https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83352
--- Comment #2 from Alexander Zaitsev <zamazan4ik at tut dot by> --- What about longer sqrt functions call? gcc(trunk) -O3 -ffast-math: double test(double a) { return sqrt(sqrt(sqrt(sqrt(sqrt(a))))); } test(double): andpd xmm0, XMMWORD PTR .LC0[rip] sqrtsd xmm0, xmm0 sqrtsd xmm0, xmm0 sqrtsd xmm0, xmm0 sqrtsd xmm0, xmm0 sqrtsd xmm0, xmm0 ret Don't see any optimizations on godbolt