Runtime optimization techniques have been developed for efficient data movement in embedded multimedia systems. As technology trends yield shorter cycle times and larger, wider datapaths in architectures for multimedia systems, the use of global broadcast networks to communicate operands is becoming a major bottleneck, limiting processor performance. This research investigates dynamic optimization mechanisms that exploit the highly regular operand distribution patterns that frequently occur in multimedia applications. Using dynamic microarchitectural mechanisms, global operand communication is converted to local transport using a lower-cost bypass network that supports instruction clustering. This technique maintains binary compatibility by applying run-time optimization rather than requiring changes to the ISA, recompilation with a vectorizing compiler, or manual retargeting and optimization. Analysis of the data access and movement patterns in video processing applications enables us to maximally exploit their inherent parallelism on current and emerging parallel execution platforms.

Publications:

• H. Kim, S. Wills and L. M. Wills, Optimizing Operand Transport using Dynamic SIMDization in Multimedia Systems, Proceedings of the IEEE International Workshop on Multimedia Signal Processing, pp. 372-377, Victoria, British Columbia, Canada, October 2006.
• H. Kim, S. Wills and L. M. Wills, Reducing Operand Communication Overhead using Instruction Clustering for Multimedia Applications, Proceedings of the IEEE International Symposium on Multimedia (ISM 2005), pp. 345-352, Irvine, California, December 2005.
• P. Sassone, S. Wills and G. Loh, Static Strands: Safely Collapsing Dependence Chains for Increasing Embedded Power Efficiency, Proceedings of the Conference on Languages, Compilers, Tools for Embedded Systems (LCTES), pp. 127-136, Chicago, Illinois, selected as one of five best papers, June 2005.
• H. Kim, D. S. Wills and L. M. Wills, Technology-based Architectural Analysis of Operand Bypass Networks for Efficient Operand Transport, Proceedings of the 4^th International Workshop on Performance Modeling, Evaluation, Optimization of Parallel and Distributed Systems (PMEO-PDS'05) , pp. 273b:1-8, Denver, Colorado, held in conjunction with the 19^th IEEE International Parallel and Distributed Processing Symposium (IPDPS'05), April 2005.
• P. G. Sassone and D. S. Wills, Dynamic Strands: Collapsing Speculative Dependence Chains for Reducing Pipeline Communication, Proceedings of the 37th Annual IEEE/ACM International Symposium on Microarchitecture, pp. 7-17, Portland, Oregon, December 2004.
• P. G. Sassone and D. S. Wills, On the Extraction and Analysis of Prevalent Dataflow Patterns, The IEEE 7th Annual Workshop on Workload Characterization (WWC-7), 8 pages, Austin, Texas, October 2004.
• H. Kim, D. S. Wills and L. M. Wills, Empirical Analysis of Operand Usage and Transport in Multimedia Applications, Proceedings of the 4^th IEEE International Workshop on System-on-Chip for Real-Time Applications(IWSOC’04), pp. 168-171, Banff, Alberta, Canada, July 2004.
• P. G. Sassone and D. S. Wills, Multicycle Broadcast Bypass: Too Readily Overlooked, Proceedings of the Workshop on Complexity Effective Design (WCED), 5 pages, Munich, Germany, June 2004.
• S. Bunchua, D. S. Wills and L. M. Wills, Reducing Operand Transport Complexity of Superscalar Processors using Distributed Register Files, Proceedings of the International Conference on Computer Design (ICCD), pp. 532-535, San Jose, California, October 2003.