static inline int32_t convert_coord(double x, double z)
{
- /* we clamp to 31 bit instead of 32 bit because we use
- a (x1-x2) shortcut when comparing coordinates
+ /* we clamp to 26 bit because:
+ a) we use a (x1-x2) shortcut when comparing coordinates
+ b) we need to be able to multiply two coordinates and store them in a double w/o loss of precision
*/
x *= z;
- if(x < -0x40000000) x = -0x40000000;
- if(x > 0x3fffffff) x = 0x3fffffff;
+ if(x < -0x2000000) x = -0x2000000;
+ if(x > 0x1ffffff) x = 0x1ffffff;
return ceil(x);
}