from what i read in the 6.1.2 sdk float message passing in 64-bit mode should now be able to preserve double precision:
"Additionally, the size of the t_atom is 8-bytes on x64, meaning that double-precision floating pointer numbers can be represented."
it seems this is not implemented correctly for all objects, (yet?)
if you run the example below in 64-bit mode, to my understanding only the calculation in the middle returns the correct result.
is this just in need of a fix, or am i missing something here?