APDL Best Games 3/APDL Best Games 3.iso/_docs/chess/documentat/changes

Text File | 1990-05-13 | 15.4 KB | 430 lines

      GNU CHESS HISTORY
    (#include "../version.h")

January 8, 1990 -- Hans Eric Sandstrom ([e-mail])
Changed the transpositon table to allow 66536 positions per player.
This means a 1.3 Mbyte hash table. Introduced simple rehashing. And
also made setting ttable 0 exclude all transposition code from
gnuchess (for machines with very little memory).


January 3, 1990 -- Tom Vijlbrief (tnosoes!tom@nluug.n)
Contributed code to fix undo of promotions.
And a bugfix RMBLY was to small.


January 27, 1990 -- Hans Eric Sandstrom ([e-mail])
Fixed nuxdsp.c to work with my changes (Saved games format)


January 26, 1990 -- Hans Eric Sandstrom ([e-mail])
Verified that the underpromotion routines works.
Found and fixed the bug that caused gnuchess to report mate
and draw prematurely.


January 25, 1990 -- Hans Eric Sandstrom ([e-mail])
Received the underpromotion additions from [e-mail]
Made some optimizations in trapped and KingScan.
Added more comments on how to improve the search.
Added info on usage of the array Tree in post mode.
Changed the index in the transposition table to unsigned short to
allow a hash table of size 65536.


January 11, 1990 -- Hans Eric Sandstrom ([e-mail])
Formatted gnuchess according to GNU standards with indent.


January 9, 1990 -- Hans Eric Sandstrom ([e-mail])
Inserted a missing comma in KTHRT.
Removed mvstr1[5],mvstr2[5] from gnuchess.c, they are only used in the
user-interface routines uxdsp.c nondsp.c anyway.
Rewrote algbr to generate move strings for underpromotion. (ph8k or h7h8k)
Some optimizations in CaptureList.
Changed row, column and locn to macros.
Made memset default and introduced a switch NOMEMSET for computers without
memset or where memset is a speed penality. I tested this on:
Sparc -  no change, DECstation - no change and Sun3/60 - 4% speed improvment.
Changed + to | in some bit-field manipulations.
Fixed the castle-with-moved-rook-bug, removed kingmoved and introduced
an array Mvboard to keep track on how many times we have moved from a
particular square.
Fixed the check for zero division in Time controls.
Added BINDIR and LIBDIR in the Makefile to set where gnuchess and
gnuchess.book gets installed.
All I have to do now is to test if all this works (-:

February 3, 1990 -- Tom Vijlbrief
Noticed an error in the rook mobility bonus. RMBLTY array was
indexed 0 to 14, but had only 14 members. Added final member.

January 1, 1990 -- Stuart Cracraft (cracraft @wheaties.ai.mit.edu)
Some special openings were inserted and some taken out.
A small change was made to allow the program to detect draws earlier
in the tree (via Stanback).
Thinking on opponent's time is now working in nondsp.c. For some
reason it had been deactivated.
Tested GNU Chess 1.55 with with Chesstool on a SUN.
Match vs. Fidelity Mach 3 (USCF 2265) resulted in 7-3 score for GNU
on a Sparcstation giving a 2325-2335 performance result.

January 1, 1990 -- Hes @log-se.sv
He contributed faster move generator routines.

August 1, 1989 -- Jay Scott
He proofread the opening book and made
corrections.

June 9, 1989 -- Tim Radzy (unet!nplab4![e-mail])
He fixed a bug in xchess/board.c. In a post-game new-game situation,
castling wouldn't be permitted under circumstances. Tim made
it possible to castle again.

May 12, 1989 -- Joe Garbarini ([e-mail])
Recommended changes to documentation vis a vis chesstool usage.

May 5, 1989 -- Jouko Holopainen ([e-mail])
Wrote code to support underpromotion.
Changed interface to accept ECO/Informator style moves.

April 30, 1989 -- Various GNU contributors
setlinebuf() modification for xchess/chesstool.
check for zero division in time printout.

