Threats started mostly as a part of search and history, but also have speed optimizations in move generation. By taking the threatened bit board - castles and king moves can be generated even faster. This showed an approximate 1% speedup locally.
Note the OB test should've been a regression, since I knew it was faster, but the bench change needed some validation. SMP was to sanity check the pseudo legal code.
Bench: 4309182
Threats started mostly as a part of search and history, but also have speed optimizations in move generation. By taking the threatened bit board - castles and king moves can be generated even faster. This showed an approximate 1% speedup locally.
ELO | 1.55 +- 1.67 (95%) SPRT | 4.0+0.04s Threads=1 Hash=8MB LLR | 1.34 (-2.94, 2.94) [0.00, 3.00] GAMES | N: 81728 W: 20207 L: 19843 D: 41678 http://chess.grantnet.us/test/33617/
ELO | 5.91 +- 14.33 (95%) CONF | 4.0+0.04s Threads=8 Hash=64MB GAMES | N: 1000 W: 230 L: 213 D: 557 http://chess.grantnet.us/test/33632/
Note the OB test should've been a regression, since I knew it was faster, but the bench change needed some validation. SMP was to sanity check the pseudo legal code.