COCO: CONTINUOUS COMPILATION

Compilers, Programming Languages, and Architecture Group
Department of Computer Science, University of Pittsburgh

Profit and Resource Management for Embedded Systems

Continuous Compilation has a number of benefits and uses for embedded software. It is particularly well suited for enabling software malleability, meeting multiple design constraints that change at run-time, and managing system resources. In this sub-project, we are developing novel solutions to these problems and applying our work with Continuous Compilation to embedded software.

One selected result from this project is to develop novel analytic models of machine resources, code optimizations, and program code that can be used to better meet performance/cost goals for embedded systems. The models can be used to quickly explore different optimization decisions (e.g., optimization configuration, interactions, etc.) to select the optimizations that best meet an objective function. Unlike other work, such as iterative compilation and experimentally based searches, our approach does not require execution feedback to evaluate whether an optimization is benficial. Hence, the approach is extremely fast and can determine the benefit of an optimization much quicker than existing approaches.

Another result shows that dynamic code translation can be used to effectively manage embedded system resources. We have developed techniques for instruction code compression/decompression in a purely software based framework. Such an approach is more flexible than a hardware based solution because it allows the compressor and decompressor to be changed. Unlike other software approaches, our technique decompresses only instructions that are very likely to execute, which helps to reduce the run-time overhead of decompression. We have developed a technique, called partial image compression, that reduces code image size by up to 56%, with a run-time decompression overhead of only 0.1 to 5.2%.

Publications

All publications are available from the CoCo publication web site. Please select the tab above for the publications.

• Planning for Code Buffer Management in Distributed Virtual Execution Environments, Shukang Zhou, Bruce R. Childers, Mary Lou Soffa ACM/USENIX Conference on Virtual Execution Environments (VEE'05), Chicago, Illinois, June 2005

• A Model-Based Framework: An Approach for Profit-driven Optimization, Min Zhao, Bruce R. Childers, and Mary Lou Soffa, ACM SIGMICRO Int'l. Conference on Code Generation and Optimization (CGO'05), San Jose, California, March 2005

• Compact Binaries with Code Compression in a Software Dynamic Transaltor, Stacey Shogan and Bruce R. Childers, Conference on Design Automation and Test in Europe (DATE'04), Paris, France, February 2004.

• Profile Guided Management of Code Partitions for Embedded Systems, Shukang Zhou, Bruce R. Childers, and Naveen Kumar, Conference on Design Automation and Test in Europe (DATE'04), Paris, France, February 2004.

• Predicting the Impact of Optimizations for Embedded Systems, Min Zhao, Bruce R. Childers, and Mary Lou Soffa, ACM Conference on Languages, Compilers, and Tools for Embedded Systems, San Diego, California, June 2003.

• Width-Sensitive Scheduling for Resource Constrained VLIW Processors Tarun Nakra, Bruce R. Childers, Mary Lou Soffa ACM Workshop on Feedback-Directed and Dynammic Optimization, during the 33rd Int'l. Symp. on Microarchitecture (MICRO-33), Monterey, California, December 2000.

• Adapting Processor Supply Voltage to Instruction-Level Parallelism Bruce R. Childers, Tony Tang, and Rami Melhem Koolchips 2000, during the 33rd Int'l. Symp. on Microarchitecture (MICRO-33), Monterey, California, December 2000.

• Memory Bus Reordering for Reduced Energy Consumption Bruce R. Childers and Tarun Nakra ACM SIGPLAN Languages, Compilers and Tools for Embedded Systems, Vancouver, Canada, June 2000.