/*! \fixme Isn't there some x86 FPU instruction for quickly
** converting a float to a rounded integer? It's worth
** looking into at some point... */
- return static_cast<int>(x+0.5f);
+ // return static_cast<int>(x+0.5f); // <-- (a) fast, but rounds -1.333 to 0!
+ // return static_cast<int>(rintf(x)); // <-- (b) slow, but correct
+ if (x>=0) return (int) (x + 0.5); // <-- slower than (a), but correct, and faster than (b)
+ else return (int) (x - 0.5);
+}
+inline int round_to_int(const double x) {
+ // return static_cast<int>(x+0.5);
+ // return static_cast<int>(rint(x));
+ if (x>=0) return (int) (x + 0.5);
+ else return (int) (x - 0.5);
}
-inline int round_to_int(const double x) { return static_cast<int>(x+0.5); }
inline int ceil_to_int(const float x) { return static_cast<int>(ceil(x)); }
inline int ceil_to_int(const double x) { return static_cast<int>(ceil(x)); }