Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Distributed perft project

Author: Michael Yee

Date: 11:37:16 12/11/03

Go up one level in this thread


On December 11, 2003 at 09:16:15, Albert Bertilsson wrote:

>On December 11, 2003 at 09:00:43, Michael Yee wrote:
>
>>On December 11, 2003 at 08:29:22, Albert Bertilsson wrote:
>>
>>>On December 11, 2003 at 08:19:02, Michael Yee wrote:
>>>
>>>>On December 11, 2003 at 08:11:22, Albert Bertilsson wrote:
>>>>
>>>>>On December 11, 2003 at 07:56:40, Jaime Benito de Valle Ruiz wrote:
>>>>>
>>>>>>On December 11, 2003 at 07:08:30, Albert Bertilsson wrote:
>>>>>>
>>>>>>>Hi!
>>>>>>>
>>>>>>>If you have any wasted cpu-resources join my distributed perft project to
>>>>>>>calculate perft(11). Any contribution large or small is welcome.
>>>>>>>
>>>>>>>The project is currently under testing to see how many will join (I'll not do it
>>>>>>>myself) and to make sure that the server is working ok on the internet, I've
>>>>>>>just tested it on LAN before.
>>>>>>>
>>>>>>>You need .Net framework from Microsoft to participate.
>>>>>>>
>>>>>>>Home page:
>>>>>>>http://www.albert.nu/programs/dperft
>>>>>>>
>>>>>>>/Regards Albert
>>>>>>
>>>>>>
>>>>>>Hi,
>>>>>>
>>>>>> I'm also curious about this "waste of time" as well, and I often leave my
>>>>>>computer downloading overnight, so I guess I wouldn't mind giving you some
>>>>>>figures.
>>>>>> I'm not that happy about running this "DistributedPerftClient.zip" in my
>>>>>>computer, though.
>>>>>> The ratio logarithm of depth to time is roughly linear, at least up to depth 8,
>>>>>>so if you give me a FEN position, and a reasonable depth to calculate , I could
>>>>>>come back to you with a figure in about 24 hours.
>>>>>> I estimate that I need at least 2 hours for a perft 8... and maybe much more as
>>>>>>the game develops (a perft 9 wouldn't let me use my computer for days). I have
>>>>>>an Athlon 1800+, and I use Movei because it is quite fast for perfts. If you
>>>>>>know a faster one, let me know, because otherwise the "four" computers that you
>>>>>>have used for your figure are about 4 or 5 times faster than mine!!!
>>>>>> Regards,
>>>>>>
>>>>>>  Jaime
>>>>>
>>>>>Although it would be possible for me to enter node counts into the database
>>>>>manually it would make the work much more tiresome. There are 8902 perft(8) in
>>>>>one perft(11) makeing manual updates not to mention setting the fen and sending
>>>>>the answer by mail quite a work.
>>>>>
>>>>>What the distribute perft client does is to recieve fen's from the server,
>>>>>calculate the nodes and send back the answer, and then get next problem. If you
>>>>>don't trust me writing a good program or doing nasty stuff I can garantee you
>>>>>that I've meant to do no such thing. If you don't want to use .Net or don't use
>>>>>windows at all you can't help me. And I can't help you because it would be to
>>>>>much work to make software for all platforms, I used .Net because it was the
>>>>>easiest for me to program.
>>>>>
>>>>>/Regards Albert
>>>>
>>>>Hi Albert,
>>>>
>>>>Have you tried Mono (www.go-mono.com) or DotGNU's pnet (www.dotgnu.org)?
>>>>Depending on how much networking stuff they have implemented, they might enable
>>>>your client to run on linux or other OSes.
>>>>
>>>>Also, I wonder if your SharpChess (and future .NET chess GUI) can run without
>>>>modifications... Then you can claim/advertise that they work under linux too.
>>>>
>>>>Michael
>>>
>>>I knew it existed but I don't know anything more about it (how far they have
>>>implemented stuff). I hope to get some time this christmas to play with linux
>>>and try Mono. I guess that #Chess could work because it is very simple and uses
>>>only some simple file I/O. I don't know about the distributed perft client
>>>because it uses .Net Remoting but if you have Mono I'd be happy to know if it
>>>works, or if it could with some minor modification.
>>>
>>>/Regards Albert
>>
>>
>>I'll give it a try later today. According to Mono's status pages,
>>System.Runtime.Remoting seems about 100% complete, so the prospects look good.
>>
>>Michael
>
>If Mono has got that far I'd be very impressed, I'd better have a look at it
>soon. Maybe soon I'll be teaching C# on Mono soon :)
>
>/Regards Albert

Hi Albert,

I guess Mono hasn't gotten that far yet! In my initial tests, I couldn't run the
client on linux (or windows) with mono. Part of the reason seems to be some
unfinished Windows.Forms stuff.

However, SharpChess 0.06 "almost" runs with mono under linux. I can at least run
perft, but "go" causes an exception. In case you're curious, here's a small
comparison:

Athlon 1.2GHz, Windows XP, Microsoft .NET

perft 4 -->  0.25s
perft 5 -->  6.06s

Athlon 1.2GHz, Linux (RH9), Mono 0.29

perft 4 -->  0.54s
perft 5 --> 13.02s

Michael




This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.