Stockfish 3 PA_GTB
-
- Site Admin
- Posts: 1226
- Joined: Wed Jun 09, 2010 7:49 am
- Real Name: Jeremy Bernstein
- Location: Berlin, Germany
- Contact:
Re: Stockfish 3 PA_GTB
user,
I did an implementation using LMDB today and it's pretty nice. Super API.
On OSX, the file size is much smaller and a better reflection of the tiny bits of data which are being stored (19 bytes per record, 8 byte key + 11 byte data). The major disadvantage vis a vis QDBM is that LMDB memory-mapped files on Windows don't grow as incrementally as the QDBM files. They start at 1MB and expand (presumably 1MB at a time) to fit the size of the data. Not a big deal, I guess, but it will change the way "Prune" works, since pruning will basically lead to the complete decimation of all data in the PH file on Windows.
On the other hand, the user can specify a hard limit to the file, and if I can come up with a good way to deal with overflow (new entries will probably just have to replace the first entry found with a lower depth), it's pretty cool. Prune will drop off, in that case, or be offered as a way to eliminate entries at a certain depth and below en masse.
I can't say all that much about speed, since the biggest time cost is doing the TT lookups on the interesting positions. PH file read/write is nothing compared to that.
Once I write a conversion from the QDBM format to LMDB and work out how to deal with pruning and overflow, I'll send you something to try, if you're interested (just PM me your system details).
Jeremy
I did an implementation using LMDB today and it's pretty nice. Super API.
On OSX, the file size is much smaller and a better reflection of the tiny bits of data which are being stored (19 bytes per record, 8 byte key + 11 byte data). The major disadvantage vis a vis QDBM is that LMDB memory-mapped files on Windows don't grow as incrementally as the QDBM files. They start at 1MB and expand (presumably 1MB at a time) to fit the size of the data. Not a big deal, I guess, but it will change the way "Prune" works, since pruning will basically lead to the complete decimation of all data in the PH file on Windows.
On the other hand, the user can specify a hard limit to the file, and if I can come up with a good way to deal with overflow (new entries will probably just have to replace the first entry found with a lower depth), it's pretty cool. Prune will drop off, in that case, or be offered as a way to eliminate entries at a certain depth and below en masse.
I can't say all that much about speed, since the biggest time cost is doing the TT lookups on the interesting positions. PH file read/write is nothing compared to that.
Once I write a conversion from the QDBM format to LMDB and work out how to deal with pruning and overflow, I'll send you something to try, if you're interested (just PM me your system details).
Jeremy
-
- Site Admin
- Posts: 1226
- Joined: Wed Jun 09, 2010 7:49 am
- Real Name: Jeremy Bernstein
- Location: Berlin, Germany
- Contact:
Re: Stockfish 3 PA_GTB
Version PA_GTB-006 is released!
v006
- Persistent Hash:
- reduced record size
- rewritten to use new database format (Kyoto Cabinet), with automatic translation from old format (QDBM)
- added new "importepd" feature, see below
- added new "Use Persistent Hash As Book Depth" UCI option
- pruning the PH file creates/appends to a "stockfish_pruned.hsh" backup of the discarded records
- "general improvements"
- further 3x repetition checking improvements
- update to the tip of the Stockfish upstream master branch
For those of you following along, you'll note that the database format has changed (not to LMDB as initially attempted, but to Kyoto Cabinet due to difficulties with the LMDB format for this application). Your old database will be automatically converted to the new format (and the old file renamed) the first time you run v006. The new database format is incompatible with v005, so you should probably install to a different folder if you intend to use both versions.
Once again, big thanks to my intrepid testers who teased a bunch of great bugs out and, more to the point, offered some fantastic suggestions in this dev cycle.
Jeremy
v006
- Persistent Hash:
- reduced record size
- rewritten to use new database format (Kyoto Cabinet), with automatic translation from old format (QDBM)
- added new "importepd" feature, see below
- added new "Use Persistent Hash As Book Depth" UCI option
- pruning the PH file creates/appends to a "stockfish_pruned.hsh" backup of the discarded records
- "general improvements"
- further 3x repetition checking improvements
- update to the tip of the Stockfish upstream master branch
For those of you following along, you'll note that the database format has changed (not to LMDB as initially attempted, but to Kyoto Cabinet due to difficulties with the LMDB format for this application). Your old database will be automatically converted to the new format (and the old file renamed) the first time you run v006. The new database format is incompatible with v005, so you should probably install to a different folder if you intend to use both versions.
Once again, big thanks to my intrepid testers who teased a bunch of great bugs out and, more to the point, offered some fantastic suggestions in this dev cycle.
