The 4th International Workshop on Performance Portable Programming models for Manycore or Accelerators (P^3MA) has issued their Call for Papers. Held in conjunction with ISC 2019, This workshop will provide a forum to bring together researchers and developers to discuss community’s proposals and solutions to performance portability.
Performance portable approaches and implementations are becoming increasingly crucial for the application developers to make the best use of modern HPC architectures. With diverging architectural designs, it can be a challenge to develop high-level abstractions that can expose all the rich features of the hardware to the programmer, especially with the introduction of heterogeneous architectures. Unfortunately if the software framework fails to expose these features to the programmer, the end solution can place a serious limitation to the maximum performance that can be achieved. The constant struggle to find the optimal abstraction to give adequate control to the programmer without overwhelming them with low level details is the key in how well an approach is adopted. There are a number of leading approaches designed to address these issues. These approaches include libraries, directive-based standardized programming models such as OpenMP and OpenACC, Kokkos, and DSLs. Although these programming approaches are evolving to address the complexities of evolving hardware, there are still several opportunities to improve their design and implementations that can lead to better adaptability and easy-of-use of these approaches by application developers. Hardware systems are taking diverse routes, some systems continue to demonstrate heterogeneity such as Summit and Sierra and some not so much such as Riken’s or UK’s Isambard that is based on Arm processors.
Topics of interest for workshop submissions include (but are not limited to):
- Experience porting applications using high-level models focused on performance portability and productivity
- Hybrid heterogeneous or many-core programming with models such as threading, message passing, and PGAS
- Continuation-style and asynchronous task-based programming
- Scientific libraries designed for performance portability on heterogeneous systems
- Interoperability between software frameworks
- Experiences in implementing compilers for performance portable programming on current and emerging architectures
- Low level communications APIs or runtimes that support accelerator architectures
- Extensions to programming models needed to support multiple memory hierarchies and accelerators
- Performance modeling and evaluation tools
- Power/energy studies
- Auto-tuning or optimization strategies
- Benchmarks and validation suites
Submissions are due April 3, 2019.
