Scorpio 2.8.7 MCTS+NN windows version

[Guenther]

a little correction of the sudden end of the pgn to make moves and evals readable...
[pgn] [Event "Computer Chess Game"] [Site "dani-s-xp"] [Date "2018.09.08"] [Round "3"] [White "scorpio"] [Black "tscp"] [Result "*"] [TimeControl "40/120"] [Annotator "1. +0.00 9... -0.84"] 1. e4 {+0.00/70} c5 2. Nf3 {+0.00/70 0.1} d6 3. d4 {+0.00/70 0.1} cxd4 4. Nxd4 {+0.00/70 0.1} Nf6 5. Nc3 {+0.00/70 0.1} Nc6 6. Bg5 {+0.00/70 0.1} e6 7. Qd2 {+0.00/70 0.1} a6 8. O-O-O {+0.00/70 0.1} h6 9. Be3 {+0.00/70 0.1} Bd7 {-0.84/6 4} 10. f3 {+0.00/70 0.1} Be7 {-0.59/6 4} 11. g4 {+0.00/70 0.1} O-O {-0.37/6 4} 12. Kb1 {+0.03/71 8} Nxd4 {-0.14/6 4} 13. Bxd4 {+1.12/71 4} e5 {-0.06/6 3} 14. Bf2 {+1.29/71 4} b5 {-0.01/5 3} 15. h4 {+1.62/71 4} Bc6 {-0.22/5 3} 16. g5 {+3.35/71 4} hxg5 {-0.15/7 3} 17. hxg5 {+3.54/71 0.9} Nh7 {-0.21/6 3} 18. Be3 {+2.96/71 4} Nxg5 {-0.21/5 2.9} 19. Qh2 {+3.75/71 4} f5 {-0.41/6 2.8} 20. Bxg5 {+4.65/71 4} Bxg5 {-0.81/6 2.7} 21. Qh7+ {+3.56/71 4} Kf7 {-0.81/5 2.6} 22. a4 {+3.14/71 4} fxe4 {+0.31/6 2.6} 23. Nxe4 {+3.54/71 4} Bxe4 {+0.33/6 2.5} 24. Qxe4 {+3.17/71 4} bxa4 {+0.43/5 2.4} 25. Bc4+ {+3.37/71 4} Ke7 {+0.43/5 2.3} 26. Rh5 {+3.64/71 4} Bf6 {+0.75/6 2.2} 27. f4 {+3.96/71 4} Rb8 {+1.53/5 2.2} 28. fxe5 {+4.01/71 4} Rxb2+ {-2.28/5 2.1} 29. Kxb2 {+7.29/71 1.0} Qb8+ {-2.67/6 2.0} 30. Ka2 {+7.09/71 4} Bxe5 {-2.64/6 1.9} 31. Rxe5+ {+5.95/71 4} dxe5 {-4.97/5 1.9} 32. Qg6 {+6.63/71 4} Rf2 {-7.94/6 1.8} 33. Qe6+ {+6.92/71 4} Kf8 {-7.94/6 1.8} 34. Qg8+ {+7.85/71 1.1} Ke7 {-9.22/6 1.7} 35. Qxb8 {+9.15/71 1.2} Rxc2+ {-10.85/6 1.6} 36. Kb1 {+11.41/71 6} Rd2 {-16.73/6 1.6} 37. Rxd2 {+11.44/71 6} Kf6 {-99.94/5 0.1} 38. Bxa6 {+14.71/71 6} g5 {-15.12/5 1.5} 39. Bb5 {+14.60/71 6} g4 {-15.76/5 1.5} 40. Rd5 {+14.59/71 7} a3 {-16.06/5 1.4} 41. Ka2 {+16.70/71 2.9} Kg6 {-99.92/6 2.0} 42. Kxa3 {+14.17/71 2.9} Kh6 {-18.11/6 5} 43. Rd8 {+18.18/71 2.9} e4 {-99.94/6 0.8} 44. Qa8 {+20.26/71 2.9} Kg5 {-18.31/7 5} 45. Qa7 {+15.70/71 2.9} Kh5 {-18.11/6 5} 46. Kb2 {+13.12/71 2.9} Kg5 {-18.00/6 5} 47. Qc5+ {+23.87/71 0.7} Kh4 {-99.94/5 1.4} 48. Qb6 {+14.84/71 2.9} Kg5 {-18.00/6 5} 49. Bc6 {+14.38/71 2.9} Kf5 {-18.05/6 4} 50. Qd4 {+16.32/71 2.9} g3 {-99.92/6 2.4} 51. Kc3 {+14.33/71 3} Kg5 {-99.92/6 1.8} 52. Rg8+ {+13.68/71 0.7} Kf5 {-99.94/4 0.1} 53. Rxg3 {+11.86/71 0.8} Ke6 {-99.96/4 0.1} 54. Bxe4 {+17.45/71 3} Ke7 {-99.94/5 0.1} 55. Qe5+ {+9.94/71 3} Kd8 {-99.96/3 0.1}* [/pgn]

