Active Page Architectures for Media Processing

In this paper, we compare the performance of various media applications on vector and VLIW Active Page OCS98] memory architectures. Active Pages is an intelligent memory system that associates simple functions with each physical page of data. Previous investigations have demonstrated substantial speedups compared to a conventional memory system for media and data-intensive applications. This study explores two processing elements for use in Active Pages: a very long instruction word processor and a vector processor. We nd that simple streaming media applications execute equally well on both of these processor types. However, irregular applications execute better on VLIW processors, whereas the compact instruction format representation of a vector processor can more ee-ciently utilize the available processor resources for certain types of applications.

[1]  R.K. Gupta,et al.  MORPH: a system architecture for robust high performance using customization (an NSF 100 TeraOps point design study) , 1996, Proceedings of 6th Symposium on the Frontiers of Massively Parallel Computation (Frontiers '96).

[2]  Vivek Sarkar,et al.  Space-time scheduling of instruction-level parallelism on a raw machine , 1998, ASPLOS VIII.

[3]  Mikko H. Lipasti,et al.  An architectural evaluation of Java TPC-W , 2001, Proceedings HPCA Seventh International Symposium on High-Performance Computer Architecture.

[4]  Abhaya Asthana,et al.  Design of an active memory system for network applications , 1994, Proceedings of IEEE International Workshop on Memory Technology, Design, and Test.

[5]  M. Oskin,et al.  Active Pages: a computation model for intelligent memory , 1998, Proceedings. 25th Annual International Symposium on Computer Architecture (Cat. No.98CB36235).