Re: [gnugo-devel] Semeai node limits

[Gunnar Farnebäck]

Dan wrote:
> Perhaps the semeai nodes should be increased but 2000 is
> too much. The current default is 500.

These are the results with semeai node limit 1000:
strategy:50     PASS Q9 [Q9]
trevord:680     FAIL R15 [S16]
nngs2:460       FAIL N15 [J4|K2|K3|K4|K5|L2]
nngs2:600       PASS B5 [B5]
9x9:490         PASS A6 [D1|A6|H1]
kgs:730         PASS R16 [R16]
4 PASS
2 FAIL
Total nodes: 1741015977 3712509 12930781 (+3.4% +12% +2.6%)

and with 1500:
strategy:50     PASS Q9 [Q9]
lazarus:7       PASS T5 [T5]
nicklas1:901    FAIL D3 [H8]
nngs:302        PASS M7 [F15|M7|B14]
nngs:510        FAIL P13 [G14]
nngs:1090       FAIL A11 [B17|B16]
strategy3:128   FAIL T11 [O8]
semeai:28       FAIL 0 0 PASS [1 1 S9]
trevord:680     FAIL R15 [S16]
nngs2:460       FAIL N15 [J4|K2|K3|K4|K5|L2]
nngs2:600       PASS B5 [B5]
nngs4:1120      PASS N7 [N7]
ninestones:570  FAIL H1 [A9|B10|D11|F11|J15|A16]
9x9:120         PASS F1 [F1]
9x9:490         PASS A6 [D1|A6|H1]
kgs:730         PASS R16 [R16]
tiny:403        FAIL D4 [B3|C2]
tiny:405        FAIL PASS [B1|B4|C1|C4]
8 PASS
10 FAIL
Total nodes: 1790551682 4047943 13247988 (+6.4% +22% +5.1%)

With 1000 nodes I measure a regression slowdown by 9% and with 1500 a
slowdown by 17%.

/Gunnar