What question did this study set out to answer?

This research aims to optimize spherical harmonic transforms for long-vector architectures and explore performance portability.

June 11, 2026Open Access

Study of Performance and Portability of a Scientific Code on Long Vector Architectures

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial evaluates performance portability in long-vector architectures, suggesting optimization strategies for HPC applications.

This research aims to optimize spherical harmonic transforms for long-vector architectures and explore performance portability.
Developed a reusable workflow based on software development vehicle (SDV) methodology.
Optimized spherical harmonic transforms in XSHELLS for the EPAC prototype.
Compared compiler auto-vectorization with architecture-specific vector intrinsics.
Code refactoring for auto-vectorization achieved speedups of up to 1.91×.
Explicit vectorization provided gains of up to 2.49×, overcoming compiler limitations.
Performance gains on NEC SX-Aurora reached 20.59×, while Intel and NVIDIA showed gains of only 0.63× and 0.87× respectively.

Paul Pölzl (2026) studied this question.

View Full Paper

Ask AI

Bookmark