Article View: rec.games.corewar
Article #216Re: Corewars on a PC
From: Stefan Strack
Date: Thu, 12 Nov 1992 20:45
Date: Thu, 12 Nov 1992 20:45
38 lines
2663 bytes
2663 bytes
In article <1992Nov9.223928.24097@cs.kuleuven.ac.be> vwulpen@cs.kuleuven.ac.be (Henk Van Wulpen) writes: >[Speaking about CoreWar Pro] >The only problem with it is that it is incredibly slow, especially when >lots of processes are SPLit off. I believe it is because it was written >in Prolog, and therefore handles linked lists (the process-queues) in a >very inefficient way. Actually, I already got an error message (from the >program) saying that a system error occured. I think it is because the >program ran out of memory trying to create a new item in the queue. >Possibly the program doesn't deallocate the space of killed processes. >The same problem occurs with KotH for PC. As long as the number of rounds >played is reasonably low and there aren't many SPLs, the program runs ok. >But with too many rounds or process-creation, the program halts saying that >it ran out of memory creating a new thread. Since increasing the number of >rounds should have no effect on the use of memory, I guess that also here the >problem is that memory isn't freed upon removal of items in the queue. >Just my opinions though. Any comments of the authors ??? > >Henk. As the author of CoreWar Pro and the "porter" of KotHPC I oughta reply to this. And yes, my name and email address is listed prominently in the doc's to both packages. Anyway, Henk is right, CoreWar Pro is slow because it is written in Prolog (PDC Prolog, former Turbo Prolog to be exact). No remedy for that unless I get the funds for the lastest version of the compiler (3.3) which has an array datatype (makes for a much more efficient implemetation of the process queue). As for KotHPC, the reason that the programs bombs earlier when "rounds" is high is that the free() in Turbo C 1.5/2.0 is not 100% efficient. Hope this helps, Stefan (stst@vuse.vanderbilt.edu) P.S.: Those in need of a speedy, good-looking PC version of corewar that has reasonable debugging features should check out Nandor Sieben's MARS. I had the pleasure to try out an alpha-version of V2.0 and it looks good. I don't know if Nandor uploaded the latest version to soda yet. Nandor?
Technical Details
Message-ID:
<BxMEz9.Ku8@vuse.vanderbilt.edu>Path:
arc1-ca.pugleaf.net!sparky!uunet!vuse.vanderbilt.edu!ststReferences:
<mq6TTB1w165w@cerianthus.pinetree.org> <&}#@byu.edu> <1992Nov9.223928.24097@cs.kuleuven.ac.be>Show Full Headers
Complete Article Headers
Newsgroups: rec.games.corewar Path: arc1-ca.pugleaf.net!sparky!uunet!vuse.vanderbilt.edu!stst From: stst@vuse.vanderbilt.edu (Stefan Strack) Subject: Re: Corewars on a PC Message-ID: <BxMEz9.Ku8@vuse.vanderbilt.edu> Sender: news@vuse.vanderbilt.edu Nntp-Posting-Host: ohfs Organization: Vanderbilt University School of Engineering, Nashville, TN, USA References: <mq6TTB1w165w@cerianthus.pinetree.org> <&}#@byu.edu> <1992Nov9.223928.24097@cs.kuleuven.ac.be> Date: Thu, 12 Nov 1992 20:45:09 GMT Lines: 39