Perft(13) [3.4 GHz Core i7-2600, 16 GB RAM]

[sje]

24 hours after the restart, about 3,100 new draft 8 records have been produced. Two more draft 11 records were output bring which bring that count to 18 of the 400 needed. These additions:

Code: Select all

rnbqkbnr/p1pppppp/1p6/8/8/1P6/P1PPPPPP/RNBQKBNR w KQkq - 0 2 11 2789110131566204
rnbqkbnr/p1pppppp/1p6/8/8/P7/1PPPPPPP/RNBQKBNR w KQkq - 0 2 11 2062282403634935

The Debian package that handles UPS management is called "nut" and has been around for a few years. But it does not handle all models of UPS units. And its main action is to email a notice of power failure and that's a bit useless to me as what I really need is for the machine to enter a graceful hibernation.

[Ajedrecista]

Hello Steven:

I am sorry that hurricane difficults everything. Your Perft count goes so fast IMO, congratulations. But I want to ask something: when Perft(13) count started, you uploaded the complete results (drafts 8, 9, ...) in your web on fcr1 file, but now there is nothing new uploaded. I think these numbers were useful to people that run Monte Carlo methods (and other kinds of methods) for Perft(13) estimations, and these people would thank you if you update intermediate results again (if possible). If not, your 24/7 count is an enough reason for gratitude.

I wanted to add my grain of salt and I run the draft 9 of the Petrov defence with the Perft counter JetChess 1.0.0.0 by Thomas Zipproth (using 1024 MB of hash). The result is here:

[d]

18,001,998,634,639 (move pathes after 9 half moves).

The 27 drafts 8 of the subsequent position are here:

Code: Select all

  1  Qd1-e2  647040707276
  2  Rh1-g1  490829624768
  3  Bf1-e2  568873960495
  4  Bf1-d3  561059003006
  5  Bf1-c4  848606605153
  6  Bf1-b5  606510996090
  7  Bf1-a6  658638271466
  8  Nb1-a3  536012964936
  9  Nb1-c3  678243895985
 10  Nf3-d4  973927879849
 11  Nf3*e5  1141766008376
 12  Nf3-g5  768354401415
 13  Nf3-h4  665523883989
 14  Nf3-g1  606215610223
 15   a2-a3  513813119181
 16   a2-a4  637345249039
 17   b2-b3  600934226082
 18   b2-b4  556993717758
 19   c2-c3  644366970275
 20   c2-c4  513136570822
 21   d2-d3  728364218922
 22   d2-d4  1359941729389
 23   g2-g3  590435349618
 24   g2-g4  543336242926
 25   h2-h3  584243924514
 26   h2-h4  601336505861
 27  Ke1-e2  376146997225

Total:   18001998634639

Time: 24749.052 s

I do not know if you have calculated this draft before. If so, my count should be a verification; if not, someone (mainly you) has to verify me. Before starting my count, I estimated (in view of other shorter counts I did) that speed would be between 700 MN/s and 725 MN/s; it was around 727.4 MN/s finally. I wonder how many minutes takes your machine (and your programme) counting this draft 9... I guess that few minutes (with a very high speed in terms of nodes/second). As you see, my count is not practical because this draft 9 took 6:52:29.052 and there are 72,078 different draft 9 positions for Perft(13). But as I said before, I wanted to add my (insignificant) grain of salt.

Good luck with hurricanes and other incidentals! Please keep the good work: it is very much appreciated.

Regards from Spain.

Ajedrecista.

[sje]

The web files haven't been updated in a while as I'm moving away from the me.com storage because Apple is dropping support. Actually, they never did it all that well, so it's not a big loss.

[d]

Code: Select all

[] emptran 9 
Bd3 561,059,003,006
Bb5 606,510,996,090
Ba6 658,638,271,466
Bc4 848,606,605,153
Be2 568,873,960,495
Na3 536,012,964,936
Ke2 376,146,997,225
Nc3 678,243,895,985
Ng1 606,215,610,223
Ng5 768,354,401,415
Nh4 665,523,883,989
Nd4 973,927,879,849
Rg1 490,829,624,768
a3 513,813,119,181
Qe2 647,040,707,276
b3 600,934,226,082
a4 637,345,249,039
b4 556,993,717,758
Nxe5 1,141,766,008,376
c3 644,366,970,275
c4 513,136,570,822
g3 590,435,349,618
d3 728,364,218,922
g4 543,336,242,926
h3 584,243,924,514
h4 601,336,505,861
d4 1,359,941,729,389
Depth: 9   Count: 18,001,998,634,639   Elapsed: 2804.11  (6.41985e+09 Hz / 1.55767e-10 s)

[Ajedrecista]

Hello Steven:

Thanks for replying so fast and for explaining the issue of chessnotation. I see that our counts are the same in the Petrov defence, so it is good. I thought that your machine was even faster than it is: 0:46:44.11 and a speed of around 6.42 GN/s, which is almost nine times faster than the perft counter I used (only in one core because JetChess has not multi-core support). I suppose that your perft counter runs in more than one core... Anyway, it is incredible how good is the performance of your programme! Have your predictions of January 2012 changed with all these hurricanes? Thanks in advance.