January 17, 1989 -- Anders Thulin
Provided extensive addition to the opening book for his
favorite opening the Vienna Game. This was drawn from ECO.

November 23, 1988 -- Stuart Cracraft
Installed new version of Xchess that is better debugged, works on
the next version of X. Thanks to Wayne Christopher and Arturo Perez.

August 28, 1988 -- Stuart Cracraft
Removed a sacrifice line from the Giuoco Piano entry in the opening
book; the program didn't seem to like the positions it got from this line.

December 30, 1987 -- John Stanback
Wrote a short blurb on the heuristics contained in GNU Chess. It resides
in the subdirectory DOCUMENTATION as the file HEURISTICS.

December 17, 1987 -- John Stanback
Modified criteria for positional evaluation in quiescence search
to include positions in which the estimated score lies within
the alpha-beta window; fixed a bug in the king proximity to pawns heuristic;
fixed a bug involving passed pawn heuristics;

December 16, 1987 -- Stuart Cracraft
Added automatic 'list' upon exit (both in display, non-display, and
chesstool mode); command-line setting of tournament time controls
bug fixed.

December 14, 1987 -- John Stanback
GNU defeated the commercial product 'Fidelity Excellence' 5.5-4.5 in
a 10-game match. It was running at about 500 nodes per second (typical
of its speed on a VAX 8650) and this would indicate its strength
would be about USCF 1875-1900.

December 4, 1987 -- John Stanback
Man page added. Command line arguments now specify regular clock
settings if so desired (useful for SUN players). Thinking
on opponent's time is now disabled by default. Estimated
rating is 1850 at 500 nodes per second.

October 20, 1987 -- Stuart Cracraft
Fixed GNU/SUN interaction. Chesstool and its features now
seem to fully work.

October 5, 1987 -- Ken Thompson
GNU beat Belle (actually drew due to a bug, but
Ken kept GNU playing through to the win) while
running on a Cray XMP-48. In this 3-1 time handicap game
Belle outsearched Cray GNU by 10-1 (even with the handicap).

September 26, 1987 -- John Stanback at HP
Hash table functioning. Thinking on opponent's
time functioning.

August 20, 1987   -- Mike Meyer at Berkeley
Mike ran GNU Chess on a Cray 1 supercomputer.
The system was very heavily loaded, so the
program was not as speedy as with the Cray below.

August 16, 1987   -- David Goldberg at SUN
He added "chesstool" support so that this
version of GNU Chess can run under the
display manager "chesstool".

August 15, 1987   -- John Stanback at HP
Hash tables, more heuristics, a modified
search which is more efficient. He also
discovered a bug in the piece-exchanger. This
would cause the program to exchange pieces suboptimally.
With this fix, the program should play much
more strongly.

August 13, 1987   -- Ken Thompson at Bell Labs
Ken ran GNU Chess on a Cray XMP supercomputer
 (among other processors). The program got
 about 3000-4000 chess positions per second
 which is comprable to today's fastest bit-slice
 commercial machines. Also, he had GNU Chess
 play two games against Belle.

July 19, 1987 -- Jay Scott & John Stanback
 Many positional heuristics have been added.

July 18, 1987 -- Stuart Cracraft
 Improvements have been made to the opening
 book. It is mostly an MCO book, containing
 major variations from many of the major openings
 and particularly in-depth on Sicilian.

May 11, 1987  -- John Stanback at HP
 He donated his chess program, a fairly mature
 and strong program.

May 1, 1987   -- Stuart Cracraft
 He added several bug fixes various people
 had reported. He also changed makemove() so that
 the calling syntax is makemove(movelist,index,board)
 rather than makemove(move,board). Having the latter
 tickled a bug in at least one manufacturer's C-compiler,
 so rather than write fancy code, we simplified it.

