SACAW 2026: 2nd Symposium on Accelerated Computing for Application Workflows

Bridging the Gap: The Three Core Communities

The core theme of this second edition centered on abstraction, intermediate representation, and advanced tooling for GPU porting. To address these complex technical challenges, the symposium united three vital, interconnected groups:

• Beginners & Researchers: Scientists looking for accessible, user-friendly tools to dramatically speed up their code without introducing unnecessary complexity into their day-to-day workflows.
• Domain Experts: Computational scientists running large-scale, real-world simulations who require raw performance and faster time-to-solution to drive their research forward.
• Tool Builders: Software developers creating the next generation of intuitive tools and acceleration libraries, looking for direct user feedback and real-world edge cases to test their systems.

Presentation Highlights

The event featured four deep-dive technical presentations exploring the frontiers of accelerated computing:

1. Data Movement, Data Layout, and Data Precision: The New Frontiers of DaCe
Alexandru Calotoiu (Senior Researcher, ETH Zürich)

Alexandru showcased DaCe (Data-Centric Parallel Programming), a powerful framework designed to optimize and accelerate Python-based applications across diverse hardware architectures. He also shared highly anticipated insights into how Large Language Models (LLMs) can be leveraged to streamline data-centric optimizations. Notably, the DaCe framework won the prestigious Gordon Bell Prize for Climate Modeling in 2025.

2. TAMALE: An Accelerated Tool for the Analysis of Large Molecular Dynamics Simulations
Ana Cunha (Oak Ridge National Laboratory, US)

Presenting online, Ana introduced TAMALE, a newly developed acceleration tool specifically engineered to tackle performance bottlenecks in the analysis of massive molecular dynamics simulations. Her presentation highlighted how high-performance analysis tools are becoming just as critical as the core simulations themselves.

3. Accelerated Chemistry with Non-Orthogonal CI
Remco Havenith (Rijksuniversiteit Groningen, NL)

Remco focused on the complexities of accelerating quantum chemistry applications, specifically detailing his work on a Non-Orthogonal Configuration Interaction (CI) program. He demonstrated how a traditional Fortran-based codebase can be successfully modernized, achieving massive scaling via MPI + OpenMP parallelization and direct GPU acceleration through OpenACC.

4. Porting CUBENS to GPUs: A High-Fidelity Solver for Non-Ideal Fluid Flows
Pietro Carlo Boldini (TU Delft, NL)

Pietro rounded out the session by detailing the architectural journey of porting CUBENS—a high-fidelity solver used to analyze non-ideal fluid dynamics—to GPUs. He provided a highly practical look at the exact methodologies, hurdles, and engineering triumphs encountered when moving complex solver pipelines onto modern accelerator hardware.

Interactive Brainstorming & Lightning Pitches

A standout feature of this year's symposium was its highly interactive format. Rather than relying solely on traditional presentations, dedicated collaborative discussion blocks were held at the conclusion of both the morning and afternoon sessions.

These structured brainstorming blocks provided an open floor for the audience and speakers to directly interact and unpack the day's core themes. Participants dove deep into technical bottlenecks, debated the merits of different abstraction layers, and shared firsthand experiences on the complexities of GPU porting, and the potential role of LLMs.

Complementing these discussions, a dynamic series of lightning pitches gave the audience a voice. Researchers took the stage to openly present their current code performance bottlenecks, sparking collaborative problem-solving sessions on what help they need to accelerate their research. Concurrently, software innovators pitched fresh frameworks and custom methodologies designed to simplify the steep learning curve of the GPU journey.

Ultimately, SACAW 2026 proved to be an invaluable knowledge exchange. Attendees walked away with actionable insights from real-world case studies, a deeper understanding of how their peers overcome computing constraints, and an expanded professional network equipped to help their research thrive.

Resources 

• Missed a session? You can find all the presentation slides, video recordings, and event photo galleries via the SACAW 2026 Event Repository: https://surfdrive.surf.nl/s/Q7J7stWmjPWTLAe

• Curious about our roots? Revisit last year's inaugural event by reading the SACAW 2025 Summary Blog to see how our community's journey first began: https://communities.surf.nl/en/future-computing-and-networking/article/…