[Werner]

I did not know, how strong these 6x64 net with LC0 is:

1 Lc0-1 9154 1CPU 3205 +436 111½1111111111101111 18.5/20
2 Scorpio 2.8 x64 1CPU 2769 -436 000½0000000000010000 1.5/20
perhaps I will make some more games with it...

sorry - wrong configuration with Scorpio here - I have to repeat this match.

Correction:
1 Scorpio 2.8 x64 1CPU 2769 +368 1½½1010½111½1½½11½11 14.5/20
2 Lc0-1 9154 1CPU 2401 -368 0½½0101½000½0½½00½00 5.5/20

[Daniel Shawul]

I have made some significant work with GPU acceleration of NN evaluation in this version.
Multi-threaded batching as I have described in the programming section helps a lot and now scorpio
can be launched with upto 256 threads even on a 4-core machine and a single GPU and perform well.
I get 10kn/s on a 6x64 net on a P100 GPU.

Multi-threaded batching also helps tremendously on the CPU. On a 32-core machine, I am now getting 6.4kn/s
on a 6x64 net which is about 64% of eps that I got on a P100 GPU. I think batching helps in two ways -- a) minimizing
data transfer costs b/n CPU and GPU and b) minimizing the tensorflow call overhead. I think the latter is why the CPU
is benefiting from batching now while the GPU also benefits from (a) as well.

@Werner, the current scorpio-mcts+nn binary has issues with finishing mates (endgames). I haven't done any games before
I release. I have fixed that now along with fixing multi-core issues. I will upload this once I find workarounds for multi-GPU issue that
may take some time to fix ...

Daniel

[AdminX]

Hi Daniel,

Quick question, Is there a way to specify which GPU to use. I have GPU0 and GPU1, I can tell it to use 1 or 2 GPU's but in the case of specifying which GPU to use I am unsure.

Thanks.

[Daniel Shawul]

I will add that feature in the next release but for now you can try to set the CUDA_VISIBLE_DEVICES environment variable for the scorpio process.
If you have 4 GPUs and you want to use /gpu:1 and /gpu:3 you just set CUDA_VISIBLE_DEVICES=1,3. Internally scorpio will still recognize these two devices as /gpu:0 and /gpu:1 because the numbering is local.

regards,
Daniel

[joechess]

Hi 2.8.8 CPU version was working fine, downloaded latest egbbdll64 and epd and binaries for win 7 today . i start engine now type "go".
crashes with
Fault Module Name: StackHash_1dc2

[Daniel Shawul]

If your CPU is older than 2011 (no AVX intructions) the new one won't work.
I will recompile and put up both of them.

[Scally]

Hi Daniel,

Is it possible to compile a version for the 32 bit CPU armv7l Raspberry Pi? I managed to compile LC0 using openblas and can get up to 100 nps on 3 cores, but I’m having problems compiling your engine.

Thanks,

Al.

[joechess]

If your CPU is older than 2011 (no AVX intructions) the new one won't work.
I will recompile and put up both of them.

That will be great - thankyou!!

[Daniel Shawul]

If your CPU is older than 2011 (no AVX intructions) the new one won't work.
I will recompile and put up both of them.

That will be great - thankyou!!

I have uploaded both now. The default supports sse2 intructions and is the one that should work for you.
Those with modern cpus can use the one in avx/ sub-folder.
Let me know if it works for you now.