Easy idea to improve infinite analysis mode!?

Discussion of chess software programming and technical issues.

Moderators: Harvey Williamson, bob, hgm

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Jouni
Posts: 2227
Joined: Wed Mar 08, 2006 7:15 pm

Easy idea to improve infinite analysis mode!?

Post by Jouni » Fri Feb 14, 2020 5:32 pm

Engine should output something after every minute! And after first hour after each 5 minutes. You get the idea. Specially SF has tendency to search explosion/bug, when nothing happens in next hour AAARGH. Now You will be sure, that it's still running OK. Also useful in long games like TCEC.
Jouni

MOBMAT
Posts: 327
Joined: Sat Feb 04, 2017 10:57 pm
Location: USA

Re: Easy idea to improve infinite analysis mode!?

Post by MOBMAT » Fri Feb 14, 2020 10:47 pm

You are right, a lot of engines get deep info a search and don't send out anything.
Some engines report PVs and other search results but are more stingy on some of the other reporting info.
Of course in their defense, outputting stuff takes CPU time.

The protocol for the UCI interface

http://wbec-ridderkerk.nl/html/UCIProtocol.html

indicates that some info should be sent out now and then.. to quote the document

"the engine should send this info regularly"

referring to:

info nodes (number of nodes searched...so far)
info hashfull (the % of the hash that is full)
info nps (the nodes per second)

any one or more of these every few seconds would be sufficient

if using TBs, the following info would suffice to let us know the engine is still alive.
info tbhits (the number of positions that were found in the endgame table bases)

As I write this, I'm watching Houdini 1.5a chomp on a mate problem. It is at ply 40 and besides the PV info, it is updating the tbhits, the hashfull, and the nps a couple of times a second. it also updates the current move and move number to show progress.

Each engine is different and the UCI protocol doesn't require much of anything that is basically for human consumption.
i7-6700K @ 4.00Ghz (using 6 threads), EGTBs on PCI SSD
Benchmark: Stockfish 11 64 bmi2 (nps): 2067669

User avatar
hgm
Posts: 25095
Joined: Fri Mar 10, 2006 9:06 am
Location: Amsterdam
Full name: H G Muller
Contact:

Re: Easy idea to improve infinite analysis mode!?

Post by hgm » Sun Feb 16, 2020 1:00 pm

Basically you are asking for support of the standard option UCI_ShowCurrentLine. Just repeating the same current move over and over again is hardly helpful. Sure, you will see the node count and time used go up, but the only thing you will learn from that is that the engine is still alive. (Which for a reliable engine should be a certainty to begin with.)
Get rid of the shit: vote for SHID!

Post Reply