readme

fortran并行计算包

Here is a brief account on what I did with the mastermind program.

 1. I speeded up the search. When a guess is found to be inconsistent with
    the board, an attempt is made to identify the leftmost column which
    needs changing. Consequently the search continues with this column
    incremented (and all subsequent columns set to 0s). This seems to be
    resulting in a significant improvement in speed, which means that
    simpler configurations run in a blink of eye :-(, and only the ones
    with 10 colours and 8 or 9 columns take an observable time.

 2. I've slightly changed the communication scheme. Slaves now poll for
    NEW_INFO messages (this now makes sense, as more info may speed up
    their search). When they submit a guess, they still wait for an answer
    (ACCEPTED, NEW_INFO (which means their guess was rejected), EXIT or WON
    (which is just a variant of EXIT, to enable the worker to display the
    last successful guess)).


 3. I've improved the graphics. The screen now looks like the following:

	<secret code>         ||        <color scale>
    ==========================||=======================
     W  <guess_1>  <score_1>  || W  <search_state_1>
			      ||  
	 ....                 ||  .........
			      ||  
			      ||  
			      || W  <search_state_n>
			      ||  
			      ||  

    On the left hand side you see the secret code and the board as
    mainained by the master. W denotes the color-coded indication of which
    worker (slave) submitted the given guess. The top right part contains a
    <color scale> showing all the colors in search order.

    The rest of the right hand side displays the state of the slaves 1...n.
    <search_state_i> shows the current guess being processed by the ith
    slave.  Below that you see a black line showing the progress of the
    slave in its portion of the search space (the full width of the guess
    is equal to the full search space of the slave). You may see short
    lines perpendicular to this line: the black ones downwards show points
    in the search space when the worker submitted a guess which wasn't
    accepted. The double height lines upwards (of the worker's own colour)
    show points when the worker submitted a guess which was
    accepted. Shorter coloured lines upwards show points when guesses
    accepted from other workers were brought to this slave's attention,
    their color shows from whom these guesses originated.

I enclose the new version below. Have a look at it if you have some time
and please let me know your comments.

I hope to be able to do some further work on this in:

	- overcoming the limitation that columns**colors < 2**31
	- performance analysis
	- improving the communication model

Best regards,

Peter