__builtin_sqrt与__builtin_sqrtf

C/C++使用__builtin_sqrt与__builtin_sqrtf。

这两个属于内置函数，使用硬件加速运算。

The __builtin_sqrt() and __builtin_sqrtf() functions are supported only when hardware floating point support is enabled. In addition, the __builtin_sqrt() function is not supported if --float_support is set to fpv4spd16.

意思是这两个函数需要硬件启用了浮点支持。

(来源：http://downloads.ti.com/docs/esd/SPNU151/built-in-functions-stdz0554909.html)

速度对比：

对1e8个数字进行根号运算。

 普通sqrt：

#include <bits/stdc++.h>
using namespace std;
int main()
{
	for (int i = 1; i <= 100000000; i++)
		double num = sqrt(1.0 * i);
	cout << clock() << "ms" << endl;
	return 0;
}

__builtin_sqrt：

#include <bits/stdc++.h>
using namespace std;
int main()
{
	for (int i = 1; i <= 100000000; i++)
		double num = __builtin_sqrt(1.0 * i);
	cout << clock() << "ms" << endl;
	return 0;
}

__builtin_sqrtf：

#include <bits/stdc++.h>
using namespace std;
int main()
{
	for (int i = 1; i <= 100000000; i++)
		double num = __builtin_sqrtf(1.0 * i);
	cout << clock() << "ms" << endl;
	return 0;
}

两者都较普通sqrt快了近十倍。

 

__builtin_sqrt与__builtin_sqrtf的区别：

#include <bits/stdc++.h>
using namespace std;
int main()
{
	cout << sizeof (__builtin_sqrt(9)) << endl;
	cout << sizeof (__builtin_sqrtf(9)) << endl;
	return 0;
}

一个返回8个字节的double型，另一个则返回4个字节的float型。