7-men Syzygy attempt

[jjoshua2]

I made a 7 man syzygy leela PR if anyone wants to test it. It picks up all the files and seems to use them without crashing at least. The probing code would be the same that SF would use almost. Root probing is a bit different because we have UCTNodes.
https://github.com/glinscott/leela-chess/pull/557

[duncan]

I've removed these from my mirror, and evidently, so has Lichess.

I have also removed those, and some regenerated files has been uploaded, please re-check and make sure they are correct now.

As expected, the new KQBNvKQN.rtbz solves the problems I noticed.

do you know if the new generator will also work for 8 pieces ?

[syzygy]

do you know if the new generator will also work for 8 pieces ?

In theory it should work. In practice you'll need a very very big machine and a lot of patience.

It would probably be a good idea to rewrite the generator to use a more space-efficient indexing scheme during generation. Generation of the biggest table should then fit in about 41 TB (so let's say 48 TB) and most tables would need half of that or less. The current scheme would need 57.75 TB (so basically a 64 TB machine).

The bigger 8-piece tables will fill more than 2^32 64-byte blocks with compressed data, which means the format will need a relatively small extension to deal with them.

[duncan]

do you know if the new generator will also work for 8 pieces ?

In theory it should work. In practice you'll need a very very big machine and a lot of patience.

It would probably be a good idea to rewrite the generator to use a more space-efficient indexing scheme during generation. Generation of the biggest table should then fit in about 41 TB (so let's say 48 TB) and most tables would need half of that or less. The current scheme would need 57.75 TB (so basically a 64 TB machine).

The bigger 8-piece tables will fill more than 2^32 64-byte blocks with compressed data, which means the format will need a relatively small extension to deal with them.

do you know how much hard disk space 7 and 8 pieces will take up and do they result in any elo gain ?

[syzygy]

do you know how much hard disk space 7 and 8 pieces will take up and do they result in any elo gain ?

A rough guess would be 20 TB and 3 PB.
Elo gain depends on access speed. How much Elo they would gain if they could be accessed without delay is anyone's guess.

[noobpwnftw]

All 4v3 pawnless tables are now built and uploaded.

I will take a break on the 4v3 pawnful tables to prepare storage and try to verify that the tables are correct while the 5v2 pawnful tables can proceed in par.

[Nordlandia]

Is set KBPPvPPK early or late in the automatic scheduler?

I'm very interested in this spesific set because it provide vital fortress information in the famous Spassky vs Fischer (1st matchgame)

If engine identify this position as draw, cfish might be able to evaluate the position after 29...Bxh2 as triple zero!



See 2k5/1p6/pP2K3/P7/8/4B3/8/8 w - - 0 0

[noobpwnftw]

Is set KBPPvPPK early or late in the automatic scheduler?

This would appear to be a late one.


Also I have rearranged folder structures on my FTP.

There will be a minor problem though, as I may be unable to host all files on one server, there may be a second server to host the remaining files.

[niklasf]

Congrats on the pawnless milestone!

All currently available tables pass some basic sanity checks, i.e. the max-DTZ positions match the stats file and their mainlines are consistent. Such a basic check is of course unlikely to catch random hardware errors, but it would have caught all previous bugs. I'll keep testing incoming tables on the Lichess mirror.

[duncan]

Congrats on the pawnless milestone!

All currently available tables pass some basic sanity checks, i.e. the max-DTZ positions match the stats file and their mainlines are consistent. Such a basic check is of course unlikely to catch random hardware errors, but it would have caught all previous bugs. I'll keep testing incoming tables on the Lichess mirror.

do you have the url ?