Regards from Spain.

Ajedrecista.

[sje]

The above result was done on a 2.66 GHz quad core Mac Pro. The machine is five years old and still performs well although it has burned out a couple of disks.

I think there's a good chance of finishing the main calculation by the end of January 2012.

At present, the run has produced 20 of the 400 draft 11 records needed:

Code: Select all

rnbqkbnr/1ppppppp/p7/8/8/N7/PPPPPPPP/R1BQKBNR w KQkq - 0 2 11 1865423942798492
rnbqkbnr/1ppppppp/p7/8/8/2N5/PPPPPPPP/R1BQKBNR w KQkq - 0 2 11 2458021022756805
rnbqkbnr/1ppppppp/p7/8/8/1P6/P1PPPPPP/RNBQKBNR w KQkq - 0 2 11 2110456650953864
rnbqkbnr/1ppppppp/p7/8/8/P7/1PPPPPPP/RNBQKBNR w KQkq - 0 2 11 1562327606120748
rnbqkbnr/1ppppppp/p7/8/8/7N/PPPPPPPP/RNBQKB1R w KQkq - 0 2 11 1882390155585040
rnbqkbnr/1ppppppp/p7/8/8/5N2/PPPPPPPP/RNBQKB1R w KQkq - 0 2 11 2373558610955814
rnbqkbnr/1ppppppp/p7/8/P7/8/1PPPPPPP/RNBQKBNR w KQkq - 0 2 11 2250225134770081
rnbqkbnr/1ppppppp/p7/8/1P6/8/P1PPPPPP/RNBQKBNR w KQkq - 0 2 11 2178802463490013
rnbqkbnr/1ppppppp/8/p7/8/N7/PPPPPPPP/R1BQKBNR w KQkq - 0 2 11 2670298917876179
rnbqkbnr/1ppppppp/8/p7/P7/8/1PPPPPPP/RNBQKBNR w KQkq - 0 2 11 2373873098604319
rnbqkbnr/1ppppppp/8/p7/8/2N5/PPPPPPPP/R1BQKBNR w KQkq - 0 2 11 3411299116725061
rnbqkbnr/1ppppppp/8/p7/8/P7/1PPPPPPP/RNBQKBNR w KQkq - 0 2 11 2171307455092881
rnbqkbnr/1ppppppp/8/p7/8/1P6/P1PPPPPP/RNBQKBNR w KQkq - 0 2 11 3039243237480582
rnbqkbnr/1ppppppp/8/p7/8/7N/PPPPPPPP/RNBQKB1R w KQkq - 0 2 11 2654960388372645
rnbqkbnr/1ppppppp/8/p7/8/5N2/PPPPPPPP/RNBQKB1R w KQkq - 0 2 11 3354897405349959
rnbqkbnr/1ppppppp/8/p7/1P6/8/P1PPPPPP/RNBQKBNR w KQkq - 0 2 11 3850353951070799
rnbqkbnr/p1pppppp/1p6/8/8/1P6/P1PPPPPP/RNBQKBNR w KQkq - 0 2 11 2789110131566204
rnbqkbnr/p1pppppp/1p6/8/8/P7/1PPPPPPP/RNBQKBNR w KQkq - 0 2 11 2062282403634935
rnbqkbnr/p1pppppp/1p6/8/8/7N/PPPPPPPP/RNBQKB1R w KQkq - 0 2 11 2486684623836455
rnbqkbnr/p1pppppp/1p6/8/8/N7/PPPPPPPP/R1BQKBNR w KQkq - 0 2 11 2429530164149293

[sje]

I've updated the online files; in the Perft->Perft13 directory, the fcr0 (last restart) and fcr1 (current output) text files give the recent progress.

The 21st of the 400 draft 11 records:
[d]
Perft(11) for the above: 3,127,051,188,240,924

I'm expecting a hurricane-related power failure soon, possibly lasting most of the week.

[Ajedrecista]

Hi Steven:

Thank you very much for updating the online files at Chess Notation. I can do nothing but see the progress of the count with these files; but other people can test their estimating algorithms and may improve them with all the data available. 21 drafts 11 are counted so far, that is great news! Please keep the good work and have good luck with hurricanes.

Regards from Spain.

Ajedrecista.

[sje]

The 22nd draft 11 record:
[d]
Perft(11) for the above: 3,098,666,191,905,313

22 down, 378 to go.

The worst of the storm should pass about 150 km to the west in about seven hours.

[sje]

The 23rd draft 11 record:
[d]
Perft(11) for the above: 3,210,768,736,129,942

23 down (5.75%), 377 (94.25%) to go.

Amazingly, the power here hasn't failed yet. In spite of some of the highest rates in the country, the local electricity suppliers are unable to keep the delivery system in repair, and so they are always calling on the Canadian lineman crews of Hydro Quebec to fix storm damage. These workers are welcome in the local restaurants as they tip well and the Canadian dollar is strong.

[sje]

After 27 days and a few power outages, the run has produced 111,602 draft 8 records. The new estimate for total run time is now 199 days, an increase of 19 days due in part to the need to recalculate the partial results lost during the outages.

The new projected completion date is January 20, 2012.