Octochess r4529

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
Codesquid
Posts: 138
Joined: Tue Aug 23, 2011 10:25 pm
Location: Germany

Re: Tim_Jim- What Am I overlooking Here!?

Post by Codesquid »

4 MB, independent on whether it's compiled as 32bit or 64bit program.
nanos gigantium humeris insidentes
User avatar
lucasart
Posts: 3232
Joined: Mon May 31, 2010 1:29 pm
Full name: lucasart

Re: Octochess r4529

Post by lucasart »

Codesquid wrote:After almost a month since the first release, there's finally a new version of Octochess.

Changes:
  • Improved playing strength
  • Experimental UCI implementation
  • Many small bugs fixed
  • Add alternative popcount implemention for systems without POPCOUNT.
  • If compiled as 32bit program, clamp transposition table size to 1GB
  • Some code refactoring
You can download the new version from http://octochess.org/octochess-r4529.7z, the archive contains both Windows and Linux binaries.
Could you implement the Hash command ? By default OctoChess seems to take over 1 GB or RAM, which means it's not possible to compare it to other engines accurately (I use 64 MB of Hash in my rating list).
User avatar
lucasart
Posts: 3232
Joined: Mon May 31, 2010 1:29 pm
Full name: lucasart

Re: Octochess r4529

Post by lucasart »

Werner wrote:
Jim Ablett wrote:I played some test games against Nejmet and it seems of similar strength.

Jim.
I have played some games too:

Code: Select all

Octochess r4558 x64 - Nejmet 3.07 : 22,0/27 19-2-6 (11=1111===111111111=110=110)  81%  +252
and I see here a quite different result :)
This is strange. Did you make sure they were using the same hash size ?
User avatar
Jim Ablett
Posts: 1383
Joined: Fri Jul 14, 2006 7:56 am
Location: London, England
Full name: Jim Ablett

Re: Octochess r4529

Post by Jim Ablett »

lucasart wrote:
Werner wrote:
Jim Ablett wrote:I played some test games against Nejmet and it seems of similar strength.

Jim.
I have played some games too:

Code: Select all

Octochess r4558 x64 - Nejmet 3.07 : 22,0/27 19-2-6 (11=1111===111111111=110=110)  81%  +252
and I see here a quite different result :)
This is strange. Did you make sure they were using the same hash size ?

I'm at fault here. I made an assumption based on watching about a dozen test games with 32 bit version.

Jim.
User avatar
Codesquid
Posts: 138
Joined: Tue Aug 23, 2011 10:25 pm
Location: Germany

Re: Octochess r4529

Post by Codesquid »

lucasart wrote:Could you implement the Hash command ?
Already working on this.
nanos gigantium humeris insidentes
User avatar
Codesquid
Posts: 138
Joined: Tue Aug 23, 2011 10:25 pm
Location: Germany

Re: Octochess r4529

Post by Codesquid »

Codesquid wrote:
lucasart wrote:Could you implement the Hash command ?
Already working on this.
It's in the repository.
nanos gigantium humeris insidentes
User avatar
geots
Posts: 4790
Joined: Sat Mar 11, 2006 12:42 am

Re: Tim_Jim- What Am I overlooking Here!?

Post by geots »

Codesquid wrote:4 MB, independent on whether it's compiled as 32bit or 64bit program.


Thank you for taking the time to help me. I appreciate it very much.


Best,

george
User avatar
Werner
Posts: 2871
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: Tim_Jim- What Am I overlooking Here!?

Post by Werner »

Codesquid wrote:
Werner wrote:
Codesquid wrote: Example for 256MB hash:

Code: Select all

octochess.exe --hash 256
...can you please give an example how this works in Arena??
I can't, I'm not familiar with Arena.
... and why does the engine sometimes in an endgame use more than 1CPU? I sometimes saw 100 % cpu usage on my quad here??
Octochess is multi-threaded. First root node is searched single-threaded, the other root nodes are searched in parallel. You can control the number of search threads with the --threads argument.
:cry: ok, this is the reason I have had such different results. I could not see the mp feature. This is normally a uci command too. So if you don´t implement it, we cannot test the engine. I am afraid I have to throw away my played games as I wanted to ply only single threaded games (1CPU - 1CPU). As you read before I too was not able to set the hash command...

best wishes
Werner
User avatar
Werner
Posts: 2871
Joined: Wed Mar 08, 2006 10:09 pm
Location: Germany
Full name: Werner Schüle

Re: Tim_Jim- What Am I overlooking Here!?

Post by Werner »

Hi Tim,
here the 64bit version does not start in console mode with this command:
octochess-r4558-64-ja.exe --hash 256
Werner
User avatar
geots
Posts: 4790
Joined: Sat Mar 11, 2006 12:42 am

Re: Tim_Jim- What Am I overlooking Here!?

Post by geots »

Werner wrote:
Codesquid wrote:
Werner wrote:
Codesquid wrote: Example for 256MB hash:

Code: Select all

octochess.exe --hash 256
...can you please give an example how this works in Arena??
I can't, I'm not familiar with Arena.
... and why does the engine sometimes in an endgame use more than 1CPU? I sometimes saw 100 % cpu usage on my quad here??
Octochess is multi-threaded. First root node is searched single-threaded, the other root nodes are searched in parallel. You can control the number of search threads with the --threads argument.
:cry: ok, this is the reason I have had such different results. I could not see the mp feature. This is normally a uci command too. So if you don´t implement it, we cannot test the engine. I am afraid I have to throw away my played games as I wanted to ply only single threaded games (1CPU - 1CPU). As you read before I too was not able to set the hash command...

best wishes
Werner

Werner, if you are saying I cannot run the engine on 1 CPU, that leaves me out also. I have 4 cores, bust most of my testing is 1CPU.


Best,

george