Xiphos 0.6

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
mt
Posts: 29
Joined: Tue Feb 27, 2018 6:00 am
Contact:

Xiphos 0.6

Post by mt » Thu Nov 07, 2019 2:12 am

I have released Xiphos 0.6:
https://github.com/milostatarevic/xiphos/releases

I would like to thank all the testers and everyone interested in this project.

User avatar
Graham Banks
Posts: 33167
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: Xiphos 0.6

Post by Graham Banks » Thu Nov 07, 2019 2:40 am

Thanks Milos. Worth the wait! :)
My email addresses:
gbanksnz at gmail.com
gbanksnz at yahoo.co.nz

Dann Corbit
Posts: 10128
Joined: Wed Mar 08, 2006 7:57 pm
Location: Redmond, WA USA
Contact:

Re: Xiphos 0.6

Post by Dann Corbit » Thu Nov 07, 2019 3:31 am

I made a statically linked version. I only built for SSE and not BMI. But if you use my makefile, it will make a static version for you for whichever you build.

https://drive.google.com/open?id=1sApMf ... ukhPYhc2au
Taking ideas is not a vice, it is a virtue. We have another word for this. It is called learning.
But sharing ideas is an even greater virtue. We have another word for this. It is called teaching.

User avatar
Werner
Posts: 2419
Joined: Wed Mar 08, 2006 9:09 pm

Re: Xiphos 0.6

Post by Werner » Thu Nov 07, 2019 8:45 am

I started all 3 versions on 1CPU parallel and got 3 different results for depth, score, nps. And the bmi2 version is not the fastest ?
I use Intel I7 7700 Kaby Lake.

edit: I will try once more an found the reason:

with default hash 128 MB the engine starts with higher nps: 2215, 2120, 2112
I have had different hash for the 3 engines in my test.

sorry - something got wrong here - now on start position I have

FEN: rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1

Xiphos 0.6 x64 1CPU BMI:
22 00:20 41.231k 1.975k +0,25 e2-e4 c7-c5 c2-c3 d7-d5 e4xd5 Dd8xd5 d2-d4 Lc8-f5 Sg1-f3 e7-e6 Lc1-e3 Sb8-c6 c3-c4 Dd5-d7 Sb1-c3 c5xd4 Sf3xd4 Sc6xd4 Dd1xd4 Dd7xd4 Le3xd4 Sg8-f6 Lf1-e2 Ta8-d8 Ld4xa7

Xiphos 0.6 x64 ori SSE 1CPU:
22 00:20 39.415k 1.888k +0,25 e2-e4 c7-c5 c2-c3 d7-d5 e4xd5 Dd8xd5 d2-d4 Lc8-f5 Sg1-f3 e7-e6 Lc1-e3 Sb8-c6 c3-c4 Dd5-d7 Sb1-c3 c5xd4 Sf3xd4 Sc6xd4 Dd1xd4 Dd7xd4 Le3xd4 Sg8-f6 Lf1-e2 Ta8-d8 Ld4xa7

Xiphos 0.6 x64 Corbit 1CPU:
22 00:20 39.450k 1.889k +0,25 e2-e4 c7-c5 c2-c3 d7-d5 e4xd5 Dd8xd5 d2-d4 Lc8-f5 Sg1-f3 e7-e6 Lc1-e3 Sb8-c6 c3-c4 Dd5-d7 Sb1-c3 c5xd4 Sf3xd4 Sc6xd4 Dd1xd4 Dd7xd4 Le3xd4 Sg8-f6 Lf1-e2 Ta8-d8 Ld4xa7
Last edited by Werner on Thu Nov 07, 2019 9:13 am, edited 1 time in total.
Werner

User avatar
Guenther
Posts: 3057
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: Xiphos 0.6

Post by Guenther » Thu Nov 07, 2019 9:05 am

Werner wrote:
Thu Nov 07, 2019 8:45 am
I started all 3 versions on 1CPU parallel and got 3 different results for depth, score, nps. And the bmi2 version is not the fastest ?
I use Intel I7 7700 Kaby Lake.
Anyone can tell me the reason? As there are different evals on the same position - is it a compile bug?
And which version should I use now ?

