Perft(14) Weekly Status Report

[sje]

I expect a significant increase in the production rate when the OpenCL version of Oscar comes online.

The set of all 965 work units is available for anyone who wishes to perform original calculations. Also, the growing set of completed work units is also available for those who would like to assist with confirmation work.

[Henk]

I did not follow but if you can detect mate do you know how many times white or black lost by check mate.

[Ajedrecista]

Hello Henk:

I did not follow but if you can detect mate do you know how many times white or black lost by check mate.

There are stats up to ply 12, but not separated by colour:

Statistics on chess games

Of course, the data of ply 4 refers to white checkmated only (eight paths of the Fool's Mate).

Regards from Spain.

Ajedrecista.

[Henk]

Maybe it is easier to compute and count check mates only. If that is true one might compute the checkmates for ply 15 first. With that figure one might perhaps give better predictions for perft where ply > 15

[sje]

Perft(14) Weekly Status 2014-12-07

There are more than 3,800,000 perft(7) results so far, about 3.94% of the 96,400,068 needed.

Day count: 121
Average throughput: 31,405 results/day

Completed work units (39): 000-032, 034-035, 037, 039-040, 964
In progress (10): 033, 036, 038, 041-047
Not yet started (916): 048-963

Work units in progress:

Code: Select all

WU#  Comp%  Machine
---  -----  -------
033  50.7%  megan
036  86.2%  gail
038  31.8%  joni
041  13.2%  melissa
042  61.4%  serra
043  72.1%  rocky
044  27.4%  cynthia
045   9.0%  betty
046  23.8%  kristen
047  15.6%  amanda

[ankan]

I modified my GPU perft program to parse and process the perft(7) records created by Steven for perft(14) calculation.

Tested with work units 0, 33, 36 and 964.

Work unit 0 took slightly more than 4 hours (on an overclocked GTX 970 graphics card). WU 33 and 36 took significantly longer (almost double the time).

I have uploaded results here:
https://drive.google.com/file/d/0B4d0eT ... sp=sharing

Currently I am running work units 38, 41 and 42.

[sje]

I modified my GPU perft program to parse and process the perft(7) records created by Steven for perft(14) calculation.

Tested with work units 0, 33, 36 and 964.

000 and 964 match.
033 and 036 are not yet finished, but the first 50,000 records of each match.

It's good to see verification.

Status snapshot:

Code: Select all

Work units in progress:

WU#  Comp%  Machine
---  -----  -------
033  53.7%  megan
036  89.8%  gail
038  34.8%  joni
041  14.8%  melissa
042  73.8%  serra
043  83.0%  rocky
044  36.0%  cynthia
045  12.5%  betty
046  39.2%  kristen
047  30.2%  amanda

[sje]

A bit further with 036; the first 89,000 records match.

[sje]

MD5 signatures of completed work units:

Code: Select all

MD5 (wu7.000.sum) = ee4d4b9dfcaa83d3ca0466e69428596d
MD5 (wu7.001.sum) = be830cf0547d160243b6a8a692fea187
MD5 (wu7.002.sum) = 829b085d4dca5d98c3659121182c8d27
MD5 (wu7.003.sum) = 346ddd03b8b1516980cded91cbf165d9
MD5 (wu7.004.sum) = ada99d99d282d6c91fc14afe270cfd0d
MD5 (wu7.005.sum) = 3f2eadd0f1211ead70ee1b142ad541e2
MD5 (wu7.006.sum) = 6856a8f0c3f8f6603ac098b4ecd441c5
MD5 (wu7.007.sum) = 0d7f614cad654c3c6ded9089dbb39440
MD5 (wu7.008.sum) = 2097681ae9c1d2804a6df2454f749a0d
MD5 (wu7.009.sum) = 32f5bae2f437938e9e1b1e074d09170b
MD5 (wu7.010.sum) = 83590d1cc6c065164889aed917cff635
MD5 (wu7.011.sum) = 6f50aa13d0a13e7fe4e348e00b0816ab
MD5 (wu7.012.sum) = 62ae3c9552d20354d2e38947bd0f2c5d
MD5 (wu7.013.sum) = 91b7af99d271fcc1bc3268ad7b52aed7
MD5 (wu7.014.sum) = 044ef13b5f89dbd7c3c96052fcaf00ea
MD5 (wu7.015.sum) = 23c8e14dc1a5effab567d19bdb16e550
MD5 (wu7.016.sum) = 4f415d18a888ebd42f34352c5b286f2d
MD5 (wu7.017.sum) = 74edf95f4f0b2cbc1f79277957e0b7ae
MD5 (wu7.018.sum) = 16201b7a1a7813f48b7f60f3d3321a8b
MD5 (wu7.019.sum) = 297402b60c9c9e6c97ecfa36aa336470
MD5 (wu7.020.sum) = 836f634bb99f38adc240174e303968f8
MD5 (wu7.021.sum) = 23179bef9df14b26bb6c8309fd6a5663
MD5 (wu7.022.sum) = 70a628e964735d8e39d112bac5eb378d
MD5 (wu7.023.sum) = efe5897882b86910f01810f56b8d6a3b
MD5 (wu7.024.sum) = 1fb3134eb53d4169056b47e052bd3828
MD5 (wu7.025.sum) = 3cd8a0c2b812f523c387e5fab1d2f44d
MD5 (wu7.026.sum) = 3b3be756f845855c36aa24df01fdf81e
MD5 (wu7.027.sum) = c819a74ef7845f5e36b3caff43e7f440
MD5 (wu7.028.sum) = 310f7b238fce2102d75c75c69bd2c79c
MD5 (wu7.029.sum) = d9bd750b4dd5a320afd46490bb4331b3
MD5 (wu7.030.sum) = 71dc01a935d4a320a7b22a90c60b1a0e
MD5 (wu7.031.sum) = 260aefcfb3f5aed63839f2af1af10cd8
MD5 (wu7.032.sum) = 926d5b436bd18163aa370bf1b662a72d
MD5 (wu7.034.sum) = 8b12fcfe348c93e93af9facc24a691eb
MD5 (wu7.035.sum) = b0e7ea473b2bf64702e65a0450ec82b3
MD5 (wu7.037.sum) = 955c98dcfb9c16ae6c0bbf6f1e199132
MD5 (wu7.039.sum) = b155795121f2ef370d0bad64f9d94df8
MD5 (wu7.040.sum) = 04755aeca1cb80032b8f60ef795b7269
MD5 (wu7.964.sum) = 1b329dcb65c93460e614a88823854578

[sje]

To calculate MD5 (Message Digest 5) checksums:

On either Mac OS/X or Linux:

Code: Select all

openssl md5 <filename>

Alternatively, on Mac OS/X:

Code: Select all

md5 <filename>

Alternatively, on Linux:

Code: Select all

md5sum <filename>

I don't know how it's done on Windows.

The MD5 checksum of a file is dependent only upon the contents of the file and not upon the host operating system or the host CPU. Using secure checksums makes it easy to compare file contents for equality without having to have both versions of a file on the same machine.