Ivanhoe plays singlecore in TCEC 13 !!

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

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
pohl4711
Posts: 996
Joined: Sat Sep 03, 2011 5:25 am
Location: Berlin, Germany
Contact:

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by pohl4711 » Sat Aug 04, 2018 10:02 am

arunsoorya1309 wrote:
Sat Aug 04, 2018 5:10 am
Anton has been informed about this, he will get back. Please don't spread false information about the issue.
As Ivanhoe was new, I did a lot of tests with different versions. And in these times, on my old 2.0GHz Quadcore-Notebook, all Ivanhoe-versions got around 4.5 Mn/s in the middlegame, running on 4 cores and 1.4 Mn/s on a single core. So, 2.5 Mn/s would fit perfectly to Ivanhoe, running on a faster and newer single core.
So, for me, that is the only possible explanation. If you have a better explanation, why Ivanhoe gets only 2.5Mn/s on 43-core machine, I would love to hear it.

User avatar
kranium
Posts: 1824
Joined: Thu May 29, 2008 8:43 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by kranium » Sat Aug 04, 2018 10:28 am

pohl4711 wrote:
Sat Aug 04, 2018 2:09 am

Then, everything is clear. In this TCEC, they give the uci-command “Threads=43“ to Ivanhoe, which is wrong (the uci-commands are shown, when an engine plays and you click on the small gearbox-icon).
And Ivanhoe doesnt know this command... so he plays in singlecore-mode! So, it is clear, why Ivanhoe calculates only 2.5 Mn/s in TCEC 13...
Correct uci-commands would have been:
TitanicMode=true
TitanicCPUSizings=43

What a ridiculous start of this season and that nobody checked the uci-commands, before starting the season. So sad. So unprofessional.

Stephan,

Please don't speculate wildly, create misleading and incorrect topics, and throw such harsh criticisms at TCEC. That's really hurtful and incredibly unfair to those working so hard behind the scenes (cudos! to all the unprofessional volunteers) to produce this incredible exciting competition for our enjoyment.

Indoda Eziphukile provided me with some info concerning the matter. Unfortunately, he can't defend himself (and he shouldn't have to!) from any of the accusations and criticisms that abound in this topic, because like Anton Mihailov, he registered several days ago for this forum but has not yet been activated.

They are using IvanHoe9.46h 01a35 x64 (which I suggested to Anton and provided)

If you look inside (this is on my 6 core w/HT PC) you'll see that it supports UCI option 'Threads':

IvanHoe9.46h 01a35 x64
WINDOWS capitalist mode in severance
Battle Mode
compiled with PREFETCH
cpu has POPCNT
Capitalist multicore mode compiled
RobboBases enabled by compile
RobboBaseLib Dynamic_Link
./RobboBaseLib.dll found (file name), to do dynamic load functions
TotalBaseCache is 1mb + (1mb)
TripleCacheSize is 1mb
TripleCacheSize is 1mb
RobboBaseLib Version 0.65c (dynamic)
CPUs Detected 12
Multicore Enabled.

uci
id name IvanHoe9.46h 01a35 x64
id author Yakov Petrovich Golyadkin, Igor Igorovich Igoronov, Roberto Pescatore, Yusuf Ralf Weisskopf, Ivan Skavinsky Skavar plus Decembrists (all)
id copyright Yakov Petrovich Golyadkin, 92th plus 93th plus 94th year from Revolution, PUBLICDOMAIN (workers)
id dedication To Vladimir Ilyich plus Zog 1 with Albania
option name Hash type spin min 1 max 65536 default 32
option name PawnsHash type spin min 1 max 1024 default 4
option name PawnsHashOneEighth type check default true
option name PVHash type spin min 1 max 1024 default 2
option name EvalHash type spin min 1 max 1048576 default 256
option name Threads type spin min 1 max 64 default 64
option name Ponder type check default false
etc.

here's a snippet from the TCEC Season 13 cutechess log file:

