Last active
January 8, 2023 04:06
-
-
Save i-saint/cbce1044c32d75a3f47828d5f39bd59d to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#include <cstdio> | |
// https://www.johndcook.com/blog/2019/11/12/rump-floating-point/ | |
template <typename T> | |
T f(T x, T y) { | |
T x2 = x*x; | |
T y2 = y*y; | |
T y4 = y2*y2; | |
T y6 = y2*y4; | |
T y8 = y2*y6; | |
return T(333.75)*y6 + x2*(T(11)*x2*y2 - y6 - T(121)*y4 - T(2)) + T(5.5)*y8 + x/(T(2)*y); | |
} | |
template <typename T> | |
__declspec(noinline) | |
T f2(T x, T y) { | |
return f(x, y); | |
} | |
int main(int argc, char **argv) | |
{ | |
constexpr double a = 77617.0; | |
constexpr double b = 33096.0; | |
printf("inline: %e\n", f(a, b)); | |
printf("no-inline: %e\n", f2(a, b)); | |
} | |
// compilers may or may not use FMA for compile-time folding. using FMA improves precision. so, inlining affects result. | |
// msvc (v19 x64) with /O2 /fp:fast /arch:AVX2 outputs: | |
// inline: -1.180592e+21 | |
// no-inline: -1.328166e+21 | |
// | |
// see also: | |
// https://godbolt.org/z/WYd6EW1G6 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment