September 28th, 2026 — I travelled to Singapore together with Evelyne Ringoot to present our joint work on KernelTuner.jl at ICPP’26 in Sustain-HPC. The paper is a result from a collaboration with the group of Prof. Alan Edelman (MIT) that started a year before when I met Evelyne at ICPP’25. In this work, We address this challenge by integrating auto-tuning directly into hardware-agnostic GPU kernels written in Julia, using our new framework, KernelTuner.jl.
Abstract
Traditionally, GPU kernels have been developed and optimized within vendor-specific programming models to achieve high performance, resulting in software that is difficult to optimize and adapt across increasingly heterogeneous computing systems. Hardware-agnostic programming models offer a more sustainable approach to GPU software development by improving portability and maintainability, but achieving efficient execution across diverse architectures remains challenging. We address this challenge by integrating auto-tuning into hardware-agnostic GPU kernels written in Julia.
We rebuild the established Kernel Tuner auto-tuning framework with Julia support, enabling systematic exploration of kernel configurations for hardware-agnostic GPU kernels targeting NVIDIA, AMD, Intel, and Apple GPUs. We demonstrate this approach on hardware-agnostic singular value decomposition (SVD) as implemented in the NextLA.jl linear algebra library.
The results show that auto-tuning is essential for creating resource-efficient hardware-agnostic GPU kernels across a variety of hardware. Optimal configurations improve kernel performance by a factor of 3 × to 7 × compared to median parameter configurations, demonstrating the substantial impact of tuning on efficient hardware utilization. Moreover, auto-tuning enables hardware-agnostic kernels to surpass manually optimized implementations in efficiency while reducing developer effort. This is made widely accessible through the first auto-tuning framework for Julia, KernelTuner.jl, bringing portable yet resource-efficient GPU kernels within reach for a wide range of applications and devices.
Citation
F.J. Willemsen, E. Ringoot, Alen Edelman “Portable to Efficient: Auto-Tuning Hardware-Agnostic GPU Kernels in Julia” Workshop Proceedings of the 55th International Conference on Parallel Processing (ICPP 2026) https://doi.org/10.1145/3816891.3834891
