optimizing CPU/FPU instructions

is it possible to code, say, an ADD that's faster than the actual instruction for ADD, only using other CPU functions (no physical access to the chip)? or are all CPU instructions always perfectly optimized?

>>14
Not that much faster (in my experience, the low values for FPU instructions are very rare in practice, and only occur on "round" numbers and special cases like 0/1.)

Nehalem:
FSINCOS ~110
FDIV 7-27
FPTAN ~115

Sandy Bridge:
FSINCOS 43-123
FDIV 10-24
FPTAN ~130

Ivy Bridge:
FDIV 10-24
FSINCOS 50-123
FPTAN ~130

Not the same precision either due to rounding and such.