Crafty 22.2 is out!

[Denis P. Mendoza]

Just catched a new update. Source is at Bob's site and executables at Peter Skinner's site. Just informing .

For those who are not familiar:

http://www.webkikr.net/ -Peter Skinner's executables

ftp://ftp.cis.uab.edu/pub/hyatt/source/ -Crafty source codes

[Volker Pittlik]

Just catched a new update. Source is at Bob's site ...

Is there a description of the personality files somewhere? Also I guess the skill command is still working because the -DSKILL compiler option still works?

Volker

[mike schoonover]

hi all,
anyone doing a compile for my older celeron 500 mhz,
runnimg windowsme 32 bit?

would apreciate it much.
regards
mike

[Denis P. Mendoza]

hi all,
anyone doing a compile for my older celeron 500 mhz,
runnimg windowsme 32 bit?

would apreciate it much.
regards
mike

I'll try my best. I thought Peter's compiles are all-purpose. I'm sure Master Jim is also listening.

Pls. wait a while as I'm profiling some beta engines. I'll study this code again as there may be some special approaches during compiling! I t reverted back to Posix threads.

Denis

[Denis P. Mendoza]

Just catched a new update. Source is at Bob's site ...

Is there a description of the personality files somewhere? Also I guess the skill command is still working because the -DSKILL compiler option still works?

Volker

I guess there's no personality description, except the definitions of the 'skill' option in lines 3543-3550 at main.c and based on v22.1:

Code: Select all

new "skill" command that can be  *
 *           used to "dumb down" crafty.  "skill <n>" where n is a number      *
 *           between 1 and 100.  100 is max (default) skill.  skill 70 will    *
 *           drop the playing Elo by about 200 points.  skill 50 will drop it  *
 *           about 400 points.  the curve is not linear, and the closer you    *
 *           get to 1, the lower the rating.  To use this feature, you need to *
 *           add -DSKILL to your Makefile options otherwise it is not included *
 *           in the executable.                                                *

So personality is between skill 1 to skill 100. I don't know if using the Crafty personality file option (*.cpf) is still usable. I read an artcile that Robert Jurjevic is still using the crafty.cpf file for his Crafty 22.xrjy personalities.

The new infos for v22.2 is this:

Code: Select all

22.2    we are now back to using POSIX threads, since all current Linux   *
 *           distributions now use the posix-conforming NTPL implementation    *
 *           which should eliminate the various compatibility issues that      *
 *           caused problems in the past.  this also should make the new       *
 *           smpnice mode work correctly for Linux and Windows since they will *
 *           now both use threads for the SMP search.  Fruit-like scoring      *
 *           (interpolation between mg and eg scores) fully implemented.  AEL  *
 *           pruning (Heinz 2000) also fully implemented (we had razoring and  *
 *           futility, but not extended futility).  "eval" command has been    *
 *           removed and combined with the "personality" command so that eval  *
 *           parameters can be modified, in addition to some search parameters *
 *           that also belong in the personality data.  mate threat extension  *
 *           and one legal reply to check extensions have been removed.  tests *
 *           proved them to be absolutely useless, over 30,000 games for each  *
 *           test showed no gain and sometimes a loss in playing strength with *
 *           them so we followed the "simple is better" and removed them.  The *
 *           fractional ply code was also removed since the only extension we  *
 *           now use is the "give check" extension which is a whole ply.  A    *
 *           significant number of evaluation parameters have been changed,    *
 *           a few even removed as cluster testing helped us find optimal      *
 *           values.  There are a few new terms, with more planned.            *

Sorry if I couldn't give you further ideas. I'll have to read the code again and test what new features it has.

Denis

[Jim Ablett]

(Great logo sent to me by Oscar A Orozco)
Crafty 22.2 by Robert Hyatt

Windows x64 & Win32 Intel compiler 10 builds.

http://www.mediafire.com/?k1tjmetnjjy

Thank-you Bob !

