Toga II 1.4 Beta 5c is Out

[Henrik Dinesen]

A pity that I have only a Quad.

Don't worry, the difference is rather small

[Uri]

Is Toga II 1.4 Beta 5c an improvement over Toga II 3.1.2SE?

[Dann Corbit]

I hated seeing all of this confusion about the threads, so I just hacked up a quick UCI option for controlling the number. I can't compile it because I don't have Windows, but it should work. So anyone who has a compiler just let me know and I'll send the source to you. Then the exe can float around to everyone like normal.

Note that MaxThreads threads are started at the beginning, but they are all suspended. Then, at the start of each search, each one is woken up. So the task manager might say there are 16 threads running, but only the number in the UCI options should be consuming CPU time.

Using your version, I ran a single thread contest against Rybka 2.3.2a last night at 300+1 and got this result:

Code: Select all

    Program            Elo    +   -   Games   Score   Av.Op.  Draws
  1 Rybkav2.3.2a.w32 : 2938   51  49    84    60.7 %   2862   54.8 %
  2 TogaIISMP        : 2862   49  51    84    39.3 %   2938   54.8 %

[jarkkop]

First Nunn2 result

Toga 1.4 Beta5c-2cores - Rybka 2.3.2 32bit, one thread

11+ 11- 18 = , all even

[Golem]

Note that MaxThreads threads are started at the beginning, but they are all suspended. Then, at the start of each search, each one is woken up. So the task manager might say there are 16 threads running, but only the number in the UCI options should be consuming CPU time.

Wow, i didn't see this part when I first read your comment... I am a little blind sometimes... So, It seems that we have the same solution with this UCI option.

I think there is no easy way to delete the inactive threads ?

[Ovyron]

Is Toga II 1.4 Beta 5c an improvement over Toga II 3.1.2SE?

No, they were done independently. However, Toga II 1.4 Beta 5c is stronger than the other one.

Thanks to everyone that brought us the strongest freeware engine!

[Dann Corbit]

Note that MaxThreads threads are started at the beginning, but they are all suspended. Then, at the start of each search, each one is woken up. So the task manager might say there are 16 threads running, but only the number in the UCI options should be consuming CPU time.

Wow, i didn't see this part when I first read your comment... I am a little blind sometimes... So, It seems that we have the same solution with this UCI option.

I think there is no easy way to delete the inactive threads ?

It's easy to delete them. However, when they are not running, they do not really consume any resources. So the added complexity of removing them is a trade-off.

They are created here:

Code: Select all

search.cpp (157): threat_handle[i-1] = (HANDLE) _beginthreadex( NULL, 0, &search_thread, &ThreadIds[i-1], CREATE_SUSPENDED, &int
ernalThreadIds[i-1]);

This is how to get rid of them:
http://msdn2.microsoft.com/en-us/librar ... S.71).aspx

[Zach Wegner]

I hated seeing all of this confusion about the threads, so I just hacked up a quick UCI option for controlling the number. I can't compile it because I don't have Windows, but it should work. So anyone who has a compiler just let me know and I'll send the source to you. Then the exe can float around to everyone like normal.

Note that MaxThreads threads are started at the beginning, but they are all suspended. Then, at the start of each search, each one is woken up. So the task manager might say there are 16 threads running, but only the number in the UCI options should be consuming CPU time.

Using your version, I ran a single thread contest against Rybka 2.3.2a last night at 300+1 and got this result:

Code: Select all

    Program            Elo    +   -   Games   Score   Av.Op.  Draws
  1 Rybkav2.3.2a.w32 : 2938   51  49    84    60.7 %   2862   54.8 %
  2 TogaIISMP        : 2862   49  51    84    39.3 %   2938   54.8 %

Cool. Pretty close. Can you confirm that it uses the requested number of CPUs?

[George Tsavdaris]

The new Toga has problems with multivariation analysis mode in Fritz GUI.
I only get one line no matter what i do.

Is this normal? Anyone with the same problem?

[Eelco de Groot]

The new Toga has problems with multivariation analysis mode in Fritz GUI.
I only get one line no matter what i do.

Is this normal? Anyone with the same problem?

George Multi PV is not yet implemented, I believe Thomas wrote that somewhere in the read.me. Question was asked at CSS forum as well, it is probably not straightforward when there are more threads, in a single thread version Multi PV could be done just as in the regular older versions. The two thread version is not really much stronger yet so maybe a single CPU multi PV version is possible. But I'm sure Thomas will take a look if it is possible to do it, after all this is just a Beta, not a finished new 1.4 yet...

Eelco