Integrating Application Specification and Performance Prediction for Strategic Design-Space Exploration

Abstract--Modeling environments and performance prediction boost application productivity, but often lack integration into an efficient and comprehensive approach to strategic design-space exploration for reconfigurable computing (RC) systems. This paper proposes a framework allowing simple yet extensible bridging of modeling environments for highlevel application specification with the authors' RC Amenability Test (RAT), an analytical model for performance prediction prior to expensive implementation. Two case studies, a modified Needleman-Wunsch application for bioinformatics and a task graph of mean-value analysis, demonstrate the efficiency of the proposed integrated framework for rapid and reasonably accurate analysis of application designs.

• Type: Journal
• Year: 2010
• Where: ERSA
• Authors: Brian Holland, Alan D. George, Herman Lam
• Keywords: ERSA 2010, Hardware, Modified Needleman-wunsch Application, Performance Prediction, RC Amenability Test