home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: rec.games.corewar
- Path: sparky!uunet!cs.utexas.edu!uwm.edu!ux1.cso.uiuc.edu!news.iastate.edu!dunix.drake.edu!acad.drake.edu!pk6811s
- From: pk6811s@acad.drake.edu
- Subject: Outlaw ImpRings? Nah!
- Message-ID: <1993Jan3.211034.1@acad.drake.edu>
- Lines: 51
- Sender: news@dunix.drake.edu (USENET News System)
- Nntp-Posting-Host: acad.drake.edu
- Organization: Drake University, Des Moines, Iowa, USA
- References: <1hukq1INNroe@MINERVA.CIS.YALE.EDU> <1993Jan2.170839.4347@vlsi.polymtl.ca> <1i610tINN9uj@MINERVA.CIS.YALE.EDU>
- Date: Mon, 4 Jan 1993 03:10:34 GMT
-
- >
- >If an impring had more points, wouldn't it be easier to kill?
- >
-
- I have submitted several variations on seven-point imps with no success.
- All the existing imps are three-pointers (yes?)
-
- It's true that an imp-ring/spiral has only one vulnerable point at any
- one time. That point may not even be the one about to execute, since
- multiple processes can be laid down over the same spiral and the loss
- of one does not affect the execution of the others who pick up the slack.
-
- The real vulnerable point of the imp is the most trailing point, which
- is always in the same stream (until killed), and it is never protected.
- It can be killed by a dat bomb, thus shortening the spiral by one.
- Unfortunately the chances of killing even a short spiral by bombing the
- trailing points one-by-one before you are overrun is minimal.
- Stunners, scanners, and Vampires have better luck by stunning the
- trailing point with a spl or spl-jmp bomb. The stone in Imprimis
- has some success by using an increment of dat #2667,nnnn; by decrementing
- locations at 2667 intervals it can run down a slow-moving imp.
-
- Gating is an effective defense against imps, but tends to make your
- program one or two instructions longer and therefore more vulnerable.
- Charon v8.0 does not use a gate but still wins the majority of its
- battles with imps. The secret is extended scanning with an increment
- that scans N and N+2666/2667/2668 (pick one) within a few hundred
- cycles of each other. You do have to be careful to allow enough time
- for core-clear and for the opponent's processes to all die, however.
-
- I still believe what I said before, the modern imps are just stones with
- a backup plan. The stone probably does most of the killing and the imp
- gives the tie in the rest of the cases. The win percentages for Emerald
- and Imprimis are nearly the same - one is a stone and the other is
- a stone-imp. Traditionally a replicator kills stones with no problem,
- so another approach would be a replicator that can somehow kill imps.
-
- It is important to protect as many of your instructions from decrements
- as possible, since half the bombs of a mov <a,b stone are single decrements.
-
- As for using some other core size, remember that a three-point imp
- does not have to use (N+1)/3 as the increment. (N*i+1)/3 works just
- as well. And changing the core size will affect ALL programs that
- depend on optimal bombing/scanning patterns.
-
- Imps are here to stay, just like stone, paper, and scissors. We don't
- need to outlaw them, just learn to live with them (or in spite of them!)
- They do make life a little more complicated.
-
- Paul Kline
- pk6811s@acad.drake.edu
-