▶ 数学函数
● 舍入函数，考虑被舍入参数有双精度浮点和单精度浮点，舍入方式有区别，舍入结果有整形、长整形和长长整形，所以共有以下舍入函数。
1 // math_functions.h 2 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl round(double x); 3 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl roundf(float x); 4 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long int __cdecl lround(double x); 5 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long int __cdecl lroundf(float x); 6 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long long int __cdecl llround(double x); 7 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long long int __cdecl llroundf(float x); 8 9 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl rint(double x); 10 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl rintf(float x); 11 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long int __cdecl lrint(double x); 12 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long int __cdecl lrintf(float x); 13 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long long int __cdecl llrint(double x); 14 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP long long int __cdecl llrintf(float x); 15 16 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl nearbyint(double x); 17 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl nearbyintf(float x); 18 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl ceil(double x) __THROW; 19 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP double __cdecl trunc(double x); 20 extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ __CUDA_MATH_CRTIMP float __cdecl truncf(float x);
● 推荐单精度浮点使用函数 rint() 系列（单指令计算），而不用函数 round() 系列（8 指令计算）。其他几个用于单精度浮点的舍入函数也是单指令计算完成的。
● 基本算术运算的舍入精度
● 指令级函数（__ 前缀数学函数），使用编译器选项 -use_fast_math 可以强制使用以下指令级函数。
● 有的指令级函数名有可选后缀，表明舍入方向
