https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116340
--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> --- When it was originally added, the conconial form was -abs<x> but then it was changed to be either -abs<x> or copysign<, -1.0> depending on if there was an optab for copysign or not. For soft-float there should always be an optab except when dealing with non IEEE floating point types where the sign bit is a bit simple.
