Portfish - C# port of Stockfish

[bpfliegel]

Dear all,

let me present you Portfish, the .Net port of Stockfish, currently as a first beta.

Github URL and readme:
https://github.com/bpfliegel/Portfish

Latest desktop releases:
https://github.com/downloads/bpfliegel/ ... elease.zip

I would be happy to receive feedback on engine performance and strength (especially on i7 and 8+ core Xeons) and compiling it on Mono.
The fastest/strongest version is 4.5 (which is currently a beta framework) and 4.0, x64.

Thanks a lot, cheers,
Balint

[gladius]

Wow, that must have been a huge amount of work! Very impressive .

[lucasart]

Dear all,

let me present you Portfish, the .Net port of Stockfish, currently as a first beta.

Github URL and readme:
https://github.com/bpfliegel/Portfish

Latest desktop releases:
https://github.com/downloads/bpfliegel/ ... elease.zip

I would be happy to receive feedback on engine performance and strength (especially on i7 and 8+ core Xeons) and compiling it on Mono.
The fastest/strongest version is 4.5 (which is currently a beta framework) and 4.0, x64.

Thanks a lot, cheers,
Balint

Did you verify the node count with a ./stockfish bench ? This is essentialt, if node counts are equal most likely all is good. If not then there is a bug somewhere, as it is not functionally equivalent to the original SF.

[pocopito]

x86 port40 seems to work with mono under ubuntu 12.04. It uses my four cores and makes arround 500 kN/s

~/PortfishLatestRelease/x86$ mono PortfishNet40.exe

Just in case it can be of any interest, this is the output of a few seconds analyzing the initial position.

