Floating point arithmetic Verilog