IEEE 754 floating point numbers use a sign bit for signed-ness (rather than something like twos complement), so if you’re sure that your compiler/platform uses that representation (very common) then you can use -std::numeric_limits<float>::max()
as you suspected.