Bytewise access is always possible and won't violate the strict aliasing/effective typing rules. The worst that could happen when messing with IEEE floating poi

Bytewise access is always possible and won't violate the strict aliasing/effective typing rules. The worst that could happen when messing with IEEE floating point values that way is creating a signalling NaN (if they are available).