Connecting FPU core on Virtex II Pro

I am interested in using a floating-point unit on a Virtex II Pro system as software emulation is very slow. Does Xilinx have any plans to develop such a unit? I am thinking about trying to implement a simple FPU in the FPGA to speed things up, but I am not sure as to the best way to connect it to the PPC core. I would like this core to be accessible to any C/Linux program running on the PowerPC unit if possible. According to the datasheet it is possible to trap floating-point operations and handle them that way, but I'm concerned that the overhead may negate any gains of hardware implementation. Any advice?