Re: [gnugo-devel] benchmarks

[Gunnar Farneback]

Inge wrote:
> Performace optimizations is of course good.  But if I have to chose, I
> would not want to exclude such nice features just because it uses some
> time.   Besides, when analyzing a game, is persistent caching used to
> its full potential?

Absolutely. The setup is designed to mimic an actual game as closely
as possible, except that the tested version may want to play different
moves. The GTP stream for the original game would go something like

play black R16    # Play a move for the opponent.
genmove white     # Generate and play a move.
play black D17
genmove white
...

while the benchmarking GTP stream goes like

play black R16    # Play a move for the opponent.
reg_genmove white # Generate but don't play a move.
play white D4     # Play the game move.
play black D17
reg_genmove white
play white Q4
...

/Gunnar