TalkChess.com

The weekend of oct 12/13 is the GSEI tournament in Turin. Although I'm not participating there this year it would be nice if the events do not happen on the same day. As far as I can see now nov30/dec1 is really bad for me. I'd prefer if nov 2/3, nov 23/24, dec 14/15 could be avoided. I will not be ...

Yes, these are the next steps I was planning to implement. I have added Qsearch() and it is acting much better now. In the initial position searched to ply 8 and ply 9 it returns 0 as expected. However, there is still a problem. In the following position. r1bqkbnr/1ppp1ppp/p1n5/1B2p3/4P3/5N2/PPPP1P...

Sven wrote: ↑

Sat Mar 02, 2019 5:13 pm

Ok. Is there a common estimate of the slowdown for a non-NUMA-aware engine on such a machine?

Torsten once tested that for me in one of the early TCEC editions. I don't rmember exactly how much, but it was significant (order of magnitude 10-20%).

Sven wrote: ↑

Sat Mar 02, 2019 11:40 am

Is NUMA awareness currently required at TCEC?

Required not. Recommended yes, as it is a dual-socket machine.

The Baron does not use neural nets (yet). The version registered to TCEC is a NUMA aware compile of 3.44
Richard.

Jesse Gersenson wrote: ↑

Mon Jan 28, 2019 12:46 am

I'm not sure how many thread the Baron supports -- wine is "limited" to 64.

I have a hardcoded limit of 32 threads in the Baron. If anyone wants to test on bigger machines: Drop me a message. Most likely you'll need a NUMA aware compile as well (the distributed version is not).

I'll take a look at it. I'm not aware of making changes that could break chess960 support, but apparently something is not working as expected. I must confess I haven't used it myself in a while. Richard. I found the problem and Ray has in the meantime checked it as well. I sent a new package to St...

I did not really plan one. I'm using the Baron with wine in linux (both 32 and 64 bits) myself without any problems, so the need for a native compile seems limited to me.

Hi Werner, Not sure what your question is. The commented out values are the defaults, but you can experiment with them. If you find other values to be clearly stronger I'd appreciate the feedback. I see you enabled Syzygy in qsearch, which can be really heavy, especially on multicore machines, but f...

I'll take a look at it. I'm not aware of making changes that could break chess960 support, but apparently something is not working as expected. I must confess I haven't used it myself in a while.
Richard.