What are the optimizations facilitated only with -ffinite-math?

Question

All the information I can find in the documentation and on the Internet for -ffinite-math-only is "Allow optimizations for floating point arithmetic, which assume that the arguments and results are not NaN or + -Infs." I don’t think so. Does anyone know exactly what optimization is? Thanks

Answer 1

Under this assumption, many small things can be optimized, for example:

• x == x → 1
• x * 1 → x
• x >= y→ !(x < y)and the like.
• x/x→ 1if the compiler can prove it x != 0.
• it can allow the compiler to use the hardware max / min commands for type expressions x > y ? x : y.
• ...

, " ", , :

• x - x → 0
• 0 / x → 0
• x * 0 → 0