As high-performance computing hardware incorporates increasing levels of heterogeneity, hierarchical organization, and complexity, parallel programming techniques necessarily grow in complexity or in their ability to abstract away complexity. The concurrent development of multi- and many-core processors, deep memory hierarchies, and accelerators and the variety of ways to combine these makes the low-level language route unmanageable for domain experts tasked with developing applications. The technologies that a competent developer might be expected to master and combine include MPI plus CUDA, OpenMP, and OpenACC, most commonly denoted MPI + X. This approach inherently saddles the developer with low-level details that might better be handled by high-level abstractions.

Higher-level parallel programming models offer rich sets of abstractions that feel natural in the intended applications. Such languages and tools include (Fortran, UPC, Julia), systems for large-scale data processing and analytics (Spark, Tensorflow, Dask), and frameworks and libraries that extend existing languages (Charm++, UPC++, Coarray C++, HPX, Legion, Global Arrays). While there are tremendous differences between these approaches, all strive to support better programmer abstractions for concerns such as data parallelism, task parallelism, dynamic load balancing, and data placement across the memory hierarchy.

This workshop will bring together applications experts who will present concrete practical examples of using such alternatives to MPI in order to illustrate the benefits of high-level approaches to scalable programming. The workshop expands upon the two similar workshops, PAW16 and PAW17, by broadening the theme beyond partitioned global address space languages. We invite you to take part in the Parallel Applications Workshop, Alternatives To MPI, and to join this vibrant and diverse community of researchers and developers.