The COMPUTE spring meeting 2025 will take place on Monday the 28th of April 2025.  The theme is "Development of Scientific Software". The event is aimed at

• all who write code to do their research and want to become better at it
• all who use code to do their research and want to learn more about where this code comes from and how it might be improved.

Program

Workshop 1: Parallel Debugging and Profiling

Joachim Hein (Centre for Mathematical Sciences, LUNARC, NAISS)

Abstract: This workshop will demonstrate parallel debuggers and parallel profilers available on LUNARC and NAISS systems.

Workshop 2: Writing efficient Python Code: From Basics to Best Practices

Jonas Lindemann (LUNARC, NAISS)

Abstract: This workshop explores techniques for writing performant and resource-efficient Python code. We will discuss selecting optimal data structures, leveraging specialised libraries, and implementing optimisation strategies. Through hands-on examples, attendees will learn how to profile code, identify bottlenecks and develop the judgment to balance code readability with performance. 

Workshop 3: Threat assessment regarding scientific software

Philipp Birken (Centre for Mathematical Sciences)

Abstract: A pillar of scientific software world wide are the US national labs, that develop, maintain and distribute a large amount of widely used open source software. It now has be questioned if or to what extent they will fulfill that role in the future. We will discuss the threat this poses to doing our research here in Lund and possible strategies to mitigate it.

Lightning talks

PhD students and PostDocs will discuss software development they are doing.

Aliaksandr Dvornik (Medical Physics): Advanced PyMC-Based Application for Lost Radioactive Sources Localization

This talk covers the ongoing development of a scalable modeling and data processing pipeline in Python. We demonstrate how Python frameworks, particularly FastAPI and PyMC, can work together to produce generic data and perform simulation tasks. Our goal is to build an effective application that is parallelized using the multiprocessing library.

Mynta Norberg (Analysis and Synthesis): Lignonaut - A cheminformatic toolkit written in R for the virtual combinatorial synthesis and exploration of lignin oligomers

Lignin is the second most abundant natural polymer. Analytical chemists use advanced equipment to study the complex structure of lignins, but this produces data that is hard to interpret. Lignonaut is a cheminformatic tool that supports this endeavour, and it was written in R due to its popularity in omics and cheminformatics. Algorithms for combinatorial synthesis, duplicate identification, and dictionary-based translation will be discussed, alongside quirks of using R for scientific programming, and approaches for optimisation.

Philipp Stürmer (Mathematical Physics): A Gradient-Based Eigenvalue Solver in C for Linear Response Problems

Physicists, materials scientists, and quantum chemists alike need to compute system excitations—how a system responds to small perturbations. This problem often reduces to solving a linear response eigenvalue problem, requiring the computation of the lowest n eigenvalues. We present a gradient-based iterative solver implemented in C that works with dense, sparse, or matrix-free matrix-vector products on shared-memory systems. Benchmark tests show that this implementation can reduce computation time from five days to just 2.5 hours for relevant systems, outperforming ARPACK-based approaches.

Valentina Schüller (Numerical Analysis and Scientific Computing): Me, Myself, and I? On Coding in Public.

Abstract: I frequently write scientific code where I am the main developer and user. This can seem pointless or lonely, but I consider it to be a fun test bed for practicing collaborative software development (the collaborators: past me and future me). I want to show some examples of me talking to myself in public, and where I have seen that this approach pays off.

Niklas Kotarsky (Numerical Analysis and Scientific Computing): Tba

Registration

Please use the registration page to register