Werner

edit: I will try once more
That's normal for Xiphos. It does some randomizing.
viewtopic.php?f=2&t=67820&p=766406&hili ... om#p766406

I get different nodes/moves/evals/time for each run.

Edit:
I see now after your editing you meant 3 different compiles.
Acually you will get different results always, even for the same compile due to randomizing,
that's what I wanted to tell!
Last edited by Guenther on Thu Nov 07, 2019 9:24 am, edited 2 times in total.
Current foe list count : [99]
http://rwbc-chess.de/chronology.htm

User avatar
Werner
Posts: 2419
Joined: Wed Mar 08, 2006 9:09 pm

Re: Xiphos 0.6

Post by Werner » Thu Nov 07, 2019 9:15 am

Thanks Guenther,
but here for my tests I have to think which hash I better use :?
Werner

User avatar
Guenther
Posts: 3057
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: Xiphos 0.6

Post by Guenther » Thu Nov 07, 2019 9:38 am

Stripped off output from Xiphos 0.6 npcnt with 'go depth 20' from the start position.
(ofc closed and restarted from cmd each time)
Run 1,2,5,6 compile from github.
Run 3,4,7,8 my compile.

Node count starts to differ already at depth 10.
In eight runs there were four different main lines with 7 different evals.

1.c4 (e5) +0.22
1.e4 (c5) +0.33
1.e4 (c5) +0.29
1.e4 (c5) +0.24
1.d4 (Nf6) +0.21
1.e4 (c5) +0.27
1.e4 (e6) +0.22
1.e4 (c5) +0.33

Code: Select all

Xiphos 0.6 NO-POPCNT by Milos Tatarevic
threads=1
hash=128MB
uci
id name Xiphos 0.6 NO-POPCNT
...
go depth 20
...
info depth 20 score cp 22 nodes 29079421 tbhits 0 time 35202 nps 826049 pv c2c4 e7e5 b1c3 g8f6 g1f3 b8c6 e2e4 d7d6 d2d3 c6d4 f3d4 e5d4 c3d5 f6d5 c4d5 f8e7 f1e2 e8g8 e1g1 h7h6 c1d2 c8d7 h2h3 f7f5 d1b3
bestmove c2c4 ponder e7e5

Xiphos 0.6 NO-POPCNT by Milos Tatarevic
threads=1
hash=128MB
uci
id name Xiphos 0.6 NO-POPCNT
...
go depth 20
info depth 1 score cp 46 nodes 20 tbhits 0 time 0 nps 20000 pv d2d4
...
info depth 20 score cp 33 nodes 14461260 tbhits 0 time 17662 nps 818731 pv e2e4 c7c5 c2c3 g8f6 e4e5 f6d5 d2d4 c5d4 g1f3 b8c6 c3d4 d7d6 f1c4 e7e6 e1g1 f8e7 b1c3 d5c3 b2c3 d6e5 d4e5 e8g8 d1e2 d8a5 f1d1
bestmove e2e4 ponder c7c5

Xiphos 0.6 NO-POPCNT by Milos Tatarevic
threads=1
hash=128MB
uci
id name Xiphos 0.6 NO-POPCNT
...
ucinewgame
go depth 20
info depth 1 score cp 46 nodes 20 tbhits 0 time 0 nps 20000 pv d2d4
...
info depth 20 score cp 29 nodes 13589876 tbhits 0 time 16959 nps 801289 pv e2e4 c7c5 c2c3 b8c6 d2d4 d7d5 e4d5 d8d5 g1f3 c8f5 f1e2 e7e6 e1g1 g8f6 c1e3 f8e7 b1d2 c5d4 f3d4 c6d4 e3d4 e8g8 f1e1 f8c8 e2f3 d5d7
bestmove e2e4 ponder c7c5

