For some reason there’s an ongoing confusion about the following matrix :
|Native 32 bit floating point calculations||Native 64 bit floating point calculations|
|32 bit memory addressing||X||X|
|64 bit memory addressing||X||X|
So, all combinations work but people regularly (even developers) have the misconception that 64 bit floating point calculations somehow
must involve a 64 bit CPU architecture (and by extension, a 64 bit operating system).