Research Insights

Dynamic binary translation (DBT), used to translate executable code from one instruction set architecture to another, supports a range of essential applications, including architectural simulators, hardware emulators, runtime optimizers, program analysis tools, software testing platforms, and more. However, existing DBT systems have not kept up with the rapid evolution of computer software and hardware. For instance, existing DBT systems suffer from significant execution inefficiencies due to the poor quality of the translated binary code. As a consequence, it becomes increasingly challenging to adopt DBT for many important applications, especially those in emerging domains, e.g., machine learning and graph processing. The goal of this project is to modernize DBT systems through a series of novel innovations. A key insight is that the translation between different computer instruction sets –usually represented in assembly language– is similar to the translation between different natural languages. Inspired by this insight, this project will invent novel translation approaches to produce high-quality binary code. The innovations developed by this project will improve the performance efficiency of DBT systems when running binary code. Moreover, they will enhance the capability of DBT systems to support important applications. More efficient, more capable, and more powerful DBT systems, as modernized by this project, will not only benefit current DBT applications, but also enable DBT applications in new frontiers, producing long-lasting impacts. The innovative technologies developed in this project will provide insights for the research community to push forward the DBT field. Additionally, this project will assemble a set of well-organized teaching and mentoring activities to promote research experiences for undergraduates and create a more diverse and inclusive educational environment.

Funder: National Science Foundation

Amount: $599,977

PI: Wenwen Wang, School of Computing