Yagiz Nizipli & Daniel Lemire | Ada: Parsing Millions of URLs per Second

With the end of Dennard scaling, the cost of computing is no longer falling at the hardware level: to improve efficiency, we need better software. Competing JavaScript runtimes are sometimes faster than Node.js: can we bridge the gap? We show that Node.js can not only match faster competitors but even surpass them given enough effort. URLs are the most fundamental element in web applications. Node.js 16 was significantly slower than competing engines (Bun and Deno) at URL parsing. By reducing the number of instructions and vectorizing sub-algorithms, we multiplied by three the speed of URL parsing in Node.js (as of Node.js 20). If you have upgraded Node.js, you have the JavaScript engine with the fastest URL parsing in the industry with uncompromising support for the latest WHATGL URL standard. We share our strategies for accelerating both C++ and JavaScript processing in practice.
About the speakers:
Yagiz Nizipli
Senior Software Engineer
I'm a Node.js core collaborator, OpenJS Foundation, and Node.js performance team member. I have a Master's in Computer Science from Fordham University, and I'm currently working as a Senior Software Engineer at Noonlight.
Daniel Lemire
Computer Science Professor, Open-Source Programmer
Université du Québec
Daniel Lemire is a computer science professor at the Data Science Laboratory of the University of Quebec (TELUQ). He is among the top 500 GitHub users worldwide in terms of follower count. He published over 80 peer-reviewed research papers, and he has been cited thousands of times. He is an editor at the journal Software: Practice and Experience (Wiley, established in 1971). In 2020, he received the University of Quebec’s 2020 Award of Excellence for Achievement in Research for his work on the acceleration of JSON parsing. His research interests include high-performance programming. He is @lemire on Twitter, and he blogs weekly at lemire.me/blog