Here be Floating Dragons: A Tale of Making OTP Faster
summary
Replaces the old Dragon4-style float-to-string path in OTP’s I/O with the Ryu algorithm and an optional C BIF, changing io_lib float formatting; benchmarks show about a 4–5x speedup in typical Erlang/Elixir workloads and up to larger gains when switching the default, PRs are open (funded by the Erlang/OTP foundation), and further wins are possible with Dragonbox-style approaches but require replacing sprintf and C-level maintenance.
AI-written summary · about
chapters (6)
- 0:0001:36 - Once Upon a Time
- 1:3705:10 - Introduction to the First Dragons
- 5:1111:25 - Why We are Exploring Dragons?
- 11:2614:22 - Dragon Rules
- 14:2320:13 - Let's Go Faster!
- 20:1435:35 - QnA
related talks
56mHow to Quickly Convert Floating-Point Numbers in C++
44mWriting fast Ruby
01h 00mFloating-Point ＜charconv＞: Making Your Code 10x Faster With C++17's Final Boss
31mBenchmarking and Optimization of Rust Libraries
24mYou Did What? Understanding Node.js Optimizations
28mFastAPI: Hitting the Performance Jackpot
40mPerf Stories
33mI Made a Slow Thing Fast