The increasing adoption of heterogeneous computing architectures, particularly GPU-accelerated and domain-specific systems, has redefined the landscape of computing, from high-performance computing (HPC) to embedded systems at the edge. However, this architectural heterogeneity introduces significant challenges in achieving performance portability. Vendor-specific programming models (e.g., CUDA, HIP) often result in tightly coupled codebases, limiting cross-platform interoperability, impeding collaborative development, and undermining reproducibility.

As scientific applications scale toward exascale and beyond, the need for performance-portable solutions becomes critical. Achieving near-peak performance across diverse architectures without sacrificing maintainability or extensibility requires abstraction of layers, domain-specific languages, and compiler/runtime innovations. Simultaneously, the energy footprint of large-scale simulations is becoming a first-order concern, necessitating energy-aware optimizations and sustainable computing practices.

This convergence of performance, portability, and sustainability is not merely a technical challenge; it is a foundational requirement for the future of computational science and engineering.

Objectives

This workshop will focus on methodologies and frameworks that enable scalable and portable scientific computing across heterogeneous platforms. Topics will include:

  • Compiler and runtime strategies for performance portability.
  • Abstractions and Domain Specific Languages for architecture-agnostic optimization.
  • Case studies demonstrating cross-platform scalability and performance tuning.
  • Lessons learnt from code bases being ported to performance-portable programming models (e.g., Kokkos, SYCL, RAJA, OpenMP offloading).
  • Portability approaches for legacy applications.
  • Metrics and methodologies for evaluating portability in HPC.
  • Co-design approaches for balancing performance, portability, and power efficiency.
  • Portability of performance as well as other metrics, including (but not limited to) energy, power, carbon efficiency, and application accuracy.
  • Portable applications for future computing systems and architectures.

The session aims to foster discussion on the trade-offs and synergies between these dimensions, highlighting emerging research and practical solutions that address the evolving demands of extreme-scale scientific computing.