Date: Fri, 7 Feb 1997 02:44:36 -0500 Message-Id: Errors-To: tuc@stormking.com Reply-To: franz@azstarnet.com Originator: corewar-l@stormking.com Sender: corewar-l@stormking.com Precedence: bulk From: franz@azstarnet.com (George Lebl) To: Multiple recipients of list Subject: Re: Core Warrior 55 X-Listprocessor-Version: 6.0b -- ListProcessor by Anastasios Kotsikonas X-Comment: Usenet News "rec.games.corewar" >> I've been working on several ideas but they just haven't worked out. One >> was a multi-process scanner which starts out as a single process and >> splits to bombing code when something is found. The bombing code sets up > >I've got a (hopefully) working self-splitting scanner mostly done. It's >a .4c cmp scanner + .2c continuous clear, somewhat inspired by Silver >Talon's design. I still haven't managed to get it switch to the clear >well, (the current version is 14 instructions long, and very fragile) >and I can't currently access the only computer I can run pmars on. :( ok .. since i just can't get it working ... this is the version that made 58 on the hill and needs huge amounts of work ;redcode-94 ;name SUB/JMN Prototype ;author Franz ;strategy A new type of scan, cca .5c sub/sub/jmn single shot ;strategy which is self-splitting. Bad thing about this routine ;strategy is it's a b (or a) scanner only. though it's fairly small ;strategy this is the first prototype of this scan/bomb routine ;strategy this will probably fail too :) stsc equ 3 sstep equ 3510 org wh ptr spl #1,#ch+300 bomb spl #2,#ch+100 bomb2 dat #0,#ch+50 for 32 dat 0,0 rof wh spl #stsc,#stsc+5 mov.i *ptr,>ptr jmz.a -1,@ch loop add.f @ch, wh sub.b @wh, ch sub.b *wh, ch ch jmn.a loop, step mov.ab wh, ptr for 51 dat 0,0 rof step dat #sstep,#sstep datz end it's a simple one shot .. using the trick of using subs to move the value of the scanned area to the jmn which in case of zero points to step and doesn't fall through ... once it gets changed a bit it will most likely point to a zero field and all processes fall through ... this was the trouble i allways had .. ALSO the scanned pointer DOES NOT get updated once the jmn has been changed so leftover proccesses on add don't change this .... this is a bad prototype .. the clear is slow and buggy, yet small ... I'm working on a much enhanced (1 instruction longer) version which is slightly faster and does faster better clear ... but it doesn't work yet .. have fun with this ... the only problem this little thing faces is fairly slow reaction and not much other aplications then a single shot ... unless you would reupdate the jmn with a correct value ... this is something i will try but i feel it will take too many instructions, since the thing is .5c right now i don't think making it longer would help it much ... Franz -- -------------------------------------------------------------------------------- Sex Pistols RULE!!! franz@azstarnet.com