Am Sonntag 27 Mai 2007 13:17 schrieb Francois Gouget:
Visual C++ 2005 does not like the way we compute the infinity and NaN values. I've tried replacing the current divisions by zero with arithmetic on FLT_MAX (at least for the infinity calculations), but it did not like that either.
Does anyone know how to make these calculations portable? How does one get at these values on Windows?
Does it like less obvious calculations, like float one = 1.0; float zero = 0.0; float infinity = one / zero; float nan = zero / zero;
I couldn't find any usable way of getting a NaN and INF, only the test for it via isNaNf. The same problem will occur in d3d8 and d3d9 I think.