Jeremy
- Attachments
-
- stockfish-3-PA_GTB-006.7z
- (2.62 MiB) Downloaded 2301 times
- Carlos Ylich
- Posts: 56
- Joined: Thu Jun 10, 2010 3:04 am
- Location: São Paulo
Re: Stockfish 3 PA_GTB
Jeremy you is great !
Great job! Thanks
Great job! Thanks
Re: Stockfish 3 PA_GTB
I did a 100 Game test of Stock fish 3 PA_GTB 006 vs. critter 1.6a.
Both engines using 5 piece TB, 4096MB hash.
40/5 time control, no persistent hash, no pondering, both engines using own book. (I downloaded the books for both engines off their respective websites).
Engine Score St Cr S-B
1: Stockfish-3-64-PA_GTB-006 57.5/100 •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 1==1=1==1=0011=======0=1=1======00==1=11=100==11===1===0==0===1=111111=1=11=1==01=00=0==1====011===0 2443.7
2: Critter_1.6a_64bit 42.5/100 0==0=0==0=1100=======1=0=0======11==0=00=011==00===0===1==1===0=000000=0=00=0==10=11=1==0====100===1 •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 2443.7
100 games played / Tournament finished
Name of the tournament: Test vs. critter
Site/ Country: THEBIGCOMPUTER, United States
Level: Tournament 40/5
Hardware: Intel(R) Core(TM) i5-3350P CPU @ 3.10GHz with 15.9 GB Memory
Operating system: Windows 8 Pro Professional (Build 9200) 64 bit
Bayeselo shows that this result is 40 points higher than critter 1.6a at 40/5 blitz on a quad core Intel.
CCRL shows critter 1.6a is 3221 on the 40/5 list, so this version is approximately 3261.
That is higher anything except Houdini 3 64bit (3316) - a bit more tweaking, and this might steal the number 1 spot!
It is also an impressive 48 ELO higher than the main branch of stock fish 3.
Both engines using 5 piece TB, 4096MB hash.
40/5 time control, no persistent hash, no pondering, both engines using own book. (I downloaded the books for both engines off their respective websites).
Engine Score St Cr S-B
1: Stockfish-3-64-PA_GTB-006 57.5/100 •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 1==1=1==1=0011=======0=1=1======00==1=11=100==11===1===0==0===1=111111=1=11=1==01=00=0==1====011===0 2443.7
2: Critter_1.6a_64bit 42.5/100 0==0=0==0=1100=======1=0=0======11==0=00=011==00===0===1==1===0=000000=0=00=0==10=11=1==0====100===1 •••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••••• 2443.7
100 games played / Tournament finished
Name of the tournament: Test vs. critter
Site/ Country: THEBIGCOMPUTER, United States
Level: Tournament 40/5
Hardware: Intel(R) Core(TM) i5-3350P CPU @ 3.10GHz with 15.9 GB Memory
Operating system: Windows 8 Pro Professional (Build 9200) 64 bit
Bayeselo shows that this result is 40 points higher than critter 1.6a at 40/5 blitz on a quad core Intel.
CCRL shows critter 1.6a is 3221 on the 40/5 list, so this version is approximately 3261.
That is higher anything except Houdini 3 64bit (3316) - a bit more tweaking, and this might steal the number 1 spot!
It is also an impressive 48 ELO higher than the main branch of stock fish 3.
Re: Stockfish 3 PA_GTB
Stockfish 120713 64-bit 8CPU is being used in the GB CHAMPIONSHIP 8CPU tournament yet I have been unable to find any download site for this one? Is this an update one of yours?
Re: Stockfish 3 PA_GTB
Thanks Mr Jeremy
Outstanding job. Very strong engine.
Outstanding job. Very strong engine.
Re: Stockfish 3 PA_GTB
Hello Jeremy,
Do you think you can bring us to the tip of the master?
Thanks
Do you think you can bring us to the tip of the master?
Thanks
Re: Stockfish 3 PA_GTB
Happy. Thanks Jeremy.Jeremy Bernstein wrote:Over 250 downloads and not a peep. It just works? Everyone's happy? Awkward silence?
jb
Best,
Gerold.
-
- Site Admin
- Posts: 1226
- Joined: Wed Jun 09, 2010 7:49 am
- Real Name: Jeremy Bernstein
- Location: Berlin, Germany
- Contact:
Re: Stockfish 3 PA_GTB
Next week, sure. Jbendings wrote:Hello Jeremy,
Do you think you can bring us to the tip of the master?
Thanks