性能 - 字符串数值类型转换

最新推荐文章于 2025-09-26 01:40:49 发布

原创最新推荐文章于 2025-09-26 01:40:49 发布 · 705 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#c++ #搜索引擎

搜索引擎实战专栏收录该内容

6 篇文章

订阅专栏

这篇博客探讨了不同数值转换方法的性能，包括sscanf、std::ostringstream和一些优化实现如Grisu2。通过在Core i7 920 @ 2.67Ghz的PC上进行基准测试，结果显示某些优化后的函数如branchlut和sse2能提供显著的性能提升，速度比sprintf快20倍以上。同时，文章还对比了C++标准库中的转换方法与特定实现，如doubleconv和fpconv，显示了特定优化技术对于提高转换效率的重要性。

字符串数值转换，性能对比，批量转换可以考虑使用sscanf

sscanf

sscanf使用demo:

split+atof 和sscanf性能对比

其他函数性能对比

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().

Function	Time (ns)	Speedup
sprintf	194.225	1.00x
vc	61.522	3.16x
naive	26.743	7.26x
count	20.552	9.45x
lut	17.810	10.91x
countlut	9.926	19.57x
branchlut	8.430	23.04x
sse2	7.614	25.51x
null	2.230	87.09x

Implementations

Function	Description
ostringstream	`std::ostringstream` in C++ standard library.
ostrstream	`std::ostrstream` in C++ standard library.
to_string	`std::to_string()` in C++11 standard library.
sprintf	`sprintf()` in C standard library
vc	Visual C++'s `_itoa()`, `_i64toa()`, `_ui64toa()`
naive	Compute division/modulo of 10 for each digit, store digits in temp array and copy to buffer in reverse order.
unnamed	Compute division/modulo of 10 for each digit, store directly in buffer
count	Count number of decimal digits first, using technique from [1].
lut	Uses lookup table (LUT) of digit pairs for division/modulo of 100. Mentioned in [2]
countlut	Combines count and lut.
branchlut	Use branching to divide-and-conquer the range of value, make computation more parallel.
sse2	Based on branchlut scheme, use SSE2 SIMD instructions to convert 8 digits in parallel. The algorithm is designed by Wojciech Muła [3]. (Experiment shows it is useful for values equal to or more than 9 digits)
null	Do nothing.

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().

Function	Time (ns)	Speedup
ostringstream	2,778.748	0.45x
ostrstream	2,628.365	0.48x
gay	1,646.310	0.76x
sprintf	1,256.376	1.00x
fpconv	273.822	4.59x
grisu2	220.251	5.70x
doubleconv	201.645	6.23x
milo	138.021	9.10x
null	2.146	585.58x

Implementations

Function	Description
ostringstream	`std::ostringstream` in C++ standard library with `setprecision(17)`.
ostrstream	`std::ostrstream` in C++ standard library with `setprecision(17)`.
sprintf	`sprintf()` in C standard library with `"%.17g"` format.
stb_sprintf	fast sprintf replacement with `"%.17g"` format.
gay	David M. Gay's `dtoa()` C implementation.
grisu2	Florian Loitsch's Grisu2 C implementation [1].
doubleconv	C++ implementation extracted from Google's V8 JavaScript Engine with `EcmaScriptConverter().ToShortest()` (based on Grisu3, fall back to slower bignum algorithm when Grisu3 failed to produce shortest implementation).
fpconv	night-shift's Grisu2 C implementation.
milo	miloyip's Grisu2 C++ header-only implementation.
null	Do nothing.

Notes:

tostring() is not tested as it does not fulfill the roundtrip requirement.
Grisu2 is chosen because it can generate better human-readable number and >99.9% of results are in shortest. Grisu3 needs another dtoa() implementation for not meeting the shortest requirement.

参考资料:

https://wiki.so.qihoo.net/pages/viewpage.action?pageId=16145758
https://github.com/compmeist/fast-atof
https://github.com/yuanyuanxiang/_atof
https://github.com/j-jorge/atoi-benchmark