[00:23] MSG: Quit: Lost terminal [01:04] Join: Caelian__ joined #corewars [01:08] MSG: Ping timeout: 240 seconds [01:38] After some thinking, I'm clueless about how to deal with papers on nano hill :P No code but a clue would be appreciated :) [01:39] (I wrote a nano paper and trying to get a nano scissor :P) [01:52] also is there a way to challenge a hill without getting on it whatever are the results? [02:26] Nick Change: Dirty|afk changed nick to Luka [02:26] Nick Change: Luka changed nick to Luka|Zzz [02:33] Join: CoreXii joined #corewars [02:33] so, what, this is a Core War channel? [02:34] anyone alive in here? [02:38] asw Caelian__ Fluffy johnkw log_guy Mizcu pak21 triple [03:02] Join: Leniad joined #corewars [03:08] if anyone has time later to look at my first warrior and completely humilliate me by pointing it's probably obvious flaws: http://pastebin.ca/494160 [03:09] thanks :D [03:09] it's your lucky day, I'm playing around with Core War as we speak [03:10] pretty big for a first warrior [03:10] well. most is decoy? :P [03:11] I just added stuff as I was reading it [03:11] no copy pasting though, I tried to write it from scratch [03:12] first warrior... damn, not bad, assuming it works [03:13] it scores 99 at wilkies [03:13] I'm not that much of a pro I could point out many flaws in it :P [03:13] in fact, maybe you could help ME with something? I tried to ask these guys but... they're just idling [03:13] sure, I can try [03:14] I think I picked this up fast because I program on real assembler... you know, the one that is actually useful ;) [03:14] http://pastebin.ca/494176 [03:14] this piece of my warrior, for some reason, goes all wrong after a small while [03:15] if I recall correctly, the spl somehow manages to get two processes beyond the door... see I'm trying to let only one pass [03:15] (start at Trap) [03:18] well [03:18] I didnt run it yet [03:18] but doesn't mov door,0 move the door to the mov? [03:18] then the next process from the spl [03:18] will run jmp -1 [03:18] exactly [03:18] k, so that is the idea [03:18] so only the first process to land on the mov Door, 0 will get past [03:18] let me run it then [03:19] I'm not sure if it works because Clear is elsewhere.. [03:19] take out the slt if it skips it [03:20] yes, it wont compile because I dont have clear [03:20] (it is to prevent it from core-clearing itself) [03:22] so yea, doesn't work quite right [03:23] starts bombing at intervals of 2 and executing the dats for some reason [03:23] number of processes goes 5, 6, 7, 6 and back to 5, then repeat [03:24] the way it should work is have the spl make new processes infinitely while only one - the first one - of them starts clearing [03:24] i'm still not sure what the slt is there for [03:25] the slt is to prevent it from overwriting itself [03:25] It will skip the jump if the next address to bomb is bigger than the distance from clear to tclear [03:26] so I'm guessing clear is somewhere before the door? [03:26] yes [03:27] but mind not that, that's not the point [03:27] the thing is that my door doesn't work, somehow more than 1 process gets through the door [03:28] and I can't figure why [03:28] I was debugging it step by step, every turn watching where all the processes are.... and then, suddenly, out of nowhere, there's another process beyond the door [03:29] ahhhhhhhh I see why [03:29] the > is indirect adressing [03:29] and it points to the b field [03:30] so the spl actually splits to what it's on it's b field [03:30] oh I see, yes that makes sense then... that's why the new process pops beyond the door [03:31] is there any reason you want to increase the spl? [03:31] maybe count something? [03:31] yes, check if there are any processes on the spl [03:31] well [03:31] you could write spl #0, >0 [03:32] won't that go into spl #0, >1 then? and next turn it doesn't post-increment its own B-value anymore, but rather the next instruction's? [03:32] true [03:32] hmm [03:33] spl #0, }0 should work [03:33] goes to spl #n, }0 then doesn't it? (where n is the number of processes passed) [03:33] yes [03:33] well [03:34] once one process get's in, it'll increase it each time it splits [03:34] yea well I actually only need to know whether there is even one or not [03:34] well then that should work [03:35] or you could make it spl #0, <0 [03:35] and not only it'll change the number first, so you know one process hit [03:35] but then your enemy will decrement the core for free [03:36] it'll change into spl #0, <7999 ...and then the B-field of 7999 gets decreased. which is the instruction above [03:37] it's the initial copy of the Door, so I guess that doesn't matter [03:38] true, it doesnt work like djn :P it's kind of obvious when you give it some thought [03:38] hehe [03:38] I'm still green [03:38] that's actually better because now the spl will only change ONCE and won't start over again after 8000 changes [03:39] but what's the idea of the trap? the enemy gets through and triggers a coreclear? [03:39] that is correct [03:39] well [03:40] it's a vampire/leech/whatever [03:40] it's kind of bogus, because the spl #0 slows the enemy so much, that the coreclear is horribly slow [03:40] yea but it's not THE method of ending it, I have another core-clear going on with my own, single process in the opposite direction ;) [03:41] this vampire will drop traps until it captures any one process, then it will force it to core-clear and switches into core-clear itself, into the opposite direction [03:42] then you should not use a spl [03:42] the spl will slow the enemy down while I'm clearing [03:42] increasing the chances that my actual core-clear on the other side indeed does overwrite enemy code [03:42] then you should only drop spl #0 <0 's [03:43] will take less time to drop [03:43] and you can still check if they trapped