The x86 JIT should implement the following optimization:
  1) aligh method entry points on a 16-byte boundary
  2) optimize within extended basic blocks. For example, currently
     for the code:
        if((a[i] < 0)
           a[i] = 0;

     the JIT will bounds check for the second reference to a[i] even though
     the check was already done in the previous line.
  3) Generate integer divisions by a constant power of 2 with shifts
     and masks instead of the "idiv" instruction.