April 25, 1987-- Jim Aspnes at MIT
He added all sorts of useful capabilities,
including positional evaluation in the tree
search using a table-driven algorithm,
modifying transposition table code in order
to work properly, though it doesn't improve
speed too much, checkmates/stalemates detected
in the search, en passant captures allowed,
detect repeated positions, iterative deepening,
quicker quiescence search, tournament time controls,
sqattacked sped up by a factor of 4, compile-time
debugging options.

January 2, 1987   -- Stuart Cracraft
 He added a few more Tal games to the collection.

January 2, 1987   -- Jim Aspnes at MIT
 He contributed MCO variations for the Catalan,
 Queen's Indian, and Reti openings.

December 29, 1986 -- Jim Aspnes at MIT
 He contributed all MCO variations of the Najdorf
 to the opening book. He also contributed a LISP
 macro (written in GNU Emacs Lisp) to convert
 xchess game formats to GNU Chess opening book
 format.

December 14, 1986 -- Ken Thompson at Bell Labs
 He contributed almost 200 games by Tal to
 our collection of Tal-games, bringing the
 total number of Tal positions in the book
 to 10,692. Total book positions now 13,207.
 These reside in bookin, bookin.bdg, bookin.tal.
 Note that presently, only bookin and bookin.tal
 can be used. The new Tal positions came in a
 slightly different format, which we have chosen
 to adopt as our standard format. All book
 games in bookin and bookin.bdg will gradually
 change into the new standard format.

December 11, 1986 -- Stuart Cracraft
 Added "averages" for node-count per move,
 cpu per move, rate per move to list_history
 and write_history.
 New version of Xchess installed.
 Started typing in Tal games into "bookin.tal".
 Added "total book positions" printout to "book"
 and "enter" statistics printout.

December 10, 1986 -- Stuart Cracraft
 Implemented aspiration search in normal
 alpha-beta search. Speedups of 3% to 40%
 have been noticed in most positions.
 Occasionally a slower search will result,
 but it is thought these are worth the
 usual speedups.

December 9, 1986  -- Stuart Cracraft
 Fixed minor bug in write_history()
 Added another Tal game, 2nd game of 1st world
 championship match with Botvinnik, a Benoni.

December 9, 1986  -- Stuart Cracraft
 Enhanced parallelism. All parallel processors
 now communicate via a shared data file and
 are kept running (in idle loops watching the
 shared data file). This saves us a few seconds
 on each move since the 'rsh' need not be invoked
 more than once (at the beginning). Since the
 shared data file is now implemented, we will
 next work towards a "parallel anarchy" in which
 any processor can use any other processor in
 order to reduce its search. The current scheme
 with the program being only as fast as its slowest
 processor, is quite inefficient.

December 1, 1986 --  Jim Aspnes at MIT
 Added a couple of Master games from
 Modern Chess Openings 12 (a Fischer game,
 and a Matanovic game).

November 30, 1986 -- Stuart Cracraft
 Added parallelism. Can now handle multiple
 processors (sharing same disk). Later we will
 add the capability to use processors not sharing
 the same disk. Modified README and MAN-PAGE.

November 26, 1986 -- Stuart Cracraft
 Fixed a few bugs in book-mailing mechanism.
 Fixed a bug regarding situations where only
 one move is available.
 Fixed a bug in read_history() that caused
 Black queenside castles to be mishandled.

November 25, 1986 -- Stuart Cracraft
 Added two pawn heuristics. Reward pawns moving into
 a phalanx of pawns. A phalanx is two or more
 horizontally-connected pawns. Likewise, penalize
 pawns leaving a phalanx of pawns. The penalty for
 leaving is a little more than the reward for
 entering.

November 24, 1986 -- Stuart Cracraft
 A user reported an unbelievable bug. Investigation
 of this bug led to the discovery that GNU Chess was
  not picking the move judged best by the tree search
 in all cases. This resulted in the bug showing
  itself which further showed that the program was
 selecting an inferior move. This may result in an
 improvement to the program's play.

