
CppCon
3 HN recommenders · recommended in 3 months, 2020–2022
1:00:17 · 41K views · Oct 11, 2019
Stephan T. Lavavej's CppCon 2019 talk on implementing floating-point to_chars and from_chars in Microsoft's C++ STL. It covers the history of float-string conversion and the algorithms that make it much faster.
HN calls it a surprisingly fun, approachable talk on a hard problem and cites it in floating-point threads.
5 comments from Hacker News
I really love this talk in particular: this video (On charconv) But any of the ones I've seen from Stephan have been fantastic, I think, although I'm not actually sure if I've seen more. He seems to have a lot of talks on very specific subjects, which can be really fun. > Hyrum Wright is in the audience yelling interjections as a result of Hyrum's law This sounds absolutely hilarious, I'll have to take a look.
Stephan T Lavavej also had a surprisingly fun and approachable talk on that problem, talking about how charconv got implemented and improved in Microsoft's C++ standard library: this video
For those interested in more information about performance was achieved, Stephan Lavavej gave a talk about this very subject during CppCon 2019 [1]. [1] this video
I'm glad you liked my charconv talk! Here's a complete list of my recorded conference talks: BoostCon/C++Now 2012: Regex In C++11 And Boost: [LINK] GoingNative 2012: STL11: Magic && Secrets: [LINK] GoingNative 2013: Don't Help The Compiler: [LINK] GoingNative 2013: rand() Considered Harmful: [LINK] CppCon 2014: STL Features And Implementation Techniques: [LINK] CppCon 2015: functional: What's New, And Proper Usage: [LINK] CppCon 2016: tuple: What's New, And How It Works: [LINK] CppCon 2018: Class Template Argument Deduction for Everyone: [LINK] CppCon 2019: Floating-Point charconv: Making Your Code 10x Faster With C++17's Final Boss: this video CppCon 2020: C++20 STL Features: 1 Year of Development on GitHub: [LINK] Pure Virtual C++ 2022: MSVC C++20/23 Update: [LINK]
Actually, there's even more recent research into doing this really efficiently, using an algorithm called Ryū. [e] [LINK] [-4] [LINK] [1.73] this video