Octochess r5178 released

[Codesquid]

Hi all,

a new version of Octochess has been released.

The main new feature in this version is full support for Fischer Random Chess, aka Chess 960. In addition, a few strength-eating search bugs have been fixed.

As always, you can download the latest version from http://octochess.org/

Please let me know if you are having any issues with the new version.

The complete list of changes:

• Full support for Fischer Random Chess, aka Chess960
• KBPvPK endgame is handled better
• Fixed too early move counter increment in search
• Fix repetition draw detection after a null-move
• Small improvements to the threading if searching with multiple cores

[Martin Thoresen]

Hi Tim,

I am confused, should I use this for TCEC instead?

[Codesquid]

Hi Tim,

I am confused, should I use this for TCEC instead?

That's not necessary. There are no functional differences between the version you have and r5178. The changes between are merely to make it easier to build the Kindle version of Octochess.

[jpqy]

Octochess 5172 (latest dev version)

Martin..maybe you can use better a dated dev version..because you never
have the latest dev version!..i also test with later dated versions from some engine in your list!
And i'am sure there are again newer dev versions!

Kind regards,
JP.

[Martin Thoresen]

Octochess 5172 (latest dev version)

Martin..maybe you can use better a dated dev version..because you never
have the latest dev version!..i also test with later dated versions from some engine in your list!
And i'am sure there are again newer dev versions!

Kind regards,
JP.

Not much I can do when programmers release new public versions before the tournament starts.

[Codesquid]

If Amazon wouldn't sneak updates onto my Kindle and thus breaking the jailbreak without my consent, I could have had the public release out a few days earlier

[Martin Thoresen]

If Amazon wouldn't sneak updates onto my Kindle and thus breaking the jailbreak without my consent, I could have had the public release out a few days earlier

No worries Tim. You are of course free to do what you want.

[Graham Banks]

Hi all,

a new version of Octochess has been released.

The main new feature in this version is full support for Fischer Random Chess, aka Chess 960. In addition, a few strength-eating search bugs have been fixed.

As always, you can download the latest version from http://octochess.org/

Thanks Tim.

[JuLieN]

Thanks Tim. Congrats and good luck for nTCEC!

But as there was no OSX binary, I tried to compile mine, and got those error messages :

Code: Select all

MBP:chess julienmarcel$ make
g++ -O3 -g -pipe -march=corei7 -Wall -Wextra -flto -static -m64 -std=gnu++0x -c -o util/logger.o util/logger.cpp
g++ -O3 -g -pipe -march=corei7 -Wall -Wextra -flto -static -m64 -std=gnu++0x -c -o util/mutex_unix.o util/mutex_unix.cpp
g++ -O3 -g -pipe -march=corei7 -Wall -Wextra -flto -static -m64 -std=gnu++0x -c -o util/platform.o util/platform.cpp
In file included from util/platform.cpp:6:0:
util/unix.cpp: In function 'int get_system_memory()':
util/unix.cpp:39:27: error: '_SC_PHYS_PAGES' was not declared in this scope
  uint64_t pages = sysconf(_SC_PHYS_PAGES);
                           ^
util/unix.cpp: In function 'void* page_aligned_malloc(uint64_t)':
util/unix.cpp:53:62: error: 'MAP_ANONYMOUS' was not declared in this scope
  void* p = mmap( 0, alloc, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0 );
                                                              ^
make: *** [util/platform.o] Error 1

Any idea on what went wrong ?

[Codesquid]

OS X is weird

For the first error, #include <sys/sysctl.h> and replace the entire function body with this:

Code: Select all

int get_system_memory()
{
	uint64_t v;size_t l=8;sysctlbyname("hw.memsize",&v,&l,0,0);return v / 1024 / 1024;
}

For the second error, replace MAP_ANONYMOUS with MAP_ANON.

I don't have access to a decent compiler on OS X, so haven't tested it myself.

After finishing compiling Octochess, make sure to run octochess test to run the selftest routines. It has to pass, otherwise the build is bad. If you then execute octochess --threads 1 --depth 18 --memory 1 --moves 1 auto, 51954300 needs to be the total number of nodes printed at the very end.