Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Singular Extension Crafties

Author: Dann Corbit

Date: 10:07:09 09/10/01

Go up one level in this thread


On September 09, 2001 at 09:55:04, Robert Hyatt wrote:
>On September 09, 2001 at 01:49:25, Slater Wold wrote:
>>On September 08, 2001 at 23:42:00, Robert Hyatt wrote:
>>>On September 08, 2001 at 17:42:44, Slater Wold wrote:
>>>>On September 08, 2001 at 09:42:42, Robert Hyatt wrote:
>>>>>On September 08, 2001 at 06:40:41, Gian-Carlo Pascutto wrote:
>>>>>>On September 08, 2001 at 06:35:24, Dann Corbit wrote:
>>>>>>>The ones with BM in the name are modelled after Bruce Moreland ideas. From my
>>>>>>>testing, these versions do better than any of the others (including crafty
>>>>>>>without extensions).
>>>>>>
>>>>>>What tests did you run?

Only tactical tests.  I have no idea if it plays better in actual games.  I
don't remember what all the scores were, but I do remember WAC, which got
295/300 (and crafty without the Bruce Moreland extensions gets 293 on my
machine).

>>>>>>What did you change?

Not sure what all I had to change.  Here is the source code:
ftp://cap.connx.com/pub/chess-engines/new-approach/C18_10.ZIP

You can change the binary that is created by changing the macro as follows:

#define DO_BM_EXTENSIONS
does Bruce Moreland ideas

#define DO_EXTENSIONS
does the regular extension ideas

Do not define both macros.

If neither macro is defined, you get the regular crafty.

>>>>>>How did you get SMP to work?

Well, it compiles, links, and runs -- but I have not tried it on a multiple CPU
machine, so maybe it does not work with SMP.

>>>>>>--
>>>>>>GCP
>>>>>
>>>>>
>>>>>He might not have noticed that "searchmp" has to be modified to make this
>>>>>work reliably.  The last time I fiddled with this stuff, all the changes
>>>>>were in search.c...  this means that if the program splits the tree into
>>>>>parallel pieces at a ply that should be "singular" then it will behave
>>>>>differently than it should...
>>>>>
>>>>>Dann:
>>>>>
>>>>>If you are convinced that this is really working reasonably on a single-cpu
>>>>>machine, let me know and let me know where the source is located.  I will see
>>>>>if I can fix it up so that it will work correctly in a parallel machine as
>>>>>well...

Here is the source:
ftp://cap.connx.com/pub/chess-engines/new-approach/C18_10.ZIP

>>>>I would _LOVE_ to see an SMP Crafty with SE.  I e-mailed you about it over a
>>>>week ago, and didn't get a response.  :(

I did not have a working SMP version one week ago.

>>>>
>>>>Slate
>>>
>>>
>>>If I didn't respond, I didn't receive the email.  I generally respond to an
>>>email within 24 hours at worst case.
>>>
>>>We have occasional email glitches (spool fills up.  Student hacked into the
>>>machine and removed entire mail queue a couple of weeks ago.  Etc).  If I
>>>don't respond quickly, it is always worth another query.
>>
>>Quick quesiton:
>>
>>Is hyatt@crafty.cis.uab.edu correct?  That's what's listed on your CCC profile.
>>But on ICC it's hyatt@cis.uab.edu.  (And that's what you responded with.)
>>
>>I take it hyatt@cis.uab.edu is the correct one.
>>
>>Thanks again.
>>
>>
>>Slate
>
>
>hyatt@cis.uab.edu is better.  If you send it directly to crafty.cis.uab.edu,
>then mail can be lost if I am working on the machine.  IE when I upgrade to a
>new version of linux, for a couple of hours there is no user "hyatt" on that
>machine, which will cause trouble.



This page took 0.01 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.