There is a bug (as reported earlier) in the Windows code.
In console mode type 'go' or try to analyze an epd and Crafty will crash.
It exhibits this behaviour in both my builds and Peter Skinner's.

I've included an additional win32 exe to test, built with Gcc-cygwin
compiler which uses the Linux src.
This builds is slower, but works fine in console mode.

Jim.

[Denis P. Mendoza]

Thanks Jim for the builds.

I had a hard time last evening figuring out the now-you-see-it, now-you-don't errors from lockl.h and utility.c (concerning 'NumaStartThread'). At least I finally got a stable all-purpose build for Mike Schoonover's old pc. It could be much slower than yours and Peter, and may have some bugs you mentioned . I don't know exactly about the bug issue but I checked in Arena using different epd positions and it analyzes with no problems with smpnice=1. And sad thing was, I forgot to add the -DSKILL option. Sorry Mike!

Anyway, here's my 32-bit build..no guarantees....yet works on my celeron too!

http://computerchessengines.mylivepage. ... ileid=4000


i'll take a look on your fixed version Jim!

I forgot...Thanks bob for this update.

[Dann Corbit]

The error occurs when addressing element 66 of the tree.

Here are the allocations for the tree in the source code:
[code]
init.c ( 128): mem = malloc(MAX_BLOCKS_PER_CPU * ((sizeof(TREE) + 2047)) & ~2047);
init.c ( 131): (TREE *) ((long) mem + j * ((sizeof(TREE) + 2047) & ~2047));
init.c ( 135): memset((void *) block[i + 1], 0, sizeof(TREE));
main.c ( 3625): (TREE *) ((~(size_t) 127) & (127 + (size_t) malloc(sizeof(TREE) + 127)));
thread.c ( 285): memset((void *) block[j * MAX_BLOCKS_PER_CPU + i + 1], 0, sizeof(TREE));
thread.c ( 316): (TREE *) ((~(size_t) 127) & (127 + (size_t) WinMalloc(sizeof(TREE) +
[code]

It looks to me like Bob is trying to align the memory on some sort of special address. I notice that we have two different types of calculations also -- one based on 127, and the other on 2047.

On line 71 of nextr.c, when i is == 65:

[code]
if (block[i]->used) // here block[i] is a null pointer.
[/code]

block[i-1] is a valid pointer.

[Dann Corbit]

Here is the problem in file init.c:

Code: Select all

#if defined(_WIN32) || defined(_WIN64)
  ThreadMalloc((int) 0);
#else


This allocates memory only for CPU 0, but not for CPUS
1 .. CPUS
The code should be:

Code: Select all

#if defined(_WIN32) || defined(_WIN64)
{
size_t i;
for (i=0; i < CPUS; i++)
  ThreadMalloc((int) 0);
}
#else


Also, crafty.c should have this inserted at the top, if you compile via that module:

Code: Select all

#ifdef _MSC_VER
#include <windows.h>
#else
#include <unistd.h>
#endif


[Dann Corbit]

Of course that should be:

Code: Select all

#if defined(_WIN32) || defined(_WIN64)
  {
   size_t index;
   for (index = 0; index < CPUS; index++)
      ThreadMalloc(index);
  }
#else


I will make a patched version available shortly.

[Dann Corbit]

Of course that should be:

Code: Select all

#if defined(_WIN32) || defined(_WIN64)
  {
   size_t index;
   for (index = 0; index < CPUS; index++)
      ThreadMalloc(index);
  }
#else


I will make a patched version available shortly.

Here is the updated source code with Microsoft VC++ solution files and a 64 bit PGO binary:
http://cap.connx.com/chess-engines/new- ... ty22-3.zip

Some output:

Code: Select all

EPD Kit revision date: 1996.04.21
found computer opening book file [./bookc.bin].

