[gnugo-devel] benchmarks

[Arend Bayer]

Here is finally some benchmarking for the break-in code. Overall, it
costs 20%, which is too much. Especially +43% in the Go4++ game is
unacceptable, of course.

For comparison, I also benchmarked 3.2. Overall, we are a lot faster
than 3.2, but the variance is extremely high.
So I think we can spend a little performance on features. This could be
increasing some depth levels, if we knew which ones are most useful.

Arend


File:
3.3.21, no breakin      3.3.21 with breakin     3.2

gnugo-3.3.15-Kwaku-200301281930.gtp.in:
1086                    1144                    1065
                        +5.3%

gnugo-3.3.15-morli-200301291908.gtp.in:
1057                    1214                    2199
                        +7.7%

gnugo-3.3.15-wingjk-200301260551.gtp.in:
1104                    1439                    1901
                        +30.3%

gnugo-go4.gtp.in:
1544                    2221                    1529
                        +43.8%

kit-gnugo-3.3.15-200301282141.gtp.in:
1114                    1425                    1107
                        +27.9%

leftd-gnugo-3.3.15-200301260333.gtp.in:
1683                    1748                    2678
                        +3.8%

manyfaces-gnugo.gtp.in:
 999                    1148                    1622
                        +14.9%

Total:
8587                    10342                   12101
                        +20.4%