Path: ibmpcug!gwen.pcug.co.uk!dish.news.pipex.net!pipex!howland.reston.ans.net!vixen.cso.uiuc.edu!newsrelay.iastate.edu!dunix.drake.edu!acad.drake.edu!pk6811s From: pk6811s@acad.drake.edu Newsgroups: rec.games.corewar Subject: Re: Swing source (Rude Wind II incl.) Date: 18 Jul 95 10:37:12 CST Organization: Drake University, Des Moines, Iowa Lines: 128 Distribution: world Message-ID: <1995Jul18.103712@acad.drake.edu> References: <1995Jul14.162510.2805@rhodes> NNTP-Posting-Host: acad.drake.edu (Randy Graham) writes: ... interesting notes on Swing's development Withershins scans forward while Swing scans backward using the same step size, so placement in core and relative size determine the outcome. Your djn.f stream successfully decoys Withershins and others, though djn.b would work better - probably would give better results against stone-based programs too since your sweep would be triggered sooner. > > W L T > Versus: === === === > Withershins Thrice 58 42 0 > Thermite 1.0 55 32 13 > quiz 55 30 15 > Rude Wind II 31 60 9 (this one surprised me. I want source ;-) Source follows. Notice that Rude Wind II scans backward using the same step size (12), but the scanning component is only 10 lines so you will miss it and start your attack further back. That gives RW a little more time to find you, as your program builds processes and starts up the sweep routine. > Agony II 17 77 6 (same problem as Paul here. Sigh...) Good grief! We'll _never_ push that guy off this way. Please kill off Swing immediately! :-) > > And source (minus EQU lines) with apologies to Paul for stealing his > multi-pass core clear from Withershins Thrice. I still have some It's only a component. 'sides I'm working on a better one now. When I saw your note about an 83.3% scanner I couldn't figure that out either. The only thing I came up with was 10 scans in a 12 instruction loop :-) Then I had a moment of panic when I saw you are using the same step size as Withershins and Rude Wind. Look out for some smart guy using a 12-instruction reflection! But you are using SNE.I and I am using SNE.X, so that should prevent a problem (for one of us :-) Some people may have picked up on the fact that this form of multi-pass sweep can be implemented in decrement mode under '88 rules. At least some of the recent comments on the Standard and ICWS Hills indicates people are trying multipass clears there. Great! Paul Kline pk6811s@acad.drake.edu ;redcode-94m quiet ;name Rude Wind II ;kill Rude Wind ;author P.Kline ;strategy once-thru bomb/scan with anti-imp attack and paper ;strategy II: 80% bomb/scan, changed to Silk-style paper ;macro incr equ 513 spacing equ 4375 step equ 12 pl1 add.ab scan,1 spl #0,#scan+5334+step+(step/2) mov p2b,>-1 ; anti-imp attack, part 1 djn -1,#3 spl 1,>200 ; spl 1,>400 ; create 7 processes mov -1,0 ; mov spacing ; 1st paper mov }p1s,>p1s mov p1b,>p1s mov p1b,>p1s+7-7*incr add #incr,@-1+spacing jmn.f @0,{p1s p1b dat <5334,<2667 for 19 dat 0,0 rof pl2 add.ab scan,1 spl #0,#scan+2667+step+(step/2) mov p1b,>-1 ; anti-imp attack, part 2 djn -1,#3 spl 1,<-100 ; spl 1,<-300 ; create 7 processes mov -1,0 mov spacing ; 2nd paper mov }p2s,>p2s mov p2b,>p2s mov p2b,>p2s+7-7*incr add #incr,@-1+spacing jmn.f @0,{p2s p2b dat <5334,<2667 for 20 dat 0,0 rof inc dat 4*step,4*step bomb dat 2*step,2*step for 16 dat 0,0 rof next sub inc,scan ; scanner bombs two and scans two locations mov bomb,@scan ; in a five-instruction loop mov bomb,*scan scan sne.x *pl1-(5*step),@pl1-(4*step) djn next,#7900/(step*4) spl #0,>scan mov p2b,>scan djn -1,#1 attack spl pl2,<3500 ; start launch code jmp pl1 ; end next+1 Paul Kline pk6811s@acad.drake.edu