Initializing multiple threads.
System is SMP, not NUMA.
hash table memory =  192M bytes.
pawn hash table memory =    8M bytes.
EGTB cache memory =   32M bytes.
ERROR "drawscore=0 " is unknown rc-file option
choose from book moves randomly (using weights.)
choose from 5 best moves.
book learning enabled
result learning enabled
resign after 5 consecutive moves with score < -9.
EGTB access enabled
using tbpath=c:\chess\winboard\Nalimov;c:\arena\engines\nalimov;
6 piece tablebase files found
72095kb of RAM used for TB indices and decompression tables
max threads set to 2

Crafty v22.3 (2 cpus)

White(1): mt 4
max threads set to 4
White(1): bench
Running benchmark. . .
......
Total nodes: 33197318
Raw nodes per second: 11102781.939799
Total elapsed time: 2.990000
White(1): st 180
search time set to 180.00.
White(1): epdpfga wac230.epd wac230.out
PFGA: EPD record: 1   ID: WAC.230
              time surplus   0.00  time limit 3:00 (+0.00) (3:00)
              depth   time  score   variation (1)
               12->   0.05  -2.47   1. ... Rf7 2. Ba3 a4 3. f4 Bb7 4. Kf3
                                    Rh7 5. Kg3 Ba6 6. Kg4 Bb5 7. f5 gxf5+
                                    8. Kf4
               13     0.05  -2.43   1. ... Rf7 2. Ba3 a4 3. f4 Bb7 4. Re2
                                    Rh7 5. Rh2 Rd7 6. Kf3 <HT>
               13->   0.06  -2.43   1. ... Rf7 2. Ba3 a4 3. f4 Bb7 4. Re2
                                    Rh7 5. Rh2 Rd7 6. Kf3 <HT> (s=2)
               14     0.08  -2.43   1. ... Rf7 2. Ba3 a4 3. f4 Bb7 4. Re2
                                    Rh7 5. Rh2 Rc7 6. Rb2 Rd7 7. Kf2 Rh7
                                    <HT>
               14     0.20  -2.59   1. ... Rh7 2. Kf4 a4 3. Ba3 Rh1 4.
                                    Rd2 Ba6 5. Bb2 Rh2 6. Ba3 Rh4+ 7. Kg3
                                    Rh5 8. Kf4
               14->   0.20  -2.59   1. ... Rh7 2. Kf4 a4 3. Ba3 Rh1 4.
                                    Rd2 Ba6 5. Bb2 Rh2 6. Ba3 Rh4+ 7. Kg3
                                    Rh5 8. Kf4
               15     0.59  -2.59   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh2 5. Kg3 Rh7 6. Kf4 Rh4+ 7. Kg3
                                    Rh8 8. Be7 Rh5 9. f4 Bb7 <HT>
               15->   0.63  -2.59   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh2 5. Kg3 Rh7 6. Kf4 Rh4+ 7. Kg3
                                    Rh8 8. Be7 Rh5 9. f4 Bb7 <HT>
               16     0.70  -2.50   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh2 5. Kg3 Rh8 6. f3 Rh7 7. Rb2
                                    Rh1 8. Kg2 Ra1 9. Bc5
               16->   0.80  -2.50   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh2 5. Kg3 Rh8 6. f3 Rh7 7. Rb2
                                    Rh1 8. Kg2 Ra1 9. Bc5 (s=2)
               17     0.94  -2.56   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh3 5. f3 Rh2 6. Rb2 Rh1 7. Rd2
                                    Bb7 8. Bb2 Rh8 9. Ke3 Ra8
               17->   1.17  -2.56   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh3 5. f3 Rh2 6. Rb2 Rh1 7. Rd2
                                    Bb7 8. Bb2 Rh8 9. Ke3 Ra8 (s=2)
