It was a weird time. I remeber that in some version you could choice to compile using the standard IEEE 754 numbers or their propietary floating point version. More details (linked in the article) https://en.wikipedia.org/wiki/Microsoft_Binary_Format