How Fast Can You Parse 1 Billion Rows in Java? – Insane Speed Test • Roy van Rijn • GOTO 2025
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
A conference talk recounting the One Billion Row Challenge (1BRC), a contest to parse 1 billion rows of weather data as fast as possible using Java. Starting from a naive baseline of ~4 minutes 50 seconds, the speaker walks through a series of optimizations: parallelism, GraalVM native compilation, the epsilon (no-op) garbage collector, integer arithmetic instead of floating point, memory-mapped files, SWAR (SIMD Within A Register) techniques for delimiter finding, branchless programming for temperature parsing, a custom open-addressing hashmap with forward probing, cache-aware chunked file reading, and intra-core instruction-level parallelism using multiple scanners per thread. The final result reached ~1.5 seconds. A standout moment is the explanation of a Vietnamese developer's single-multiplication trick for parsing temperature values, which earned him a job at Oracle working on the JVM Vector API.