Signed-off-by: Matteo Bruni <mbruni at codeweavers.com> --- Note to myself, mostly: this is only the float version, we still need to generate imax / umax for integers (together with all the other integer ops).