Crafty-22.9 is available

[Bo Persson]

crafty has started to use c++ features in variable declarations?

iterate.c

if (iteration_depth > 3) {
time_used = ReadClock() - start_time;
int time_left = time_limit - time_used;
int prev_ply = ReadClock() - prev_time;

Aha. That is something Jason changed. Peter Skinner was having problems compiling, probably with a compiler that didn't appreciate this. I will change it and re-copy the source so that no one will have difficulty...

This is C99, so it has been part of C for 10 years already.

Yoy have to chose your compatibility level, of course.

[swami]

Crafty 23.0 is playing at ICC?! Only a day ago, 22.9 was released... now It's new version already... What changes have been made, Bob?

[Dirt]

Crafty 23.0 is playing at ICC?! Only a day ago, 22.9 was released... now It's new version already... What changes have been made, Bob?

That's a little surprising, as he said he was planning on releasing 22.10 to address the draw offer problem. However, he has said that 23.0 would appear soon, the new full version number to highlight book incompatibility.

[bob]

Crafty 23.0 is playing at ICC?! Only a day ago, 22.9 was released... now It's new version already... What changes have been made, Bob?

New hash random numbers statically initialized, plus a couple of minor other things, nothing significant yet. I slightly tweaked the PRNs to get rid of a couple that were just a few bits different, but that wrecks the book completely, hence the new version change. If you get a 23.x version, pre-23.x version books (.bin files) will not work...

[bob]

Crafty 23.0 is playing at ICC?! Only a day ago, 22.9 was released... now It's new version already... What changes have been made, Bob?

That's a little surprising, as he said he was planning on releasing 22.10 to address the draw offer problem. However, he has said that 23.0 would appear soon, the new full version number to highlight book incompatibility.

I had already done this (new version). If Crafty does OK during the night, I will release a 22.10 tomorrow with the draw offer change made...

[bob]

crafty has started to use c++ features in variable declarations?

iterate.c

if (iteration_depth > 3) {
time_used = ReadClock() - start_time;
int time_left = time_limit - time_used;
int prev_ply = ReadClock() - prev_time;

Aha. That is something Jason changed. Peter Skinner was having problems compiling, probably with a compiler that didn't appreciate this. I will change it and re-copy the source so that no one will have difficulty...

This is C99, so it has been part of C for 10 years already.

Yoy have to chose your compatibility level, of course.

You would not believe how many still use very old versions of MSVC that won't accept this (among other things).

[Dann Corbit]

On my machine it is 3% faster for bench, and I am comparing it to an Intel build which usually clobbers my build (by anywhere from 10-30%).

C:\chess\winboard\crafty>crafty-22.9
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 4

Crafty v22.9 (4 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 132210433
Raw nodes per second: 12019130
Total elapsed time: 11.82
White(1): quit

C:\chess\winboard\crafty>crafty-22.8-win64.exe
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
72127kb of RAM used for TB indices and decompression tables
max threads set to 4

Crafty v22.8 (4 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 128172256
Raw nodes per second: 11652023
Total elapsed time: 11.45
White(1):

Here is the JA compile on my machine:
C:\chess\winboard\crafty_229_ja\speed compiles>crafty_229_x64_ja.exe
unable to open book file [./book.bin].
book is disabled
unable to open book file [./books.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
book file enabled.
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
72192kb of RAM used for TB indices and decompression tables
max threads set to 4

Crafty v22.9 JA (4 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 104852679
Raw nodes per second: 13106584
Total elapsed time: 8.87
White(1):

[Werewolf]

This is all very interesting but this latest version of Crafty on my Quad is still being hammered by the ancient Fritz 7 (running on only 1 cpu) on my tactics test...

[glorfindel]

This is all very interesting but this latest version of Crafty on my Quad is still being hammered by the ancient Fritz 7 (running on only 1 cpu) on my tactics test...

Is your tactics test secret?

[bob]

That could well be. Crafty has been de-tuned for tactics, because all my tuning has been in real games on my cluster. Most search extensions are gone, which will hurt tactical test trials, but which helps real game results.

It all depends on what you are looking for. If Fritz 7 can beat it in real games, that would be important news. But not in tactical tests...