Finished game 2 (DeusX 1.0 vs Tucano 7.05): 1-0 {White wins by adjudication: TCEC win rule}
5195015 >Ivanhoe 999946h(2): uci
5195034 >Senpai 2.0(3): uci
5195039 <Ivanhoe 999946h(2): IvanHoe9.46h 01a35 x64
5195039 <Ivanhoe 999946h(2): WINDOWS capitalist mode in severance
5195039 <Ivanhoe 999946h(2): Battle Mode
5195040 <Ivanhoe 999946h(2): compiled with PREFETCH
5195040 <Ivanhoe 999946h(2): cpu has POPCNT
5195040 <Ivanhoe 999946h(2): Capitalist multicore mode compiled
5195040 <Ivanhoe 999946h(2): RobboBases enabled by compile
5195270 <Ivanhoe 999946h(2): RobboBaseLib Dynamic_Link
5195270 <Ivanhoe 999946h(2): ./RobboBaseLib.dll found (file name), to do dynamic load functions
5195272 <Ivanhoe 999946h(2): TotalBaseCache is 1mb + (1mb)
5195272 <Ivanhoe 999946h(2): TripleCacheSize is 1mb
5195272 <Ivanhoe 999946h(2): TripleCacheSize is 1mb
5195308 <Ivanhoe 999946h(2): RobboBaseLib Version 0.65c (dynamic)
5195308 <Ivanhoe 999946h(2): CPUs Detected 44
5195308 <Ivanhoe 999946h(2): Multicore Enabled.
5195309 <Ivanhoe 999946h(2): id name IvanHoe9.46h 01a35 x64
5195309 <Ivanhoe 999946h(2): id author Yakov Petrovich Golyadkin, Igor Igorovich Igoronov, Roberto Pescatore, Yusuf Ralf Weisskopf, Ivan Skavinsky Skavar plus Decembrists (all)
5195309 <Ivanhoe 999946h(2): id copyright Yakov Petrovich Golyadkin, 92th plus 93th plus 94th year from Revolution, PUBLICDOMAIN (workers)
5195309 <Ivanhoe 999946h(2): id dedication To Vladimir Ilyich plus Zog 1 with Albania

If the NPS reported on the TCEC web site doesn't seem right to you, I'm not surprised...it could be many factors. For ex: it's a fairly well-know fact that some info is interpreted and presented oddly by the highly customized TCEC version of cutechess.

It's should be a fun event, I really recommend you take a chill pill, relax, and enjoy the show!

pohl4711
Posts: 996
Joined: Sat Sep 03, 2011 5:25 am
Location: Berlin, Germany
Contact:

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by pohl4711 » Sat Aug 04, 2018 10:42 am

kranium wrote:
Sat Aug 04, 2018 10:28 am


If the NPS reported on the TCEC web site doesn't seem right to you, I'm not surprised...it could be many factors. For ex: it's a fairly well-know fact that some info is interpreted and presented oddly by the highly customized TCEC version of cutechess.

And how you explain, that Ivanhoe shows a search depth of only 19-22 plies in the early middlegame (download the played games so far and look for yourself), with queens still on board in this TCEC on 43-core machine? Perhaps cutechess cant display the depths correctly, too ???
You should know Ivanhoe/Robbolito better than anybody else and so you should know, that these depths should be reached in 1-3 seconds by Ivanhoe, if he would run with 43 threads.

For me, TCEC is dead, if the Ivanhoe-games will not be replayed with correct settings, because otherwise it is no competiton, but a bad joke.

Over and out. Hoping for a restart.

User avatar
kranium
Posts: 1824
Joined: Thu May 29, 2008 8:43 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by kranium » Sat Aug 04, 2018 11:04 am

As I said, "...it could be many factors"
and as the person who submitted the engine to TCEC for consideration, I don't consider it significant.

Don't forget this is a fun and free 'unofficial' competition for our entertainment.

I'm just not interested in speculating about it...I'll leave that to you.

I'm enjoying the competition immensely, especially the NN games!

