Syzygy EGTB's via Torrent Thread

[velmarin]

I got it,

For months I had problems with earlier versions.
So far I have compiled this version works fine.
I'd say very good, I like to try on another engine.
Let's see if I find time.

[jshriver]

Just wanted to again say thank you to everyone who is participating in this.

See we are now around 5mb a second, and have 5 seeds for the 6men WDL, an amazing 10 + webseed for 3-4-5, and soon to have the huge DTZ also seeded. From the server I see right now 3 ppl stand at 64.4%, and another over 50%, and gaining speed every day.

I really hope others continue to seed once done. This is a big and initial push to get them out to whomever has it. It's only because of people doing this and sharing their bandwidth that makes this possible. There is no way sadly I could have done this on my own.

Once the dust settles, I will always have my webseed for 3-4-5, but 6men will be a lot harder and definitely a lot slower than what our group is offering right now.

Grab them while you can
-Josh

[phenri]

Until the 6 piece DTZ torrent is completed i just patched the probing code to only access the DTZ for 5 pieces or less. Now everything seems to work.

What part of the code should be changed to 5 pieces? is it needed? Because there is an UCI option to change that.

option name Probe Syzygybases type spin default 6 min 0 max 6

[IQ]

It is ONLY needed if you use 6 piece WDL Syzygy with only 5 piece DTZ Syzygy. Then you have to change the source or otherwise stockfish will just exit when it tries to load the 6 piece DTZ if a 6 piece positiion is reached (not in search as the WDL work fine).

You would have to change search.cpp where the root position is checked if it has <= 6 pieces.... this is hardcoded as a "<=6". I just changed the 6 to a 5 and stockfish no longer exits on a missing DTZ file. If you cannot find it, i can look up the exact line number - but as it is the only hardcoded "6" its not difficult to find. Beware though, there are side effects - sometimes stockfish does not win a won 6 piece ending. If you use the fritz interface with partial nalimov 6 pieces - this should happen very rarely as the gui takes over when the nalimov are hit. Even if you stumble upon a 6 piece position which your nalimov set does not cover, search and the Syzygy WDL often still win. But i also saw 2-3 games where stockfish refused to make progress.... so this is just a hack until the torrent with the DTZ bases completes.


On a side note - ignoring the 6 piece WDL without DTZ issue for a moment: maybe the hardcoded 6 should generally replaced with the "Probe Syzygybases" value or the real DTZ number of pieces present.

Until the 6 piece DTZ torrent is completed i just patched the probing code to only access the DTZ for 5 pieces or less. Now everything seems to work.

What part of the code should be changed to 5 pieces? is it needed? Because there is an UCI option to change that.

option name Probe Syzygybases type spin default 6 min 0 max 6

[RJN]

Just wanted to again say thank you to everyone who is participating in this.

See we are now around 5mb a second, and have 5 seeds for the 6men WDL, an amazing 10 + webseed for 3-4-5, and soon to have the huge DTZ also seeded. From the server I see right now 3 ppl stand at 64.4%, and another over 50%, and gaining speed every day.

I really hope others continue to seed once done. This is a big and initial push to get them out to whomever has it. It's only because of people doing this and sharing their bandwidth that makes this possible. There is no way sadly I could have done this on my own.

Once the dust settles, I will always have my webseed for 3-4-5, but 6men will be a lot harder and definitely a lot slower than what our group is offering right now.

Grab them while you can
-Josh

No problem, thanks for getting the ball rolling. I've been mentioning it occasionally on the TCEC chat, so more people have the link.

BTW, I saw speed was up to about 9MB/s tonight, and over 1.7TB transferred so far. Very nice!

I also saw it mentioned on Immortal Chess, so the torrent should keep growing, at least for now.

"Big Things have small beginnings"

[RJN]

Amazing, speed has almost doubled since my last post 4 hours ago:

17.8 MB/sec

Name/Info Hash Seeders Leechers Completed D/Ls Bytes Transferred Speed (rough estimate)

Syzygy 3-4-5 Individual Files - (Download Torrent)
938.39 MB
13 2 30 29.24 GB 344.21 KB/sec

Syzygy 6men DTZ - (Download Torrent)
81.39 GB
1 19 0 582.14 GB 6.73 MB/sec

Syzygy 6men WDL - (Download Torrent)
67.83 GB
6 16 9 1333.08 GB 10.73 MB/sec
Space Used: 150.14 GB 20 37 39 1944.46 GB 17.8 MB/sec

[phenri]

Amazing, speed has almost doubled since my last post 4

I made an announcement 8 hours ago in another forum

[Kirill Kryukov]

Well the entire set is up now and seems to be running solid. Even with only 1 seed right now, it went up from 20-30k to now about 1/2 meg. As more people join in it'll only get faster.

Enjoy! Let me know if there are any problems or concerns. I made sure to run a verify right before creating the torrents to ensure data quality.

-Josh

Downloaded the complete set, both WDL and DTZ. Thanks to Josh, Ronald, and all participants!

[RJN]

I have now downloaded them all too. No problems, checksums match.

Great team participation by all. I will keep seeding for now, but have to interrupt from time to time. (my share ratio is almost 6 already, in just over a week, but I can't keep that upload rate going indefinitely)

BTW, does anyone have a recent Stockfish windows 64 bit compile, with the path option enabled? Preferably one of the recent SF dev masters.

I would like to try putting the WDL on SSD, while keeping the DTZ on HDD, and can't do that with the no-path compile.

Again, a big thanks to everyone, especially the generous creator, Ronald de Man.

PS: I know the SF stance is not to use TBs, but what is the harm in offering the option in the next official release? Not criticizing the amazing SF team, but I think that would be a great addition, to give a choice.

[phenri]

Stockfish-master_130919_sbase_support.7z