Concurrency and Programming Languages

Current Projects

SCOOP: Compiler support for task-parallelism. We are developing SCOOP, a compiler for task-parallel programs that produces code for both kinds of cores in a Cell processor, and transparently inserts all needed calls for inter-core communication. The compiler targets a runtime system that automatically resolves all task-dependencies and guarantees deterministic execution of the program.
Static dependency analysis: We are developing a static analysis that infers task-dependencies statically, reducing the runtime cost of synchronization, without sacrificing the soundness and determinism of the program.
Runtime support for parallel programming: We are developing a language runtime that implements novel abstractions for parallel programming. A scheduler for task-parallel programs that provably guarantees a deterministic order of execution, and a memory-menagement library that can hide the NUMA or distributed memory hierarchy from the programmer.