fdivs
Integer division by a constant melts into shifts. Floats don't get off that easy. fdivs is a real divide in silicon, and that's what the compiler reaches for.
fdivs f1, f1, f2 # f1 = a / b
blr
The catch: a constant divisor almost never stays a divide. Write x / 2.0f and the compiler turns it into x * 0.5f, a multiply by the reciprocal that's much cheaper. The divide instruction only shows up when b is computed at runtime.
Your task
Write func_80090ab8 to match the fdivs above.