Re:Exercises in Emulation: Xbox 360’s FMA Instruction
by toast0 on Thursday March 21, 2019 @11:25AM
(#19453396)
There's not really a bug here. Some algorithms based on floating point numbers require floating point operations to work in certain ways. x86 floating point operates differently than most other floating point systems, in large part because of when it was developed, and that it can't change because it needs to remain compatible with existing code.
It would be a bug to release code that required x86 floating point behavior on a non-x86 system, but in this case, it was easily caught during early testing.
It would be a bug to release code that required x86 floating point behavior on a non-x86 system, but in this case, it was easily caught during early testing.