Code: Select all

1 [+0.72]  1.Nf3  (0.07)
 2 [+0.12]  1.Nf3 Nf6  (0.07)
 3 [+0.68]  1.Nf3 Nf6 2.Nc3  (0.07)
 4 [+0.12]  1.Nf3 Nf6 2.Nc3 Nc6  (1.50)
 5 [+0.56]  1.Nf3 Nf6 2.Nc3 d5 3.d4  (1.53)
 6 [+0.12]  1.Nf3 Nf6 2.Nc3 d5 3.d4 Nc6  (1.67)
 7 [+0.44]  1.e4 d5 2.exd5 Nf6 3.d4 Nxd5 4.Nf3  (1.82)
 8 [+0.36]  1.e4 d5 2.exd5 Nf6 3.d4 Nxd5 4.Nf3 Nc6 5.Nc3 e6  (2.02)
 9 [+0.44]  1.e4 d5 2.exd5 Nf6 3.d4 Nxd5 4.Nf3 Nc6 5.Nc3 e6 6.Bb5 Bb4  (2.11)
 9 [+0.52]  1.e4 d5 2.exd5 Nf6 3.d4 Nxd5 4.Nf3 Nc6 5.Nc3 e6 6.Bg5  (2.12)
 9 [+0.28]  1.e4 e6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Nc6 5.Nf3 Nf6  (2.20)
 9 [+0.40]  1.e4 e6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Nc6 5.Nf3 Nf6 6.Bg5  (2.36)
11 [+0.48]  1.e4 e6 2.Nc3 Nc6 3.d4 d5 4.e5 f6 5.Bb5 fxe5 6.Bxc6+  (2.44)
11 [+0.48]  1.e4 e6 2.Nc3 Nc6 3.d4 d5 4.e5 f6 5.Bb5 Ne7 6.Nf3 a6 7.Ba4  (2.51)
12 [+0.44]  1.e4 e5 2.Nf3 Nf6 3.Nc3 Nc6 4.d4 exd4 5.Nxd4 Bb4 6.Nxc6 Bxc3+ 7.bxc3 bxc6 8.e5  (3.03)
14 [+0.36]  1.e4 e5 2.Nf3 Nf6 3.Nc3 Nc6 4.d4 exd4 5.Nxd4 Bb4 6.Nxc6 Bxc3+ 7.bxc3 bxc6 8.e5 Qe7  (3.72)
14 [+0.24]  1.e4 e5 2.Nf3 Nc6 3.d4 exd4 4.Nxd4 Nf6 5.Nxc6 bxc6 6.Nc3 Bd6 7.Bd3 O-O 8.O-O  (4.33)
15 [+0.32]  1.e4 e5 2.Nf3 Nc6 3.Bb5 f5 4.d3 fxe4 5.dxe4  (4.86)
15 [+0.40]  1.e4 e5 2.Nf3 Nc6 3.Bb5 Nf6 4.d3 Bd6 5.Nbd2 a6 6.Bxc6 dxc6 7.Nc4 Qe7 8.Bd2  (5.62)
16 [+0.32]  1.e4 e5 2.Nf3 Nc6 3.Bb5 Nf6 4.d3 Bd6 5.Nbd2 a6 6.Bxc6 dxc6 7.Nc4 Bg4 8.Ncxe5 Bxe5  (6.12)
16 [+0.40]  1.e4 e5 2.Nf3 Nf6 3.Nxe5 d6 4.Nf3 Nxe4 5.Nc3 Nxc3 6.dxc3 Be7 7.Bc4 Nc6 8.Be3 Be6 9.Bxe6 fxe6 10.O-O  (7.27)
17 [+0.32]  1.e4 e5 2.Nf3 Nf6 3.Nxe5 d6 4.Nf3 Nxe4 5.Nc3 Nxc3 6.dxc3 Be7 7.Bc4 Nc6 8.Be3 Be6 9.Bxe6 fxe6 10.O-O O-O  (8.86)
17 [+0.24]  1.e4 e6 2.d4 d5 3.exd5 exd5 4.Nf3 Nf6 5.Bd3 Bb4+ 6.c3 Be7 7.Na3 O-O 8.O-O Bxa3 9.bxa3 Ne4  (11.65)
17 [+0.28]  1.e4 e6 2.d4 d5 3.exd5 exd5 4.Nf3 Nf6 5.Bb5+ c6 6.Bd3 Bb4+ 7.c3 Qe7+ 8.Ne5 Bd6 9.O-O Bxe5 10.dxe5 Ng4 11.f4 Qc5+ 12.Kh1 O-O  (12.67)
18 [+0.20]  1.e4 e6 2.d4 d5 3.exd5 exd5 4.Nf3 Nf6 5.Bb5+ c6 6.Bd3 Bb4+ 7.c3 Qe7+ 8.Ne5 Bd6 9.O-O Nbd7 10.Nxd7 Bxd7 11.Re1 Be6  (16.01)
18 [+0.36]  1.e4 e6 2.d4 d5 3.exd5 exd5 4.Nf3 Nf6 5.Bb5+ c6 6.Bd3 Bb4+ 7.Bd2 Qe7+ 8.Qe2 Qxe2+ 9.Kxe2 Bd6 10.Re1 Be6 11.Kf1 O-O 12.Kg1  (18.57)
18 [+0.20]  1.e4 e6 2.d4 d5 3.e5 c5 4.Nf3 Nc6 5.c3 Bd7 6.Bd3 cxd4 7.cxd4 Qb6 8.Bc2 f6 9.O-O fxe5 10.dxe5  (19.45)

EDIT: just to add the results from the x64 in the same conditions:

