Introduction - PAW 2017: The 2nd Annual PGAS Applications Workshop
Workshop: PAW 2017: The 2nd Annual PGAS Applications Workshop
Abstract: In order to manage the complexity of emerging architectures and technologies, parallel programmers require effective approaches for developing large-scale applications that fully exploit these transformative features. Ideally, such approaches should elevate the programmer’s expression of the computation from mechanism-specific abstractions (messages, kernels) to higher-level abstractions, similar to how Fortran and C permitted programmers to think in terms of variables rather than registers. The Partitioned Global Address Space (PGAS) family of programming models provides such an approach. It simplifies the complexity of parallel programming across distinct memories by making it more similar to shared-memory computation, while enhancing scalability by exposing data locality through its semantics. This workshop will bring together applications experts who will present concrete practical examples to illustrate the benefits of exploiting PGAS languages including Fortran, Unified Parallel C (UPC), X10, and Chapel as well as libraries such as Unified Parallel C++ (UPC++), Coarray C++, OpenSHMEM, MPI-3, and Global Arrays.