floating-point arithmetic (Annex F)). An exception is Microsoft Visual
C++ for x86, which makes long double a synonym for double."
I worked around this MSVC "feature" recently in WinUAE by using
(optionally) tiny x86/x64 helper functions that directly execute native
FPU instructins.
Also, AFAIK, UAE FPU emulation (before softfloat was introduced) was
never designed to be very accurate, only fast, for example only basic
flags that can be used accurately with native FPU (without extra
complexity) are emulated. Everything else is simply ignored because 99%+
of programs don't care (at least when talking about real world Amiga
software).