Milos
Posts: 3348
Joined: Wed Nov 25, 2009 12:47 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by Milos » Sat Aug 04, 2018 11:56 am

kranium wrote:
Sat Aug 04, 2018 10:28 am
They are using IvanHoe9.46h 01a35 x64 (which I suggested to Anton and provided)

If you look inside (this is on my 6 core w/HT PC) you'll see that it supports UCI option 'Threads':

IvanHoe9.46h 01a35 x64
WINDOWS capitalist mode in severance
Battle Mode
compiled with PREFETCH
cpu has POPCNT
Capitalist multicore mode compiled
RobboBases enabled by compile
RobboBaseLib Dynamic_Link
./RobboBaseLib.dll found (file name), to do dynamic load functions
TotalBaseCache is 1mb + (1mb)
TripleCacheSize is 1mb
TripleCacheSize is 1mb
RobboBaseLib Version 0.65c (dynamic)
CPUs Detected 12
Multicore Enabled.

uci
id name IvanHoe9.46h 01a35 x64
id author Yakov Petrovich Golyadkin, Igor Igorovich Igoronov, Roberto Pescatore, Yusuf Ralf Weisskopf, Ivan Skavinsky Skavar plus Decembrists (all)
id copyright Yakov Petrovich Golyadkin, 92th plus 93th plus 94th year from Revolution, PUBLICDOMAIN (workers)
id dedication To Vladimir Ilyich plus Zog 1 with Albania
option name Hash type spin min 1 max 65536 default 32
option name PawnsHash type spin min 1 max 1024 default 4
option name PawnsHashOneEighth type check default true
option name PVHash type spin min 1 max 1024 default 2
option name EvalHash type spin min 1 max 1048576 default 256
option name Threads type spin min 1 max 64 default 64
option name Ponder type check default false
etc.

here's a snippet from the TCEC Season 13 cutechess log file:

Finished game 2 (DeusX 1.0 vs Tucano 7.05): 1-0 {White wins by adjudication: TCEC win rule}
5195015 >Ivanhoe 999946h(2): uci
5195034 >Senpai 2.0(3): uci
5195039 <Ivanhoe 999946h(2): IvanHoe9.46h 01a35 x64
5195039 <Ivanhoe 999946h(2): WINDOWS capitalist mode in severance
5195039 <Ivanhoe 999946h(2): Battle Mode
5195040 <Ivanhoe 999946h(2): compiled with PREFETCH
5195040 <Ivanhoe 999946h(2): cpu has POPCNT
5195040 <Ivanhoe 999946h(2): Capitalist multicore mode compiled
5195040 <Ivanhoe 999946h(2): RobboBases enabled by compile
5195270 <Ivanhoe 999946h(2): RobboBaseLib Dynamic_Link
5195270 <Ivanhoe 999946h(2): ./RobboBaseLib.dll found (file name), to do dynamic load functions
5195272 <Ivanhoe 999946h(2): TotalBaseCache is 1mb + (1mb)
5195272 <Ivanhoe 999946h(2): TripleCacheSize is 1mb
5195272 <Ivanhoe 999946h(2): TripleCacheSize is 1mb
5195308 <Ivanhoe 999946h(2): RobboBaseLib Version 0.65c (dynamic)
5195308 <Ivanhoe 999946h(2): CPUs Detected 44
5195308 <Ivanhoe 999946h(2): Multicore Enabled.
5195309 <Ivanhoe 999946h(2): id name IvanHoe9.46h 01a35 x64
5195309 <Ivanhoe 999946h(2): id author Yakov Petrovich Golyadkin, Igor Igorovich Igoronov, Roberto Pescatore, Yusuf Ralf Weisskopf, Ivan Skavinsky Skavar plus Decembrists (all)
5195309 <Ivanhoe 999946h(2): id copyright Yakov Petrovich Golyadkin, 92th plus 93th plus 94th year from Revolution, PUBLICDOMAIN (workers)
5195309 <Ivanhoe 999946h(2): id dedication To Vladimir Ilyich plus Zog 1 with Albania

If the NPS reported on the TCEC web site doesn't seem right to you, I'm not surprised...it could be many factors. For ex: it's a fairly well-know fact that some info is interpreted and presented oddly by the highly customized TCEC version of cutechess.

