Fizbo 1.5 released

Discussion of anything and everything relating to chess playing software and machines.

Moderators: Harvey Williamson, bob, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
User avatar
Graham Banks
Posts: 34539
Joined: Sun Feb 26, 2006 9:52 am
Location: Auckland, NZ

Re: Fizbo 1.5 released

Post by Graham Banks » Wed Aug 19, 2015 3:55 am

Graham Banks wrote:
ThatsIt wrote:Another problem: too many time losses @ level xy move in zz time repeated !
I got 8 within only 200 games.

Best wishes,
G.S.
I've alerted Fabio to all the concerns in this thread.
From Youri:
I added option "SyzygyPath". It does the same thing as "NalimovPath". The engine only supports syzygy EGTBs, but their path can be passed to engine under either of those two options. If there are no syzygy files in that path, then syzygy EGTBs will not be used.

Updated executables are posed on https://sites.google.com/site/fizbochessengine/.

Time losses: i did not see them in my testing. I can only hope that they are caused by small excess of time over allowed limit, so i increased the time margin in my engine - hopefully this will fix the problem. If the problem persists, i would probably need to see a log of all traffic between GUI and engine.

Youri.
gbanksnz at gmail.com

Modern Times
Posts: 2604
Joined: Thu Jun 07, 2012 9:02 pm

Re: Fizbo 1.5 released

Post by Modern Times » Wed Aug 19, 2015 5:06 am

SzG wrote: I'd remove that Nalimovpath option. Why mislead people?
Agreed.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

User avatar
Matthias Gemuh
Posts: 3238
Joined: Thu Mar 09, 2006 8:10 am
Contact:

Re: Fizbo 1.5 released

Post by Matthias Gemuh » Wed Aug 19, 2015 6:25 am

SzG wrote:
ThatsIt wrote:I added option "SyzygyPath". It does the same thing as "NalimovPath". The engine only supports syzygy EGTBs, but their path can be passed to engine under either of those two options. If there are no syzygy files in that path, then syzygy EGTBs will not be used.

...

Youri.
I'd remove that Nalimovpath option. Why mislead people?
It will mislead engines too.
My engine was quite strong till I added knowledge to it.
http://www.chess.hylogic.de

ymatioun
Posts: 64
Joined: Fri Oct 18, 2013 9:40 pm
Location: New York

Re: Fizbo 1.5 released

Post by ymatioun » Wed Aug 19, 2015 10:12 pm

OK, i will remove "Nalimov" option right away - updated files will be on my site within an hour.

Youri.

carldaman
Posts: 2004
Joined: Sat Jun 02, 2012 12:13 am

Re: Fizbo 1.5 released

Post by carldaman » Wed Aug 19, 2015 11:16 pm

ymatioun wrote:OK, i will remove "Nalimov" option right away - updated files will be on my site within an hour.

Youri.
Thanks for giving us Fizbo, Youri! :)

I'm noticing that enabling Syzygy usage causes Fizbo to gobble up more and more RAM. The amount of RAM used keeps increasing as Fizbo is calculating.

There is also a big drop in nps and depth in the endgames.

The first released binary of Fizbo 1.5 did not have these issues, but, of course, could not access the Syzygy tablebases, as it was hardcoded to look for them on the H: drive.

I'm wondering if others are witnessing the same behavior.

(Win 7, 64-bit system here.)

Regards,
CL

ymatioun
Posts: 64
Joined: Fri Oct 18, 2013 9:40 pm
Location: New York

Re: Fizbo 1.5 released

Post by ymatioun » Wed Aug 19, 2015 11:48 pm

I just ran a few games with the latest file (to make sure i did not mess up compile) and i don't see any memory increase. Normally i run 1,000 games in testing, and that does not lead no any memory increase.

Slow search with EGTB turned on: i don't see it either. I just ran search on a simple position (8/4k3/3pp3/8/8/3PP3/4K3/8 w - - 0 1): with SYZYGY, i see depth 43 (and stop) after 0.1 sec and 326,448 nodes. Without EGTB i stop at depth 43 after 12 seconds and 48 Million nodes.

But i think EGTB files have to be on a media with low latency - SSD or similar. If you have them on a hard drive, they may significantly reduce performance. Other than this, i have no ides why you see these problems. All i can say is that i don't see anything like that.

Youri.

carldaman
Posts: 2004
Joined: Sat Jun 02, 2012 12:13 am

Re: Fizbo 1.5 released

Post by carldaman » Wed Aug 19, 2015 11:55 pm

ymatioun wrote:I just ran a few games with the latest file (to make sure i did not mess up compile) and i don't see any memory increase. Normally i run 1,000 games in testing, and that does not lead no any memory increase.

Slow search with EGTB turned on: i don't see it either. I just ran search on a simple position (8/4k3/3pp3/8/8/3PP3/4K3/8 w - - 0 1): with SYZYGY, i see depth 43 (and stop) after 0.1 sec and 326,448 nodes. Without EGTB i stop at depth 43 after 12 seconds and 48 Million nodes.

But i think EGTB files have to be on a media with low latency - SSD or similar. If you have them on a hard drive, they may significantly reduce performance. Other than this, i have no ides why you see these problems. All i can say is that i don't see anything like that.

Youri.
They are 5-man syzygy tbs on a regular HD, but these are not supposed to impact performance as much as 6-man syzygy (which need SSD), unless I'm mistaken.

ymatioun
Posts: 64
Joined: Fri Oct 18, 2013 9:40 pm
Location: New York

Re: Fizbo 1.5 released

Post by ymatioun » Thu Aug 20, 2015 12:02 am

Clarification:

Fizbo uses EGTBs two ways: W/L/D tables are used inside search. So they get accessed as soon as number on pieces on board drops to 5. This part requires tables to be on SSD, because they are accessed a lot.

Second way is using depth-to-zero tables when there are 5 or fewer piece on the board. This piece would work from hard drive.

Unfortunately, currently i have no way to only do the second part without the first one. So if files are not on SSD then endgame search will be very slow.

Youri.

ymatioun
Posts: 64
Joined: Fri Oct 18, 2013 9:40 pm
Location: New York

Re: Fizbo 1.5 released

Post by ymatioun » Thu Aug 20, 2015 10:03 pm

No, only 5 and fewer.

Post Reply