geots wrote: Flip a coin..... Thing is, quality control is headed south of the border. It's a nice thought, and you can take your best shot at it- but it'll be quasi at best. Too many hands, too many different personalities and too much different hardware. Everything comes with a price.
Oh my gosh !
I would say that _your_ comments on this subject are a coin flip. Not that you are necessarily wrong, but given your background on fishtest, for you it is really like to flip a coin when you write a line about this stuff.
The "flip a coin" comment was just a tongue in cheek response that meant nothing. Marco, I have high respect for you and the Stockfish team. As Don said, this idea has been around a number of years- probably not with your refinements. I know that Don used something similar last year beta testing for Komodo 5, as he mentioned. I just think if he had considered it that successful- he would be using it now. (I was beta testing for him then, and am now as we speak.. But I did not sign on for that part last year) He mentioned that he was not trying to rain on your parade, and I certainly am not either. I wish you all the success in the world- and hope this works well for you and Tord. I would have no reason to feel any other way. I just have misgivings- which I stated. That along with Robert's thoughts just sum up my take on it. I certainly did not intend to offend anyone. I would add that people who mention this might be the wave of the future haven't thought it thru well. What it might be is the wave of the future for "open source engines". I can't imagine a commercial programmer near the top letting 30 or 40 people have access to his most private work.
At any rate- my apologies if I offended anyone on the Stockfish team. It was not my intention.
geots wrote: At any rate- my apologies if I offended anyone on the Stockfish team. It was not my intention.
George, you didn't offended anybody, of course. And don't need to apologize.
I agree idea of distributing jobs across machines is not new. It is really not new ! It was already used far before even Don was borne
But again, what amazes me is _not_ the technical achievement, but its openness. If Gary had setup such a framework, contacting privately me and some other guy, and keeping the UI access under a ssh connection (a private network) the technical result would have been exactly the same...but this would have _not_ be a milestone in my view.
geots wrote: At any rate- my apologies if I offended anyone on the Stockfish team. It was not my intention.
George, you didn't offended anybody, of course. And don't need to apologize.
I agree idea of distributing jobs across machines is not new. It is really not new ! It was already used far before even Don was borne
But again, what amazes me is _not_ the technical achievement, but its openness. If Gary had setup such a framework, contacting privately me and some other guy, and keeping the UI access under a ssh connection (a private network) the technical result would have been exactly the same...but this would have _not_ be a milestone in my view.
I think the fact that so many people are willing to join in and help is something you can be very proud of.
Marco, I am heading to get some sleep, and when I wake I imagine the match will not be far from over. On the XP quad (it does not run sse or pop) I started a 30 game match with Stockfish 2.2.2 x64 vs. Ivanhoe B46fE 0.2 x64, with both using 4 cores. The control is 40/20 repeating benched to CCRL standards and run at 40/10.
I will email you and probably give you a link to the complete PGN set. BUT IT IS GAME 2 you need to play thru very carefully, and take your time. There is a wealth of info in this game that spells out some particular vulnerabilities in Stockfish's game. I don't have to tell you that you learn the most from your losses. Only 3 games have been played- and I am headed to get some sleep. No idea how it will actually end up. But this PeterPan compile is the 2nd strongest Ivanhoe in existence at 64 bit. And he played a textbook game on how you attack the kingside- and Stockfish was too late understanding the trouble he was in. It reminded me of all the Fisher games I played thru where he opened up the rook file and pounded pawns down on his opponents' king.
I am not worth a lot at analysis- but this could be the most important loss you ever studied. There is room to make corrections here in his game.
Ok now I managed to install the client. It seems one just has to install the python packages
mongodb
requests
Not so complicated after all.
However http://54.235.120.254:6543/signup does nothing for me. It just brings up a login screen. Since I do not have a username/password yet it is useless to me.
Actually currently (since few days) it is still easier, you just don't need to install nor mongo nor requests. We have updated the readme to reflect that.
And regarding the readme there is a chapter that says:
Get username/password
Please e-mail us, and we will give you your username and password.
So please drop an email to Gary for the credentials. Sign up page it works...but only for administrators
We prefer to send directly the credentials to new users to avoid spamming / bots / etc..
The only thing I'd recommend is, please, do not run other stuff in background while testing. We prefer you dedicate your machine only a small time or only at limited intervals, but please, when you connect to fishtest framework don't do anything else with that machine.
So what should one do if one wants to take a machine offline. Would a partially completed work unit be lost or would it be assigned to another machine?
Michel wrote:So what should one do if one wants to take a machine offline. Would a partially completed work unit be lost or would it be assigned to another machine?
All the work is saved as games are completed, so it's safe to kill the worker at any time.
File "worker.py", line 35, in worker
update()
File "/home/vdbergh/SRC/CHESS/fishtest/worker/updater.py", line 32, in update
zip_file = ZipFile(worker_zip)
File "/usr/lib64/python2.6/zipfile.py", line 696, in __init__
self._GetContents()
File "/usr/lib64/python2.6/zipfile.py", line 716, in _GetContents
self._RealGetContents()
File "/usr/lib64/python2.6/zipfile.py", line 728, in _RealGetContents
raise BadZipfile, "File is not a zip file"
BadZipfile: File is not a zip file
