Here is what I used:

   void NvWriteByte(U8 xdata *dptr, U8 val)
   {
   WMCON |= EEMWE_;
   *dptr = val;   
   while((WMCON&EERDY_) == 1)          // Note that each byte now takes
      ;                                // 4ms instead of 2.5ms.
   while((WMCON&EERDY_) == 0)
      ;
   WMCON &= ~EEMWE_;
   }