I did not have time to check the last version thoroughly, I only let HIARCS and Zappa repeat their match and they were using all 4 threads again when set to 2.
Hopefully I will have more time this evening to take a closer look.
Best,
Hi Gabor,
did you really download ChessGUI again before testing ?
I would be very surprised.
Best,
Matthias.
SzG wrote:
Hmm, you don't trust me... date of the exe is May 18 22:55. But as stated above, I was in a hurry so a mistake may have occurred*. I can tell you more in 3 hour's time.
*Most probably, I did not delete the ChsGUI.uci files from the engine directories before running the match.
Cheers,
May 18 22:55 is correct. Deleting the ChsGUI.uci files is not necessary.
Still surprised ,
Matthias.
SzG wrote:
I regret to report it but still no success. HIARCS is using all 4 threads whatever I do. So does Zappa.
In the ucilog.txt I sent you I find the following 2 lines:
Sun May 18 22:02:19 2008: from HIARCS 12 2CPU (0): option name Core Threads type spin min 1 max 4 default 4
and
Sun May 18 22:02:20 2008: to HIARCS 12 2CPU (0): setoption name Core Threads value 2
Is this not the key to the problem? I find the first line in ChsGUI.uci for HIARCS but I do not find the second one.
One more thing: when the ChsGUI.uci file HIARCS is created, it contains this line:
option name Core Threads type spin min 1 max 4 default 4
After playing a game and changing the UCI options, the line changes to:
option name Core Threads type spin default 2 min 1 max 4
As you can see, the order of parameters has changed. I edited the second line to the previous format but no success with that either.
"setoption name Core Threads value 2" is the correct command that was not being sent. Nothing else matters. It is all that the engine needs.
The place to look is in ChessGUI debug file.
"As you can see, the order of parameters has changed." That is OK.
Who else out there has many cores and mp engines to help solve this mystery ? Just find out if your mp engines use only the limited number of cores you define in ChessGUI 0.095 !
Matthias Gemuh wrote:
Who else out there has many cores and mp engines to help solve this mystery ? Just find out if your mp engines use only the limited number of cores you define in ChessGUI 0.095 !
Best,
Matthias.
Just what I wanted to ask. Note that I only have this problem with HIARCS 12 and Zappa Mexico II.
ChessGUI 0.095 updated. No fix yet for CPU threads issue.
Matthias Gemuh wrote:
Yet another minor bugfix : if an engine lost on time, ChessGUI could mess things up by writing pgn score as "time" instead of "1-0" or "0-1".
Matthias.
ChessGUI had ceased using GUI remis book correctly for new tourneys.
Fixed.