               18     1.47  -2.55   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh3 5. f3 Rh2 6. Rb2 Rh1 7. Rd2
                                    Bb7 8. Rg2 Rh4+ 9. Kg3 Rh5 10. Kf4
                                    Rh1
               18->   2.05  -2.55   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh3 5. f3 Rh2 6. Rb2 Rh1 7. Rd2
                                    Bb7 8. Rg2 Rh4+ 9. Kg3 Rh5 10. Kf4
                                    Rh1 (s=2)
               19     2.65  -2.56   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh3 5. f3 Rh2 6. Rb2 Rh1 7. Rd2
                                    Bb7 8. Rg2 Rh4+ 9. Ke3 Rh5 10. Bb2
                                    Rh1 11. Kf4
               19->   3.93  -2.56   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh3 5. f3 Rh2 6. Rb2 Rh1 7. Rd2
                                    Bb7 8. Rg2 Rh4+ 9. Ke3 Rh5 10. Bb2
                                    Rh1 11. Kf4 (s=4)
               20     5.29  -2.56   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh3 5. f3 Rh2 6. Rb2 Rh1 7. Rd2
                                    Bb7 8. Bb2 Rh4+ 9. Ke3 Rh5 10. Kf4
                                    Rh8 11. Ke3 <HT> (s=3)
               20->   7.65  -2.56   1. ... Rh7 2. Rb1 a4 3. Ba3 Rh5 4.
                                    Kf4 Rh3 5. f3 Rh2 6. Rb2 Rh1 7. Rd2
                                    Bb7 8. Bb2 Rh4+ 9. Ke3 Rh5 10. Kf4
                                    Rh8 11. Ke3 <HT> (s=3)
               21    11.97  -2.57   1. ... Rh7 2. Rb1 a4 3. Ba3 Kb5 4.
                                    Kg4 Ba6 5. Kf4 Rh2 6. Kg3 Rh5 7. Kg4
                                    Bb7 <HT> (s=2)
               21->  15.09  -2.57   1. ... Rh7 2. Rb1 a4 3. Ba3 Kb5 4.
                                    Kg4 Ba6 5. Kf4 Rh2 6. Kg3 Rh5 7. Kg4
                                    Bb7 <HT> (s=2)
               22    20.22  -2.57   1. ... Rh7 2. Rb1 a4 3. Ba3 Kb5 4.
                                    Kg4 Ba6 5. Kf4 Rh2 6. Kg3 Rh5 7. Kg4
                                    Bb7 8. Be7 <HT>
               22->  23.75  -2.57   1. ... Rh7 2. Rb1 a4 3. Ba3 Kb5 4.
                                    Kg4 Ba6 5. Kf4 Rh2 6. Kg3 Rh5 7. Kg4
                                    Bb7 8. Be7 <HT>
               23    34.43  -2.57   1. ... Rh7 2. Rb1 a4 3. Ba3 Kb5 4.
                                    Kg4 Ba6 5. Kf4 Rh2 6. Kg3 Rh5 7. Kg4
                                    Bb7 <HT>
               23->  44.21  -2.57   1. ... Rh7 2. Rb1 a4 3. Ba3 Kb5 4.
                                    Kg4 Ba6 5. Kf4 Rh2 6. Kg3 Rh5 7. Kg4
                                    Bb7 8. Bd6 <HT>
               24     2:05  -2.49   1. ... Rh7 2. Rb1 a4 3. Ba3 Bd7 4.
                                    Rf1 Rh5 5. Bc1 Be8 6. Kg2 Rh7 7. Rh1
                                    Ra7 8. Ba3 Rf7 9. Bc1 Rf5 10. Rh8 Kd7
                                    11. Rh7+ Kc8 12. Ra7 Kb8 13. Re7 Bf7
              time=3:00  mat=-1  n=2781817512  fh=89%  nps=15.4M
              ext-> check=70.7M qcheck=84.0M reduce=1503.7M/424.9M
              predicted=0  evals=1674.7M  50move=0  EGTBprobes=39  hits=39
              SMP->  splits=26011  aborts=1962  data=23/256  elap=3:00
terminating SMP processes.Black(1):