The workshop will encompass large-scale scientific applications targeting HPC architectures, present and future. While the organizers represent three DOE supercomputer centers, we invite participation from computational and computer scientists worldwide. Through presentations and a panel discussion, we hope to survey the state of portability among different HPC architecturesa€”past and present, by sharing lessons learned and best practices making applications run on distinct existing architectures; and future by discussing long-term plans to develop and re-develop scientific applications to prepare for the next couple of generations of supercomputers. One anchor point for discussion is the distinct supercomputer architectures represented by coming next-generation machines at the Department of Energya€™s (DOEa€™s) supercomputer center sites: the GPU-CPU system Summit at Oak Ridge Leadership Computing Facility, the Aurora and Theta systems coming to Argonne Leadership Computing Facility, and the Cori system coming to the National Energy Research Scientific Computing Center (NERSC). Long-term portability of applications among these architectures, and anticipated exascale architectures, is an area of strategic importance to DOE. We have ongoing workshops and training events under that umbrella, but in this workshop we hope to share ideas and learn from the wider scientific HPC community nationwide and worldwide. The portability issue is a tough one. Rewriting or refactoring large scientific applications for each new supercomputer generation is untenable. We may today be at a crossroads, where by investing in portable development approaches now we can a€œfuture-proofa€ key applications through the exascale era. Portability here means some optimal combination of source-code portability and performance portabilitya€”performance that is a€œgood enougha€ across the full spectrum of distinct HPC architectures.