November 24, 1986 -- Stuart Cracraft
 Added two heuristics. Penalize king moves if
 the king hasn't castled yet. Also, penalize pawn
 moves which produce doubled pawns. Should
 probably have something for isolated pawns
 too.

November 23, 1986 -- Wayne Christopher at Berkeley
 New version of X chess display front-end.
 Fixed bugs include multiple pieces, runs
 on SUNS & Bobcats, loads saved games.

November 23, 1986 -- Stuart Cracraft
 Cleaned up some minor bugs regarding history.
 Added "Illegal command" error message at Wayne's
 request.

November 22, 1986 -- David Goldberg at SUN Microsystems
 He complained that GNU Chess was memory-hungry.
 A few minor modifications to hash.c reduced
 uninitialized data space 87% and text space
 12%. This should make it easier for GNU Chess
 to run on small computers.

November 22, 1986 -- Stuart Cracraft
 "read" command was working, but needed
 additional tweaking so that history
 array would be printed by list_history().

November 19, 1986 -- Stuart Cracraft
 Added "read" command which reads a history
 file (game listing) and restores the board
 to as if the person was still playing that.
 particular game. Generally cleaned up
 history mechanism, made it more orthogonal.
 Revised README. Added doc to MAN-PAGE.

November 16, 1986 -- Stuart Cracraft
 More opening book bugs found and fixed.
 Added capability to accept abbreviated-algebraic notation
 for entering "book" games from files.
 Added approximately 2500 new positions to
 opening book from games involving the
 opening called Blackmar-Diemer Gambit,
 a hoary line developed by Diemer in
 Germany years ago.

November 15, 1986 -- Wayne Christopher at Berkeley
 He modified the move generator, resulting in
 a 28% speedup.

November 14, 1986 -- Stuart Cracraft
 He documented a lot of the GNU Chess modules
 with brief comments for each function. More
 extensive internal documentation may go in
 later.

November 14, 1986 -- Wayne Christopher at Berkeley
 He created the Xchess interface for
 GNU Chess to have windowing with X windows.

November 14, 1986 -- Stuart Cracraft
 He added a "randomization" feature to
 the opening book. This will cause the
 program to select randomly from alternate
 variations whenever -DBEST is removed
 from Makefile's CFLAGS. If this is not
 removed, the opening play selects the
 first move found in the book as it appears
 "in order" in the human-readable book.

November 14, 1986 -- David Goldberg at SUN Microsystems
 He responded to a query about dbm(3) which
 eventually resulted in the fixing of a subtle
 bug in the book code which was causing the
 program to sometimes hash to the incorrect
 address and thereby produce a book move which
 didn't even exist in the book. Thanks David!

November 14, 1986 -- Stuart Cracraft
 He added the "oboard" routine in util.c. This
 is the reverse of the already extant "iboard"
 (same module). These two routines translate
 between GNU Chess internal format and
 Forsythe notation.

November 10, 1986 -- Stuart Cracraft
 He added the "enter" command. This causes
 the current game to be entered in the book.
 Then, GNU Chess tries to mail this new entry
 to the book maintainers (for inclusion in
 the master copy of the book).

November 9, 1986 -- Stuart Cracraft
He added code for an opening book. MAN-PAGE
and README were modified accordingly.

November 8, 1986 -- Stuart Cracraft
Checks and mates are now noticed at ply-1.
This is a more complete fix to the Oct 31 fix.

October 31, 1986 -- Stuart Cracraft
First attempt at fix to bug which causes
program to check human's king when program
itself is in check.

October 31, 1986 -- Mly at MIT
Reported a bug which caused program to crash
when an illegal human move was played. Fixed.
Also, program was unable to play as White. Fixed.

October 22, 1986 -- Stuart Cracraft
Pps now rewards moves which liberate bishops.

October 19, 1986 -- Stuart Cracraft
Added bitmapper routines to distribution.
Added version notice.

October 19, 1986 -- David Goldberg at SUN Microsystems
Interfaced GNU Chess with SUN's chesstool.

October 18, 1986 -- Initial release date.