This project brings together several important research areas, including parallel compilers, Single-Program Multiple-Data (SPMD) libraries, and object-oriented programming models. The project combines these ideas to achieve high performance on parallel machines in a robust fashion over a range of languages.
Building on earlier research in the Parallel Compiler Runtime Consortium (PCRC) project, this project will investigate a language model that combines data-parallel features from High Performance Fortran (HPF) with an explicitly parallel SPMD programming style. This "HPspmd" model allows direct calls to many established libraries for parallel programming with distributed data. By relaxing HPF's single-threaded semantics and having the programmer explicitly manage non-local references, the task of the compiler is hugely simplified. However, good library bindings (e.g. class libraries) supply the required expressiveness for programming. This project will develop and optimize a Java version of the language model (HPJava); design and implement bindings to parallel libraries such as CHAOS, DAGH, Global Arrays, MPI,and ScaLAPACK; and implement large applications within the resulting framework. The implementations are expected to drive further refinement of the language model, especially in support of irregular problems.
