Most VHDL and Verilog simulators do support floating point. Even if your FPGA logic uses integer arithmetic, floating point is very useful in the simulation testbench to generate stimulus waveforms, and to help analyze the results.
Your ISE 7.1i may have an older version of the floating-point core library. I don't recall which version of ISE introduced it.
I expect Xilinx or Al.tera will someday put floating-point arithmetic blocks into their FPGAs. When that happens, they will probably add floating-point support to their VHDL/Verilog synthesizer.