Towards Porting Operating Systems with Program Synthesis

The end of Moore’s Law has ushered in a diversity hardware not seen decades. Operating system (OS) (and software) portability is accordingly becoming increasingly critical. Simultaneously, there been tremendous progress program synthesis. We set out to explore the feasibility using modern synthesis generate machine-dependent parts an operating system. Our ultimate goal new ports automatically from descriptions machines. One issues involved writing specifications, both for functionality and instruction architectures. designed two domain-specific languages: Alewife machine-independent specifications Cassiopea describing architecture semantics. Automated porting also requires implementation. developed toolchain that, given specification machine description, specializes target synthesizes implementation assembly language with customized symbolic execution engine. Using this approach, we demonstrate successful total 140 OS components pre-existing OSes four real platforms. several optimization methods OS-related improve scalability. effectiveness our languages ability synthesize code all evidence code. However, many research challenges remain; discuss benefits limitations synthesis-based approach automated porting.