Study of performance and portability of a scientific code on long vector architectures

Abstract

Long-vector architectures are currently experiencing a resurgence in high performance computing (HPC) as they promise massive parallelism and portable code when combined with auto-vectorization. The European processor accelerators (EPAC) prototype, developed by the European Processor Initiative(EPI), combines this principle with the open-source RISC-V "V" (RVV) extension. While the hardware is actively developed, most HPC software targets heterogeneous host-device platforms and is poorly prepared for long-vector hardware. Current literature specifically lacks full-scale optimization studies of applications dominated by spherical harmonic transforms or similar spectral methods. Therefore, the vectorization challenges and cross-platform performance portability of these workloads remain unclear. To address this gap, this thesis formalizes a reusable workflow based on the software development vehicle (SDV) methodology to systematically optimize the spherical harmonic transforms inside XSHELLSfor the EPAC prototype, explicitly comparing the trade-offs between compiler auto-vectorization andarchitecture-specific vector intrinsics. Performance evaluations reveal that refactoring code to assist auto-vectorization yields overall speedups of up to 1.91×. Explicit vectorization overcomes severe compiler limitations in nested loops and delivers overall gains of up to 2.49×. Although these code adaptations translate effectively to the NEC SX-Aurora (yielding gains up to 20.59× over the scalar baseline), they produce significant loop overhead, degrading performance to 0.63× and 0.87× of the auto-vectorized baseline on the Intel Sapphire Rapids and NVIDIA Grace CPUs, respectively. Ultimately, this research produces an SDV-based optimization blueprint for the EPAC platform and identifies three generalizable code patterns critical for vectorization efficiency. The results demonstrate that maximizing hardware utilization on current long-vector architectures requires manual intrinsic vectorization, as compiler support for nested multi-dimensional loops remains a critical bottleneck.