Xiphos 0.6 NO-POPCNT by Milos Tatarevic
threads=1
hash=128MB
uci
id name Xiphos 0.6 NO-POPCNT
...
ucinewgame
go depth 20
info depth 1 score cp 46 nodes 20 tbhits 0 time 1 nps 10000 pv d2d4
...
info depth 20 score cp 24 nodes 19679886 tbhits 0 time 24260 nps 811173 pv e2e4 c7c5 g1f3 e7e6 b1c3 b8c6 f1e2 e6e5 e1g1 g8f6 d2d3 f8e7 c3d5 e8g8 c2c4 d7d6 c1d2 h7h6 a2a3 a7a6 h2h3 b7b5 a1c1 b5c4 c1c4 f6d5
bestmove e2e4 ponder c7c5

Xiphos 0.6 NO-POPCNT by Milos Tatarevic
threads=1
hash=128MB
uci
id name Xiphos 0.6 NO-POPCNT
...
go depth 20
info depth 1 score cp 46 nodes 20 tbhits 0 time 0 nps 20000 pv d2d4
...
info depth 20 score cp 21 nodes 19721115 tbhits 0 time 24159 nps 816271 pv d2d4 g8f6 c2c4 e7e6 g1f3 d7d5 b1c3 f8b4 c1g5 c7c5 c4d5 e6d5 d4c5 e8g8 a1c1 c8e6 g5f6 d8f6 e2e3 b8c6 f1e2 f8d8 e1g1 b4c5
bestmove d2d4 ponder g8f6

Xiphos 0.6 NO-POPCNT by Milos Tatarevic
threads=1
hash=128MB
uci
id name Xiphos 0.6 NO-POPCNT
...
go depth 20
info depth 1 score cp 46 nodes 20 tbhits 0 time 1 nps 10000 pv d2d4
...
info depth 20 score cp 27 nodes 20657295 tbhits 0 time 25163 nps 820906 pv e2e4 c7c5 g1f3 e7e6 f1e2 b8c6 c2c4 d7d5 e4d5 e6d5 c4d5 d8d5 e1g1 g8f6 b1c3 d5d8 e2b5 f8e7 d2d4 e8g8 d4d5 c6b4 b5c4 b7b5 c4b5 f6d5 f1e1 c8e6 c3d5
bestmove e2e4 ponder c7c5

Xiphos 0.6 NO-POPCNT by Milos Tatarevic
threads=1
hash=128MB
uci
id name Xiphos 0.6 NO-POPCNT
...
go depth 20
info depth 1 score cp 46 nodes 20 tbhits 0 time 1 nps 10000 pv d2d4
...
info depth 20 score cp 22 nodes 29864183 tbhits 0 time 36744 nps 812741 pv e2e4 e7e6 d2d4 d7d5 e4d5 e6d5 g1f3 g8f6 f1d3 f8d6 d1e2 d6e7 e1g1 e8g8 c1e3 f8e8 f1e1 e7d6 b1d2 h7h6 h2h3 b8c6 c2c3 f6e4 d3c2 c8f5 d2e4 f5e4
bestmove e2e4 ponder e7e6

Xiphos 0.6 NO-POPCNT by Milos Tatarevic
threads=1
hash=128MB
uci
id name Xiphos 0.6 NO-POPCNT
...
go depth 20
info depth 1 score cp 46 nodes 20 tbhits 0 time 0 nps 20000 pv d2d4
...
info depth 20 score cp 33 nodes 12820228 tbhits 0 time 15902 nps 806151 pv e2e4 c7c5 c2c3 d7d5 e4d5 d8d5 d2d4 b8c6 g1f3 c8f5 f1e2 e7e6 e1g1 g8f6 c1e3 f8e7 b1d2 e8g8 e2c4 d5d7 d4c5 a8d8 b2b4 h7h6 h2h3
bestmove e2e4 ponder c7c5
Current foe list count : [99]
http://rwbc-chess.de/chronology.htm

tomitank
Posts: 201
Joined: Sat Mar 04, 2017 11:24 am
Location: Hungary

Re: Xiphos 0.6

Post by tomitank » Thu Nov 07, 2019 7:04 pm

Nice progress!

Try this, maybe give you some elo:
https://github.com/official-stockfish/S ... h.cpp#L782

Post Reply