The post explores the phenomenon of an additional build step making a Zig app run faster. It discusses how a bash pipeline works and why the extra step improves performance. The author seeks help from the Zig community and makes changes to the benchmark to fix the issue. They also discover another performance fix related to syscalls.
Table of contents
Creating a minimal reproduction of the phenomenon 🔗︎Asking the Zig community for help 🔗︎My mental model of bash pipelines is wrong 🔗︎Revisiting my byte counter 🔗︎Fixing my benchmark 🔗︎Applying Andrew Kelly’s performance fix 🔗︎Benchmarking a larger input 🔗︎Cheating my way to maximum performance 🔗︎Conclusion 🔗︎Sort: