This will be used to implement the half-precision floating-point variants of FCMEQ in following changes.