itoa Benchmark

Copyright(c) 2014-2016 Milo Yip (miloyip@gmail.com)

Introduction

This benchmark evaluates the performance of conversion from 32-bit/64-bit integer to ASCII string in decimal. The function prototypes are:

void u32toa(uint32_t value, char* buffer);
void i32toa(int32_t value, char* buffer);
void u64toa(uint64_t value, char* buffer);
void i64toa(int64_t value, char* buffer);

Note that itoa() is not a standard function in C and C++, but provided by some compilers.

Procedure

Firstly the program verifies the correctness of implementations.

Then, two cases for benchmark are carried out:

1. Sequential: Converts consecutive values in same number of decimal digits.

   For u32toa(), the tested values are { 1, 2, ..., 9 }, {10, 11, ..., 99 }, ... { 4000000000, 4000000001, ..., 4294967296}, i.e., groups of 1 to 10 decimal digits.

   For signed versions, use alternate signs, e.g. { 1, -2, 3, -4, ... 9 }.

   For 64-bit integer, there are groups of 1 to 20 decimal digits.

2. Random: Converts the shuffled sequence of the first case.

Each digit group is run for 100000 times. The minimum time duration is measured for 10 trials.

Build and Run

1. Obtain premake5.
2. make

Results

The following are sequential results measured on a PC (Core i7 920 @2.67Ghz), where u32toa() is compiled by Visual C++ 2013 and run on Windows 64-bit. The speedup is based on sprintf().

Note that the null implementation does nothing. It measures the overheads of looping and function call.

Since the C++ standard library implementations (ostringstream, ostrstream, to_string) are slow, they are turned off by default. User can re-enable them by defining RUN_CPPITOA macro.

Some results of various configurations are located at itoa-benchmark/result. They can be accessed online, with interactivity provided by Google Charts:

• corei7920@2.67_win32_vc2013
• corei7920@2.67_win64_vc2013
• corei7920@2.67_cygwin32_gcc4.8
• corei7920@2.67_cygwin64_gcc4.8

Implementations

FAQ

1. How to add an implementation?

   You may clone an existing implementation file (e.g. naive.cpp). And then modify it. Re-run premake to add it to project or makefile. Note that it will automatically register to the benchmark by macro REGISTER_TEST(name).

   Making pull request of new implementations is welcome.

2. Why not converting integers to std::string?

   It may introduce heap allocation, which is a big overhead. User can easily wrap these low-level functions to return std::string, if needed.

3. Why fast itoa() functions is needed?

   They are a very common operations in writing data in text format. The standard way of sprintf(), std::stringstream, std::to_string(int) (C++11) often provides poor performance. The author of this benchmark would optimize the "naive" implementation in RapidJSON, thus he creates this project.

References

[1] Anderson, Bit Twiddling Hacks, 1997.

[2] Alexandrescu, Three Optimization Tips for C++, 2012.

[3] Muła, SSE: conversion integers to decimal representation, 2011.

Related Benchmarks and Discussions

• [The String Formatters of Manor Farm] (http://www.gotw.ca/publications/mill19.htm) by Herb Sutter, 2001.
• C++ itoa benchmark by localvoid
• Stackoverflow: C++ performance challenge: integer to std::string conversion