2007 | OriginalPaper | Buchkapitel
Architecture-Based Optimization for Mapping Scientific Applications to Imagine
verfasst von : Jing Du, Xuejun Yang, Guibin Wang, Tao Tang, Kun Zeng
Erschienen in: Parallel and Distributed Processing and Applications
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
It is a challenging issue whether scientific applications are suitable for Imagine architecture. To address this problem, this paper presents a novel architecture-based optimization for the key techniques of mapping scientific applications to Imagine. Our specific contributions include that we achieve fine kernel granularity and choose necessary arrays to organize appropriate streams. Specially, we develop a new stream program generation algorithm based on the architecture-based optimization. We implement our algorithm to some representative scientific applications on ISIM simulation of Imagine, compared the corresponding FORTRAN programs running on Itanium 2. The experimental results show that the optimizing stream programs can efficiently improve computational intensiveness, enhance locality of LRF and SRF, avoid index stream overhead and enable parallelism to utilize ALUs. It is certain that Imagine is efficient for many scientific applications.