Hi Antonio!Antonio Torrecillas wrote: Can you please report how it scale with more cpus?
I don't want a complete study, say from a concrete position
how many nodes or expands are explored in 10 seconds with 1,2,..,n cpus.
Here is Shredder-GUI's output with Rocinante counting 1 minute each trial at 1, 2, 4, 8 and 12 cores from starting position.
Pity I didn't change number of threads too each time, so it remained at 12 always, I can easily repeat the test with changing that, but wanted to show what I've got anyhow first, to let you know, if it's what you meant at all.
Engine: Rocinante 2.0 JAPD3 (0 MB)
von Antonio Torrecillas
3/03 0:02 +0.11 1.Sc3 Sc6 2.d4 (25) 0
4/04 0:02 +0.19 1.d4 d5 2.Sc3 Sc6 (31) 0
4/04 0:02 +0.19 1.Sf3 d5 2.d4 Sc6 (35) 0
3/03 0:03 +0.13 1.d4 Sf6 2.Sc3 (36) 0
4/04 0:03 +0.17 1.e4 d5 2.e5 d4 (43) 0
4/04 0:04 +0.19 1.d4 d5 2.Sc3 Sc6 (46) 0
4/04 0:07 +0.19 1.Sf3 Sc6 2.Sc3 d5 (78) 0
4/04 0:08 +0.19 1.Sc3 Sc6 2.d4 d5 (87) 0
4/04 0:08 +0.19 1.Sf3 Sc6 2.Sc3 d5 (90) 0
4/04 0:16 +0.19 1.d4 d5 2.Sc3 Sc6 (159) 0
5/05 0:24 +0.18 1.e4 d5 2.e5 Sc6 3.d4 (229) 0
7/07 1:00 +0.24 1.e4 d5 2.exd5 Dxd5 3.Sc3 De6+ 4.Sge2 (464) 0
Bester Zug: e2-e4 Zeit: 1:00.388 min K/s: 7 Knoten: 464
4/04 0:01 +0.19 1.Sf3 d5 2.d4 Sc6 (35) 0
3/03 0:01 +0.13 1.d4 Sf6 2.Sc3 (36) 0
4/04 0:02 +0.17 1.e4 d5 2.e5 d4 (43) 0
4/04 0:02 +0.19 1.d4 d5 2.Sc3 Sc6 (46) 0
4/04 0:04 +0.19 1.Sf3 Sc6 2.Sc3 d5 (78) 0
4/04 0:04 +0.19 1.Sc3 Sc6 2.d4 d5 (87) 0
4/04 0:05 +0.19 1.Sf3 Sc6 2.Sc3 d5 (90) 0
4/04 0:09 +0.19 1.d4 d5 2.Sc3 Sc6 (159) 0
5/05 0:14 +0.18 1.e4 d5 2.e5 Sc6 3.d4 (229) 0
7/07 0:59 +0.21 1.e4 e5 2.Sf3 d5 3.Sxe5 dxe4 4.Lc4 (644) 0
Bester Zug: e2-e4 Zeit: 0:59.998 min K/s: 10 Knoten: 644
4/04 0:02 +0.19 1.Sf3 Sc6 2.Sc3 d5 (78) 0
4/04 0:02 +0.19 1.Sc3 Sc6 2.d4 d5 (87) 0
4/04 0:02 +0.19 1.Sf3 Sc6 2.Sc3 d5 (90) 0
4/04 0:05 +0.19 1.d4 d5 2.Sc3 Sc6 (159) 0
5/05 0:08 +0.18 1.e4 d5 2.e5 Sc6 3.d4 (229) 0
6/06 0:59 +0.17 1.e4 e5 2.Sf3 Sc6 3.Sc3 Lc5 (1.119) 0
Bester Zug: e2-e4 Zeit: 0:59.998 min K/s: 18 Knoten: 1.119
4/04 0:01 +0.19 1.Sc3 Sc6 2.d4 d5 (87) 0
4/04 0:01 +0.19 1.Sf3 Sc6 2.Sc3 d5 (90) 0
4/04 0:03 +0.19 1.d4 d5 2.Sc3 Sc6 (159) 0
5/05 0:04 +0.18 1.e4 d5 2.e5 Sc6 3.d4 (229) 0
8/08 1:00 +0.23 1.e4 e5 2.d4 exd4 3.Dxd4 Sc6 4.Dc4 Sf6 (1.590) 0
Bester Zug: e2-e4 Zeit: 1:00.030 min K/s: 26 Knoten: 1.590
4/04 0:02 +0.19 1.d4 d5 2.Sc3 Sc6 (159) 0
5/05 0:03 +0.18 1.e4 d5 2.e5 Sc6 3.d4 (229) 0
8/08 1:00 +0.21 1.e4 e5 2.Sc3 Sc6 3.Sf3 Sf6 4.d4 exd4 (1.849) 0
Bester Zug: e2-e4 Zeit: 0:59.997 min K/s: 30 Knoten: 1.849
Each line beginning with "Bester Zug..." shows change of number of cores.
Rest of settings was threads 12, klevel 1, mcts_ab true and probedepth 3.
Let me know, if it helps or not