Distributed perft work unit file record format

[sje]

Some ideas for the distributed perft work unit file record format:

1) ASCII file, line oriented

2) Each record is one position

3) Fields are separated by a space character

4) First six fields are the FEN for the position

5) Field seven is the integer draft requirement

6) Field eight is the integer count of occurrences

7) [Calculated] Field nine is the integer perft(draft) of the position

8) [Calculated] Field ten is the integer product of fields six and seven

For the FEN half move clock (FEN field five), it is possible for its value to be dependent upon the path taken but without any difference in available moves. To ensure consistency, of all the different half move clock values possible, the lowest one will be output.

Fields past field ten may be strings which describe the items about the processing status and work assignment.

A soon-to-be-released version of CookieCat will produce these files upon command. There will likely be some limit as to the depth, perhaps 8 or 9 ply.

[sje]

Looking at this a bit more, I think that field seven should be either swapped with field eight or be deleted. The idea here is that the first N consecutive fields per record should be the same regardless of the actual depth in play for a perft(depth) calculation.

And given some arcane en passant situations, the half move clock should take the highest of several effectively equivalent values for a given position.

[stegemma]

Some ideas for the distributed perft work unit file record format:

1) ASCII file, line oriented

2) Each record is one position

3) Fields are separated by a space character

4) First six fields are the FEN for the position

5) Field seven is the integer draft requirement

6) Field eight is the integer count of occurrences

7) [Calculated] Field nine is the integer perft(draft) of the position

8) [Calculated] Field ten is the integer product of fields six and seven

For the FEN half move clock (FEN field five), it is possible for its value to be dependent upon the path taken but without any difference in available moves. To ensure consistency, of all the different half move clock values possible, the lowest one will be output.

Fields past field ten may be strings which describe the items about the processing status and work assignment.

A soon-to-be-released version of CookieCat will produce these files upon command. There will likely be some limit as to the depth, perhaps 8 or 9 ply.

I would suggest:

0) a way to define the version of the format, to handle further modification and still can accept old version results, maybe something in the first line?
1) you should well define line termination: cr, lf or cr/lf?
3) using tab instead of space allows to have FEN occupies just one field