Typed linear algebra composes linear algebra libraries with strongly typed quantity, unit libraries to add compile-time guarantees and developer ergonomics for safer physical linear algebra systems. We present a first, free, and open-source implementation that integrates dimensional analysis in linear algebra computations through the type system while preserving the performance of established numerical backends.
The talk motivates the safety proposition. We will work through the typed matrix definition and examples. We will evaluate ergonomics and compatibility with std::linalg, std::mdspan, Eigen, mp-units. We will note lessons learned, tradeoffs, frictions. We will look to other additional safeties, open problems, and opportunities for better linear algebra in C++ applications.
View Slides
François Carouge is a staff software engineer at Johnson & Johnson. François focuses on software infrastructure, quality, and cybersecurity. François spent the last 15 years defining and supporting large scale distributed control and robotics software systems in regulated industries.
François holds masters of science in computer engineering, embedded systems from ESIGELEC and Lehigh University. François contributes to the free and open source C++ software, robotics, and control ecosystems through his local meetups and public repositories.