Octochess r4690 released

[geots]

Hi Tim,
hi Jim,
thanks for the updated engine and the compiles.
I now have the problem, which engine we should test:
Octochess r4690 or
Octochess r4692 JA
... what are the differences?

best wishes
Werner

Here is log of differences between the revisions.

Code: Select all

4692:  Fix problem with not ending a transaction.
4691:  Inside comments there can be [
4690:  Release of r4690

I don't think they affect playing strength.
I would use Tim's builds as it is official release unless you have a legacy system that won't run his versions.
Tim's builds require minimum 64-bit Vista/Win7. My builds will run older systems too such as Win98/WinXP 32 bit.

Jim.

Jim, glad I found you in this discussion. As for Octochess r4692, I am using your 64 bit compile. I see where CCRL shows games run with 64bit on 1CPU. When I load it, it has a choice of 1-4 cores. But no matter what I set it on, it is stuck on 2 cores, as in Task Mgr. it shows usage is 50- which is half of the 4 I have- no matter what I set it on. I know this is sort of an old thread, but maybe you or Tim will see it and offer some help. Denis Mendoza and Adam Hair have an maybe still are trying to help me, but we have come up with nothing yet.

Hope you see this thread.



Best,

george

[Graham Banks]

I see where CCRL shows games run with 64bit on 1CPU

Those are likely to be Octochess r4690 games, as we're combining the results.

[geots]

If its not DiscoCheck 3.6.2 and its 64bit exe, or Octochess sticking on 2 cores, it is Sloppy 0.2.3 64bit wont show the right hash, and wont access egbb when the previous version 0.2.0, does it all perfectly.

I wanted to load at least 1 version of each engine and run some of them, but damn if I will spend 3 days on one Sloppy version. Maybe it is just me, but sometimes authors say they didn't have a way to check what it would do in chessbase guis. But they could have easily sent it to CCRL and asked them to check it real quick. Just to see if it obeys the config or ini files. Instead some just put it out unchecked and hope for the best- at least it sure seems that way. And the lower down the list you go- the worse it gets.

The only thing I know to do is throw out Octochess and Sloppy and Scorpio and Dirty can run 64bit with no egbb access. Not my problem. Since I know how to load them right- if they don't work the first time- it's goodbye to the recycle bin and let someone else worry with them.

I'm not trying to be ugly- but I wanted to give them some exposure. But that doesn't include beta testing for the programmers.


gts

[Codesquid]

Jim, glad I found you in this discussion. As for Octochess r4692, I am using your 64 bit compile. I see where CCRL shows games run with 64bit on 1CPU. When I load it, it has a choice of 1-4 cores. But no matter what I set it on, it is stuck on 2 cores, as in Task Mgr. it shows usage is 50- which is half of the 4 I have- no matter what I set it on. I know this is sort of an old thread, but maybe you or Tim will see it and offer some help. Denis Mendoza and Adam Hair have an maybe still are trying to help me, but we have come up with nothing yet.

george, I just tested it myself. The official Octochess builds do not have this problems, it is only Jim's builds that have this problem.

As far as I can tell, it really is only a single thread playing chess, as desired. What is consuming the extra CPU time is the management thread. It's designed to wake up only when the allotted move time has expired or if the search has finished. It's this waiting that probably does not work here, resulting in constant polling.

@Jim: Please try the following code:

Code: Select all

mutex m;
condition c;
timestamp start;
scoped_lock l(m);
c.wait(l, 10000);
timestamp stop;
std::cerr << "Elapsed: " << (stop-start).milliseconds() << " ms " << std::endl;

It should sleep for 10 seconds.

[geots]

Jim, glad I found you in this discussion. As for Octochess r4692, I am using your 64 bit compile. I see where CCRL shows games run with 64bit on 1CPU. When I load it, it has a choice of 1-4 cores. But no matter what I set it on, it is stuck on 2 cores, as in Task Mgr. it shows usage is 50- which is half of the 4 I have- no matter what I set it on. I know this is sort of an old thread, but maybe you or Tim will see it and offer some help. Denis Mendoza and Adam Hair have an maybe still are trying to help me, but we have come up with nothing yet.

george, I just tested it myself. The official Octochess builds do not have this problems, it is only Jim's builds that have this problem.

As far as I can tell, it really is only a single thread playing chess, as desired. What is consuming the extra CPU time is the management thread. It's designed to wake up only when the allotted move time has expired or if the search has finished. It's this waiting that probably does not work here, resulting in constant polling.

@Jim: Please try the following code:

Code: Select all

mutex m;
condition c;
timestamp start;
scoped_lock l(m);
c.wait(l, 10000);
timestamp stop;
std::cerr << "Elapsed: " << (stop-start).milliseconds() << " ms " << std::endl;

It should sleep for 10 seconds.

Tim, thing is it is showing it is using 50% of my available cores. You are right when you say that may not be true, but I don't have anything else to go by. I was thinking that I tried to download from your site- but it directed me to Ablett's (whose work by the way is usually the best). But that may have been another engine. In case I can't find a link to your r4692 64-bit, would you mind putting a link here. I really hate to have to pass over this engine. I want to use it. And thank you for your response.


Best,

george

[geots]

I just went to your site and I remember the problem now. I cannot find any r4692 64-bit to download. You don't go any further than 4690 that I can find.


george

[Codesquid]

I just went to your site and I remember the problem now. I cannot find any r4692 64-bit to download. You don't go any further than 4690 that I can find.


george

r4692 has never been officially released. In playing strength, r4690 and r4692 are absolutely identical, only difference is in utility code not relevant for play.

[geots]

I just went to your site and I remember the problem now. I cannot find any r4692 64-bit to download. You don't go any further than 4690 that I can find.


george

r4692 has never been officially released. In playing strength, r4690 and r4692 are absolutely identical, only difference is in utility code not relevant for play.

Well that solves all the problems with it, then. I will get r4690 and "let 'er rip"

Thanks for taking the time to help me Tim. I really do have an interest in testing your engine.


Best,

george

[geots]

I just went to your site and I remember the problem now. I cannot find any r4692 64-bit to download. You don't go any further than 4690 that I can find.


george

r4692 has never been officially released. In playing strength, r4690 and r4692 are absolutely identical, only difference is in utility code not relevant for play.

Tim, I got r4690 and no problems with the threads. Works fine and thank you. I would like to mention that I did not use the "generic" exe, but rather I used the sse4 exe. I hope that is the correct one.

I do have one question. I want to be sure. There are not many, but I have seen a couple 32bit exes that use SSE4, Rybka being one- so I am asking if your sse4 is 64 bit or not. Like I said, I certainly would guess it was- I just want to double check it with you.

Heading to get some sleep and I will check back tonight. Thanks again for your time.


All the best,

george