The term "deprecated" (for x87) was officially written in context of MSVC,
and has a longer history: MSVC was already without float80 (except in x87 asm) on win32.
For MSVC users, x87 float80 asm has been rare since win32, thus Microsoft can simply say:
The x87 support in MSVC, just needed for float80 in asm (not cross-platform),
is overly expensive, and unrequired in the longer term.
For all compilers, SSE (float32, float64) is preferred,
so float80 needs a separate implementation (x87 FPU or different).
The general plan is that compilers implement float80, float128 to be cross-platform.
With FPC, this currently means (untested by me): softfpu.pp for basic arithmetic
(overloaded operators in ufloatx80), and external library for other math functions.