------------------------------------------------------------
IF
    you work on a Linux or Unix system with Python installed
  AND
    copies of my files
        "sample.py", "explore.py", "params.py",
        "after.py" and "sle0.dat"
    are aleady in your working directory
THEN
    you should be able to use my programs, as follows.
------------------------------------------------------------

1. Copy "sle0.dat" to "sle.dat" by the command
      >cp sle0.dat sle.dat

2. Run the "sample" program by the command
      >python sample.py
   It should type on your terminal lines like these:

      [0, 50] 16 1 0.03 0 0.4

      0 24 0.22
      1 54 0.44
      2 78 0.82
        . . .
      47 504 9.52
      48 54 9.81
      49 553 10.01
      clock:  10.01

    The real numbers may differ (these are time, in seconds),
    but the integers must conform. If they do, the program
    works as it should, and you can use it yourself, as
    explained below.
------------------------------------------------------------

1. Initialize the system by
      >cp sle0.dat sle.dat

2. Choose the parameters by editing the file "params.py":
    the parameter "n" of the paper (called "size" in the
      programs) will be computed as 2**size_code (enter
      size_code accordingly),
    the parameter "k" of the paper (called "perturb" in the
      programs) will be computed as 2**perturb_code (enter
      perturb_code accordingly),
    "trials=[0,50]" means that trials number 0,1,2,...,48,49
      will be performed; enter another interval as you wish,
      but do not request trials number 250 and more,
    and leave intact the last three parameters (unless you
      know what are you doing).

3. Run the "sample" program by
      >python sample.py

4. Before finishing, the program writes the results to
       "sle.dat";
     normally, older results remain intact;
     but if you use, say, "trials=[0,50]" and later
       "trials=[40,100]" for the same size_code and
       perturb_code, then trials number 40,...,49 are
       repeated, and new results overwrite old results
       (but they must be the same).

5. Accumulate as many results as you wish.

6. Analyse the results by the program "after"; for example,
   the command
      >python after.py 4 0
   checks all results for size_code=4 and perturb_code=0
   (that is, n=16 and k=1) and gives you some statistics:
   quantiles 10%, 25%, 50% (the median), 75%, 90%,
   numbers of the corresponding trials, etc.

   However, all results larger than 0.4 (it is the last
   parameter "bound" in "params.py") are replaced with 0.4.
   Increase the bound if you wish; but then the program
   "sample" will work slower.
------------------------------------------------------------
Further, use the force - read the source.

         Enjoy!  Yours, Boris Tsirelson.
------------------------------------------------------------
            --------------------------------