From Segfault to Safety @Conciso

Wir Sprechen über das migrieren von Code auf neue Tech-Stacks. Erst erst eine Übersicht dann (1.Talk, Tim Janus)
30 Jahre altem C-Nasa Code zu Rust (2. Talk, Dr. Q)

Wrapping legacy SuperLU vs Rewriting in Rust.
Economics of a Rewrite with Agentic Coding.

Talk 2:
Two years ago, our team at Dr.Q wanted to use SuperLU in our Rust stack. SuperLU is a powerful and mature sparse linear solver developed at Lawrence Berkeley National Laboratory.
We followed the conventional advice for integrating complex legacy software: keep the battle-tested implementation, put a narrow Rust interface around it, and defend that boundary with thorough tests.
After roughly 18 developer-months of working on this approach, we changed course. Instead of continuing to wrap SuperLU, we rewrote it in Rust.
The rewrite took less than six developer-months.
More importantly, finishing the rewrite was not the end of the story. Once we owned the solver as a native Rust codebase, we were able to understand, modify, and improve it far more quickly than before.
In this talk, we’ll share how we approached the rewrite, where agentic coding tools changed its economics, and what we learned about an assumption that has guided software engineering for decades: “Wrapping complex, mature dependencies is safer and cheaper than rewriting them.”