In the XEM6310 UCF file (maybe others, haven’t looked) the period constraint on okHostClk is 9.96 ns. It should be 9.92 ns, corresponding to 100.8 MHz. Shouldn’t really be a problem (are you really going to miss timing by 0.4 MHz?) but it did generate a warning in ISE about periods not matching when I put my own CLKIN1_PERIOD generic in a PLL.