search

Matthies / RubiChess

Another chess engine
GNU General Public License v3.0
150 stars 16 forks source link

C++20 #456

Open Matthies opened 4 months ago

Matthies commented 4 months ago

Non-regression test: Elo | 1.99 +- 3.56 (95%) SPRT | 10.0+0.10s Threads=1 Hash=8MB LLR | 2.96 (-2.94, 2.94) [-3.00, 1.00] Games | N: 17838 W: 4402 L: 4300 D: 9136 Penta | [53, 1922, 4879, 2000, 65]

Matthies commented 4 months ago

There are several (almost all) workers at OpenBench using gcc 8 or 9 which don't understand -std=c++20 and suggest using -std=c++2a. Not working in OB is a 'no go' merging this branch but I will give c++2a a try.

Matthies commented 1 month ago

Switching to c++2a for workers with older gcc9 works technically but the performance of the code of this old compiler seems very bad looking at a fixed 1000 games test:

Elo   | -16.66 +- 10.95 (95%)
Conf  | 10.0+0.10s Threads=1 Hash=8MB
Games | N: 1002 W: 231 L: 279 D: 492
Penta | [11, 121, 275, 93, 1]
http://chess.grantnet.us/test/35899/

So I will stay at C++11 for now.