Code: Select all

1 [+0.72]  1.Nf3  (0.07)
 2 [+0.12]  1.Nf3 Nf6  (0.08)
 3 [+0.68]  1.Nf3 Nf6 2.Nc3  (0.08)
 4 [+0.12]  1.Nf3 Nf6 2.Nc3 Nc6  (1.52)
 5 [+0.56]  1.Nf3 Nf6 2.Nc3 d5 3.d4  (1.63)
 6 [+0.12]  1.Nf3 Nf6 2.Nc3 d5 3.d4 Nc6  (1.70)
 7 [+0.40]  1.e4 Nc6 2.Nf3 d5 3.exd5 Qxd5 4.Nc3  (2.08)
 7 [+0.44]  1.e4 d5 2.exd5 Nf6 3.d4 Nxd5 4.Nf3  (2.09)
 8 [+0.36]  1.e4 Nc6 2.Nf3 f5 3.Nc3 fxe4 4.Nxe4 Nf6  (2.22)
 8 [+0.52]  1.e4 Nc6 2.d4 d5 3.exd5 Qxd5 4.Nf3 e6 5.Nc3  (2.25)
 8 [+0.36]  1.e4 d5 2.exd5 Nf6 3.d4 Nxd5 4.Nf3 Nc6 5.Nc3 e6 6.Bg5  (2.27)
 9 [+0.40]  1.e4 e6 2.Nc3 d5 3.d4 Nf6 4.e5 Ne4 5.Nxe4 dxe4  (2.33)
10 [+0.48]  1.e4 e6 2.Nc3 d5 3.Nf3 dxe4 4.Nxe4 f5 5.Nc3 Nc6 6.d4  (2.49)
10 [+0.36]  1.e4 d5 2.exd5 c6 3.dxc6 Nxc6 4.Nc3 Nf6 5.Nf3 e5  (2.55)
11 [+0.44]  1.e4 d5 2.exd5 c6 3.dxc6 Nxc6 4.Nc3 Nf6 5.Nf3 e5 6.d3  (2.74)
11 [+0.52]  1.e4 d5 2.exd5 c6 3.dxc6 Nxc6 4.Nc3 Nf6 5.Nf3 e5 6.Bd3  (2.90)
11 [+0.36]  1.e4 e6 2.Nc3 d5 3.Nf3 Nc6 4.Bb5 Nf6 5.e5 Ne4 6.O-O  (3.35)
12 [+0.28]  1.e4 e6 2.Nc3 d5 3.Nf3 Nc6 4.Bb5 Nf6 5.e5 Ne4 6.O-O a6  (3.56)
12 [+0.44]  1.e4 e6 2.Nf3 d5 3.e5 f6 4.d4 Nc6 5.Bb5 Bd7 6.Nc3 fxe5 7.Bxc6 bxc6 8.Nxe5  (3.70)
12 [+0.40]  1.e4 e6 2.Nf3 d5 3.e5 f6 4.d4 fxe5 5.Nxe5 Nf6 6.Be2 Nc6 7.O-O Nxe5 8.dxe5  (3.85)
13 [+0.40]  1.e4 e6 2.Nc3 d5 3.Nf3 dxe4 4.Nxe4 Nf6 5.Bd3 Nxe4 6.Bxe4 Nc6 7.O-O  (4.22)
14 [+0.32]  1.e4 e6 2.Nc3 d5 3.Nf3 dxe4 4.Nxe4 Nf6 5.Bd3 Nc6 6.O-O Nb4  (4.43)
14 [+0.32]  1.e4 e6 2.Nf3 d5 3.e5 f6 4.d4 fxe5 5.Nxe5 Nd7 6.Be3 Nxe5 7.dxe5 Nh6 8.Bxh6 gxh6 9.Qh5+ Kd7  (5.22)
15 [+0.40]  1.e4 e6 2.Nf3 d5 3.e5 f6 4.d4 fxe5 5.Nxe5 Nd7 6.Qh5+ g6 7.Nxg6 Bb4+ 8.Nc3 Ngf6 9.Qh6  (5.79)
