TalkChess.com

Hi Daniel, I can only show you the results between 1CPU and 8Threads (4CPU): Ver 0.94 1CPU 512 MB hash T30 Startposition 142 sec +0.22 e4 1318nps after Nh3 with black 204 sec -0.43 d5 1302 nps Ver 0.94 8Thr 2048 MB hash T30 Startposition 114 sec +0.32 e4 6812 nps after Nh3 with black 120 sec -0.44 ...

arunsoorya1309 wrote: ↑

Mon Dec 10, 2018 3:35 am

we are testing in ccc to move to Linux, I’ll try to get it in

Regards
Arun

Thanks! Please, tell me something a little before starting the tournament, as I will send you my latest version.

The linux version was Ubuntu. I added to the same file a Centos 7 version. Otherwise it said GLIBC... not found.

Hello! Maybe someone can test if this test version of Andscacs scales well in nps with more than 64 threads in Windows? Also included is a Linux version that should work well, using events instead of sleep for the waiting threads (Windows version already had this for quite some time), and maybe some...

Hello Matthew, nice to see you back after so long !! Does this mean Giraffe will get some updates in the future ? :D Thanks! Afraid not! AlphaZero is taking up all my time these days, and it's a very exciting project with lots of uncharted territory ahead :). AlphaZero is basically what I ever want...

chrisw wrote: ↑

Sat Dec 01, 2018 8:42 pm

...LC0 evaluation is optimistic usually, because, er, dunno. Not really thought about it.

I guess it's the result of that the positions it reaches tend to work against himself in selfplay.

Many thanks to all! I will review all this.

Hello.
To use threads the same way in Windows and in Linux, I will use pevents:
https://github.com/neosmart/pevents
that emulates events the Windows way in Linux.
Someone has used it? It works well?
Many thanks.

Hello.
Many engines use prefetch, but how about prefetchw?

https://www.felixcloutier.com/x86/PREFETCHW.html

Maybe can help to win some speed also.

In Andscacs there is only one iterative deepening loop in the main thread, which signals all the threads to start searching. About depth of the threads: for (int i = 1; i < NumThreads; i++) { incrementthread[i] = i & 1; } if (NumThreads <= 4) { incrementthread[0] = 0; incrementthread[1] = 1; increme...