feclearexcept
来自 cppreference.com
| 在头文件 <fenv.h> 中定义 |
||
| int feclearexcept( int excepts ); |
(自 C99) | |
尝试清除位掩码参数 excepts 中列出的浮点异常，该参数是 浮点异常宏 的按位或运算。
内容 |
[编辑] 参数
| excepts | - | 列出要清除的异常标志的位掩码 |
[编辑] 返回值
0 如果所有指示的异常都已成功清除，或者如果 excepts 为零。如果出错，则返回非零值。
[编辑] 示例
运行此代码
#include <fenv.h> #include <stdio.h> #include <math.h> #include <float.h> /* * A possible implementation of hypot which makes use of many advanced * floating-point features. */ double hypot_demo(double a, double b) { const int range_problem = FE_OVERFLOW | FE_UNDERFLOW; feclearexcept(range_problem); // try a fast algorithm double result = sqrt(a * a + b * b); if (!fetestexcept(range_problem)) // no overflow or underflow return result; // return the fast result // do a more complicated calculation to avoid overflow or underflow int a_exponent,b_exponent; frexp(a, &a_exponent); frexp(b, &b_exponent); if (a_exponent - b_exponent > DBL_MAX_EXP) return fabs(a) + fabs(b); // we can ignore the smaller value // scale so that fabs(a) is near 1 double a_scaled = scalbn(a, -a_exponent); double b_scaled = scalbn(b, -a_exponent); // overflow and underflow is now impossible result = sqrt(a_scaled * a_scaled + b_scaled * b_scaled); // undo scaling return scalbn(result, a_exponent); } int main(void) { // Normal case takes the fast route printf("hypot(%f, %f) = %f\n", 3.0, 4.0, hypot_demo(3.0, 4.0)); // Extreme case takes the slow but more accurate route printf("hypot(%e, %e) = %e\n", DBL_MAX / 2.0, DBL_MAX / 2.0, hypot_demo(DBL_MAX / 2.0, DBL_MAX / 2.0)); return 0; }
输出
hypot(3.000000, 4.000000) = 5.000000 hypot(8.988466e+307, 8.988466e+307) = 1.271161e+308