15 [+0.48]  1.e4 e6 2.Nc3 d5 3.Nf3 dxe4 4.Nxe4 Nf6 5.Nxf6+ gxf6 6.d4 Nc6 7.Be2 e5 8.d5  (6.34)
15 [+0.40]  1.e4 e6 2.Nc3 d5 3.Nf3 dxe4 4.Nxe4 Nf6 5.Nxf6+ gxf6 6.Bb5+ Bd7 7.Bxd7+ Qxd7 8.d4 Nc6 9.c3 Rg8 10.O-O  (6.79)
16 [+0.32]  1.e4 e6 2.Nc3 d5 3.Nf3 Nf6 4.e5 Ne4 5.Bd3 f5 6.O-O Nc6  (7.94)
16 [+0.32]  1.e4 e6 2.Nf3 d5 3.exd5 exd5 4.d4 Nf6 5.Bd3 Bb4+ 6.c3 Qe7+ 7.Be3 Bd6 8.O-O O-O 9.Re1 Nc6  (10.68)
17 [+0.24]  1.e4 e5 2.Nf3 Nc6 3.Bb5 f5 4.d3 fxe4 5.dxe4 Nf6 6.Bxc6 bxc6 7.Nxe5 Bb4+ 8.c3 Bd6  (12.57)
17 [+0.16]  1.e4 e5 2.Nf3 Nc6 3.Bb5 f5 4.d4 exd4 5.e5 Nge7 6.O-O a6 7.Bxc6 Nxc6 8.Nxd4 Nxe5 9.Nxf5  (15.39)
18 [+0.24]  1.e4 e5 2.Nf3 Nc6 3.Bb5 f5 4.d4 exd4 5.e5 Nge7 6.O-O a6 7.Bxc6 Nxc6 8.Nxd4 Nxe5 9.Re1  (16.78)
18 [+0.24]  1.e4 e5 2.Nf3 Nc6 3.Bb5 f5 4.d4 exd4 5.e5 Nge7 6.O-O a6 7.Bxc6 Nxc6 8.Nxd4 Nxe5 9.f4 Ng6 10.Nxf5 Bc5+ 11.Kh1  (19.19)
19 [+0.32]  1.e4 e5 2.Nf3 Nc6 3.Bb5 f5 4.d4 exd4 5.e5 Nge7 6.O-O a6 7.Bxc6 Nxc6 8.Nxd4 Nxe5 9.f4 Ng6 10.Re1+ Ne7 11.Nc3  (20.86)
19 [+0.20]  1.e4 e6 2.Nf3 d5 3.exd5 exd5 4.d4 Nf6 5.Bd3 Bd6 6.Nc3 O-O 7.O-O Nc6 8.Re1 a6 9.Ne5 Nxd4 10.Bxh7+ Nxh7 11.Qxd4  (26.86)
20 [+0.28]  1.e4 e6 2.Nf3 d5 3.exd5 exd5 4.d4 Nf6 5.Bd3 Bd6 6.Nc3 Be6 7.O-O O-O 8.Re1 Nc6 9.Nb5 Nb4 10.Bg5 Nxd3 11.Qxd3  (29.11)
20 [+0.36]  1.e4 e6 2.d4 d5 3.Nc3 Nf6 4.e5 Ne4 5.Nxe4 dxe4 6.Be3 Nc6 7.c3 Qh4 8.g3  (35.43)
20 [+0.16]  1.e4 e6 2.d4 d5 3.exd5 exd5 4.Nf3 Nf6 5.Bd3 Bd6 6.Nc3 Be6 7.O-O O-O 8.Re1 Nc6 9.Nb5 Nb4 10.Nxd6 Qxd6 11.Ne5 Rae8 12.Bf4 Nxd3 13.Nxd3 Qb6  (40.97)
21 [+0.24]  1.e4 e6 2.d4 d5 3.exd5 exd5 4.Nf3 Nf6 5.Bb5+ c6 6.Bd3 Qe7+ 7.Be2 Qc7 8.c4 Bb4+ 9.Nc3 Ne4 10.Qb3 Bxc3+ 11.bxc3 O-O 12.O-O  (46.88)

