Computer Engineering Seminar

Efficient and Predictable Execution of Object-Oriented Code on the SHAP Platform

Thomas PreusserResearcher and LecturerTechnische Universitt Dresden

This talk presents SHAP, a highly-customizable, Java-programmed embedded CLDC platform. SHAP aims at an efficient and predictable execution of compact bytecode and pioneers several new approaches for the implementation of high-level object-oriented functionality to reach this goal. It will be shown how the sequential bytecode execution is optimized on SHAP and how system services commonly expected in modern object-oriented environments are isolated from the execution of the application by hardware assistance.

The presentation will detail three main features of the implementation of the SHAP platform. Firstly, a bump-pointer code caching technique will be presented that realizes a fully predictable cache behavior. Secondly, the technique of reference coloring as pioneered by SHAP will be introduced, which enables a space-efficient and constant-time interface method dispatch. And thirdly, the object heap management with its deeply integrated garbage collector will be described. The support for soft references and reference queues of the latter extends the capabilities of SHAP even beyond those of a plain CLDC platform.

Dr. Preuer earned his Bachelor and his Diploma degrees in Computer Science after his studies at Technische Universitt Dresden and a year as an exchange student to Cockrell School of Engineering at the University of Texas at Austin. In 2011, he received his PhD working on Increasing the Performance and Predictability of Code Execution on an Embedded java Platform. His current research focuses on embedded systems and the efficient implementation of computations in hardware. Together with Dr. Martin Zabel, he is a lead designer of the embedded SHAP Java platform. He has made major contributions in the exploitation of FPGA carry chain structure and has been holding the world record in the enumeration of the Queens Puzzle since 2009, which was achieved by an FPGA computation. Currently, he is a researcher and lecturer at Technische Universitt Dresden where he has been giving classes in Computer Engineering for 7 years and is coordinating international academic cooperation projects funded by the European Union.

Sponsored by

Igor Markov