, primitive to most Digital Signal Processing algorithms. Further result of multithreaded application on Multicore processor compared with single core is demonstrated for lower and upper limit of granularity for application fragmentation. This work suggests the need for design of memory manager for Multithreading to exploit it more effectively. General Terms Multithreading multicore architectures