[mcostalba]

Wow, that must have been a huge amount of work! Very impressive .

Yes, it has been.

I have been in contact with Balint during these months and I know the effort has been very big. He not only did a perfect porting so that Portish it is 100% functional equivalent to current development version of Stockfish, but he also tried hard to optimize it for .Net platform. I am not talking of micro-optimization, but of big reworks as memory allocation.

I really would like to congratulate with him also publicly (privately I had already done): Balint, you really made something awesome.

Thanks
Marco

[bpfliegel]

Marco, thanks a lot for your kind words!

When reaching up to the Apr 17 version the port was tested with the Arasan 14 (175 positions) against the original version - this allowed to find one issue earlier that the standard benchmark did not enable (just simply not covering that btw very specific case). It should be alright functionally, I will recheck it on the weekend again to be certain.

It was some effort, 4 months overall.

Emilio, thanks for the feedback on Mono, looks good, somewhat slow. What type of CPUs do you use? I get 900-1000 kps on my laptop i5 in 4 CPU mode for the 4.0/4.5 x64.

Cheers, Balint

[pocopito]

Marco, thanks a lot for your kind words!
Emilio, thanks for the feedback on Mono, looks good, somewhat slow. What type of CPUs do you use? I get 900-1000 kps on my laptop i5 in 4 CPU mode for the 4.0/4.5 x64.
Cheers, Balint

My laptop is an i5 4 cores. I've (re)tested 4.0 and 4.5 on both x86 and x64 (2.0 doesn't seem to work on none of them), with similar results: after the first plies the nps is around 500.000.
It's a bit odd this difference between both computers. I've double checked the 4 threads options is selected, and my cpu graph shows all the 4 cores are being used during the analysis.

I've also tested in "normal" positions (not only the starting one) and the results are quite similar.

If you think of any test I can do just ask for it.

BTW, thanks for the effort and congratulations for the results. I'm none to say anything like this, but even for a mediocre programmer like me, it's evident that it must have been a hell of a work.

Regards

E Diaz

[bpfliegel]

Marco, thanks a lot for your kind words!
Emilio, thanks for the feedback on Mono, looks good, somewhat slow. What type of CPUs do you use? I get 900-1000 kps on my laptop i5 in 4 CPU mode for the 4.0/4.5 x64.
Cheers, Balint

My laptop is an i5 4 cores. I've (re)tested 4.0 and 4.5 on both x86 and x64 (2.0 doesn't seem to work on none of them), with similar results: after the first plies the nps is around 500.000.
It's a bit odd this difference between both computers. I've double checked the 4 threads options is selected, and my cpu graph shows all the 4 cores are being used during the analysis.

I've also tested in "normal" positions (not only the starting one) and the results are quite similar.

If you think of any test I can do just ask for it.

BTW, thanks for the effort and congratulations for the results. I'm none to say anything like this, but even for a mediocre programmer like me, it's evident that it must have been a hell of a work.

Regards

E Diaz

It needs some time to warm up as usual - in case of .Net this is more true than by C engines, so nps is maxed out usually only around depths 22-23. Do you have a log on what nps is reached at specific depths?

Also wonder why the 2.0 version does not work, what error did you get (if any)?

I'm looking for objective tests on playing strength - as I've seen there are a lot of people testing engine versions here on CCC.

Thanks a lot for your kind help!
Balint

[pocopito]

I've analysed the starting position deeper and nps are higher:

Code: Select all

[+0.28 ]  Depth: 26/37  Nodes: 446568K (740 kn/s) Time: 603.06 s
NPS: 740500  Hash: ----  Load: ---- TB hits: ---- Current: d4 (1/20) 

It's not the million nps that you reach, but the increase is quite a bit.

By the way, running on just 1 thread, I get these numbers:

Code: Select all

[+0.24 ]  Depth: 23/35  Nodes:  37637K (368 kn/s) Time: 102.15 s
NPS: 368464  Hash: ----  Load: ---- TB hits: ---- Current: Nf3 (2/20) 

I'm running these tests under scid-vs-pc, and AFAIK there's no option to get a log of the nps at each depth.


EDIT- Just forget the next part: I solved this problem installing some extra mono libraries for the 2.0 version.


And these are the errors for the 2.0 versions:
For x64:

Code: Select all

$ mono PortfishNet20.exe 

Unhandled Exception: System.TypeLoadException: A type load exception has occurred.
  at System.Threading.Thread.StartInternal () &#91;0x00000&#93; in <filename unknown>&#58;0 
&#91;ERROR&#93; FATAL UNHANDLED EXCEPTION&#58; System.TypeLoadException&#58; A type load exception has occurred.
  at System.Threading.Thread.StartInternal () &#91;0x00000&#93; in <filename unknown>&#58;0 

For x86:

Code: Select all

$ mono PortfishNet20.exe 

Unhandled Exception: System.TypeLoadException: A type load exception has occurred.
  at System.Threading.Thread.StartInternal () &#91;0x00000&#93; in <filename unknown>&#58;0 
&#91;ERROR&#93; FATAL UNHANDLED EXCEPTION&#58; System.TypeLoadException&#58; A type load exception has occurred.
  at System.Threading.Thread.StartInternal () &#91;0x00000&#93; in <filename unknown>&#58;0

[bpfliegel]

Thanks a lot Emilio!

Should be somewhat better based on the latest performance reports I read over Mono, but this is ok. I'm just installing an Ubuntu 12.04 to able to check on it.

Also checked the Arasan test suite against the very latest Stockfish source. Numbers are equal.

Cheers, Balint