ABSTRACT Background When sharing or logging numerical data, we must convert binary floating‐point numbers into their decimal string representations. For example, the number π might become 3.1415927. Engineers have perfected many algorithms for producing such accurate, short strings. Aims We present an empirical comparison across diverse hardware architectures and datasets. Methods We benchmarked several established and recent algorithms for converting binary floating‐point numbers (IEEE 754 double‐precision) to their decimal string representations. We executed the conversions across multiple CPU microarchitectures, including recent Intel (Alder Lake, Skylake), AMD (Zen 3, Zen 4), and ARM (Apple M1/M2, Neoverse) processors, using recent versions of GCC, Clang, and platform‐specific compilers and several datasets. Results and Conclusions Cutting‐edge techniques like Schubfach and Dragonbox achieve up to a tenfold speedup over Steele and White's Dragon4, executing as few as 210 instructions per conversion compared to Dragon4's 1500–5000 instructions. Often per their specification, none of the implementations we surveyed consistently produced the shortest possible strings—some generate outputs up to 30% longer than optimal. We find that standard library implementations in languages such as C++ and Swift execute significantly more instructions than the fastest methods, with performance gaps varying across CPU architectures and compilers. We suggest some optimization targets for future research.
Gareau et al. (Thu,) studied this question.