Joost VandeVondele

Adjudication is something I don't really like, but certainly when tested in the past, it turned out to be really accurate. It tends to speedup testing by 20% (again, based...

so this PR is fine with me.

Is this issue specific to FRC or is this also affecting non-variant (standard) chess? I tried to understand the patch but wasn't sure.

I'll move some data here from the fishtest issue where we started discussing: On a machine with 128cores/256threads I look at timeout stats as a function of cutechess concurrency (threads...

If I remove output of the pgn, and just looking at the short TC games, it appears to improve already a bit: ``` threads TC timed-out / total 240 4.0+0.04...

@alwey suggested that this could be related to parsing PVs. I couldn't test an earlier cutechess version, but I realized that I could test the same engine (stockfish) but with...

This particular test was on a dual-socket EPYC 7742, CentOS Linux 7. However, machines running on fishtest are other kinds as well. E.g. look at the bottom of this page...

@lucasart yes I've done that a couple of days ago. I have 10% time losses with normal IO and about 5% with limited IO. (TC 4+0.04, concurrency 250).

I think I ran c-chess-cli without the pgn option: ``` ./c-chess-cli -each tc=4+0.04 option.Hash=8 option.Threads=1 -engine cmd=./stockfish.master -engine cmd=./stockfish.master -games 2000 -concurrency 250 -openings noob_3moves.epd -random -repeat -resign 3,700 -draw...

BTW, just to make sure, when I run 25 c-chess-cli process in parallel, each of concurrency 10, there is no single loss in time. ``` for i in `seq 1...