It's should be a fun event, I really recommend you take a chill pill, relax, and enjoy the show!
Official source code doesn't have option Threads but TitanicMode and TitanicCPUSizings. Threads was KLO's change. Without actual checks it is impossible to know if everything is working correct in the binary you provided. Judging by the output of cutechesscli, it wasn't. It is anyway not from KLO but someone's, who knows who's, compile.
I believe you had fair intentions when you provided the binary, but I am pretty sure that you wouldn't be reacting with "take a chill pill, relax, and enjoy the show" if Fire was running on a single core or some random compile that was behaving oddly instead of the one that you submitted.
This is a disgrace for TCEC and extreme unprofessionalism. And ofc it could never happen with Martin.
Problem with crippled Ivanhoe is that it enables both NN's to pass to division 3. Considering all the controversy already involved regarding DeusX "engine" it is hard not to be paranoid.

peter
Posts: 1748
Joined: Sat Feb 16, 2008 6:38 am
Full name: Peter Martan

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by peter » Sat Aug 04, 2018 12:03 pm

kranium wrote:
Sat Aug 04, 2018 11:04 am
As I said, "...it could be many factors"
But seeing these stored time to depth- values, as Stefan Pohl already pointed out, there's just one factor probable, Ivan ran single core, Norman, don't you really think so too?
Even if UCI command for number of threads is there, maybe they simply forgot to also activate multicore mode, or it's not correctly coded or compiled.
Last edited by peter on Sat Aug 04, 2018 12:17 pm, edited 1 time in total.
Peter.

User avatar
kranium
Posts: 1824
Joined: Thu May 29, 2008 8:43 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by kranium » Sat Aug 04, 2018 12:15 pm

peter wrote:
Sat Aug 04, 2018 12:03 pm
kranium wrote:
Sat Aug 04, 2018 11:04 am
As I said, "...it could be many factors"
But seeing these stored time to depth- values, as Stefan Pohl already pointed out, there's just one factor probable, Ivan ran single core, Norman, don't you really think so too?
Even if UCI command for number of threads is there, maybe they simply forgot to also activate multicore mode, or it's not correctly compiled.
from the log file above:
5195308 <Ivanhoe 999946h(2): Multicore Enabled.

User avatar
kranium
Posts: 1824
Joined: Thu May 29, 2008 8:43 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by kranium » Sat Aug 04, 2018 12:21 pm

Milos wrote:
Sat Aug 04, 2018 11:56 am
Problem with crippled Ivanhoe is that it enables both NN's to pass to division 3. Considering all the controversy already involved regarding DeusX "engine" it is hard not to be paranoid.
Whether IvanHoe plays at full strength or not, is not THE factor that decides if the 2 NN engines advance
There are 8 engines in the group
Ivanhoe is just one of the opponents they need to beat..

The 2 engines that play the best vs all engines will go thru
and if both NNs go thru..it's ok with me, I would actually enjoy that
Last edited by kranium on Sat Aug 04, 2018 12:28 pm, edited 1 time in total.

User avatar
Laskos
Posts: 9137
Joined: Wed Jul 26, 2006 8:21 pm
Full name: Kai Laskos

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by Laskos » Sat Aug 04, 2018 12:25 pm

kranium wrote:
Sat Aug 04, 2018 12:21 pm
Milos wrote:
Sat Aug 04, 2018 11:56 am
Problem with crippled Ivanhoe is that it enables both NN's to pass to division 3. Considering all the controversy already involved regarding DeusX "engine" it is hard not to be paranoid.
Whether IvanHoe plays at full strength or not, is not THE factor that decides if the 2 NN engines advance
There are 8 engines in the group
The 2 engines that play the best vs all engines will go thru
:lol:

User avatar
kranium
Posts: 1824
Joined: Thu May 29, 2008 8:43 am

Re: Ivanhoe plays singlecore in TCEC 13 !!

Post by kranium » Sat Aug 04, 2018 12:36 pm

I get the feeling that some here are angry about the possibility that Ivanhoe is not running as well as it should...
because they strongly do not want both NN engine to advance, perhaps hoping that 1 fails miserably?

If so, that's unfortunate...personally I'm rooting for both to go thru

Post Reply