anyone [03:43] hmm, explain? [03:43] well, dropping the trap you made will take many mov instructions, one per line [03:44] and the advantage is that you dont just make the enemy slower, but you also make him clear for you [03:44] the problem is that the spl slows him so much the clearing is useless [03:44] oh, no, I didn't mean I'm dropping THAT trap [03:44] ah [03:44] ok then [03:44] the bomb is the basic vampire type, a jmp to the trap [03:50] well I think it's time for me to go to sleep [03:51] thank for the help [03:51] thanks* [03:51] and to the rest of the channel, you can look at my warrior too ;) [03:51] Nick Change: Leniad changed nick to zzzLeniad [04:38] MSG: Quit: 3( ‹ BaFeScript Beta9.1 3›-‹ www.bafescript.org 3› ) [07:48] Nick Change: Caelian__ changed nick to Caelian [08:50] MSG: Quit: Changing server [08:54] Join: pak21 joined #corewars [09:03] MSG: Ping timeout: 240 seconds [10:04] Join: fiveop joined #corewars [10:48] Nick Change: Luka|Zzz changed nick to Luka [10:50] h L [10:54] h M :P [11:14] * Luka is trying some evolving for the nano hill, rather fun :) [11:54] Good luck, Luka. I hope that you'll soon get a warrior onto nano. Unfortunately it is a little difficult, because every warrior is highly optimized [11:59] saw that :) [12:02] will soon give one of my evo warriors a try :D [12:09] What program do you use for evolving? Your own? [12:11] a mix ;) [12:13] mostly Yace [12:18] Join: Core_old joined #corewars [12:19] do you have to write you own evolver to submit evolved warriors? (not like I'm just watching the process tho :P) [12:19] nope [12:20] ok thanks :) [12:20] of course, Yace could use some tweaking etc [12:43] entered nano, #46 ;) [12:43] not bad [12:44] probably an imp-clear by the scores [12:44] doesnt exactly score like an nanopaper does [12:45] impressive score reading :D [12:46] nano is relatively easy to read, especially, because there aren't than many different good "strategies" [12:47] I see [12:47] mostly imp and imp-clear? [12:47] coreclear, impclear, paper, scanner, nanoqscan [12:48] Simply create some more warriors and you'll soon know how to read the scores [12:48] scanner? wow [12:48] hehe [12:48] Mizcu: Don't forgert the scanners and mad bombers [12:48] *forget [12:49] Luka: well, we did challenge PVK about two years ago to make small scanners [12:49] ... and oneshots, altough I don't see why you make a distinction between scanners and oneshots on nano [12:49] PVK? [12:49] One of the authors of nPaper II [12:50] I see :) [12:50] which was a quite good warrior on '94nop [12:51] he did succeed making a spl-bombing paper in 4 lines, though it sucked by performance [12:51] and then you can have linear scanning, which sucks even more, unless making a bomb-dodger [12:53] spl-bombing paper in 4 lines ... Crazyness [12:53] :) [12:53] well that was a type [12:53] spl-bombing scanner [12:54] i am multitasking, ok? [12:55] sure [13:31] MSG: Quit: Trillian (http://www.ceruleanstudios.com [13:43] Join: Leniad joined #corewars [13:45] Hi Leniad [13:45] heya [13:46] my warrior scores 99 against wilkies and is #13 in the beginners hill :D [13:46] http://pastebin.ca/494160 [13:47] hmm ... kind of hard to read [13:48] yes, it misaligned it [13:48] but if you read on the box where you can send changes, it is perfectly aligned [13:48] ah ... ok [13:51] you should restrict yourself to max. 80 characters per line [13:51] :) [13:52] true. I'm developing it in windows though :P [13:53] That's no reason not to use max. 80 characters per line. That's more or less the default witdth for all editors [13:58] your decoy is quite near to your warrior [13:58] well it somehow entered on my editor, that's why it's that long [13:58] I'm using ARES [13:59] Why don't you try to write a more sophisticated clear [13:59] ? [13:59] dunno, how would that be? [13:59] how about a sd-clear? [13:59] and you could add a light impspiral.. [13:59] and I think the warrior is long enough as it is [14:00] fir it would wipe the core with spl # 0 [14:00] and then it would switch to dat [14:00] hmm [14:00] i'll see how I can try to write that [14:00] btw, an imp spiral always wins against my warrior [14:00] what could I do to prevent that? [14:00] that way can stun more opponents and then you finally kill them with the d-clear [14:02] try to change the direction of your clear [14:03] an splclear into d-clear, you say eh [14:03] and instead of using spl/mov/dat as a clear why not spl/mov/jmp -1/dat [14:03] Mizcu: Anything wrong? [14:04] I thought spl/mov/dat was faster [14:04] have fun implementing that [14:05] spl mov jmp is as fast as spl mov dat [14:05] Leniad: Why don't you benchmark both versions (spl/mov/dat and spl/mov/jmp/dat) [14:05] ? [14:07] oh, you're right. It has to execute the dat anyways [14:07] well [14:07] I'll implement that when I get some time [14:07] :) [14:07] now I have to run to the uni :P [14:07] Nick Change: Leniad changed nick to Leniad_un [14:07] thx for the help! [15:13] Join: John joined #corewars [15:13] Hi :-) [15:14] h [15:14] Hi Mizcu [15:14] I'm going to have a go at your challenge on my day off :-) [15:15] 50 lines isn't very long though [15:15] should be sufficient [15:16] I wanted to try something more complex :-P [15:16] How long is your solution? [15:16] i havent even tried [15:17] though i think 20 might be enough [15:17] if speed is not an issue [15:17] 20 might be enough if there is always a solution with equal totals [15:21] well, you could probably make a sort of generic sorting which you run multiple times, switching numbers in table etc to close up with your goal [15:22] Urgh, sounds horrible ;-) [15:22] its easy to think of a small, and of a fast solution, but the average is hard [15:24] Welcome to the nano hill Luka [15:52] * John waves [15:52] MSG: Quit: mov.i #1,1 [16:48] Join: Leniad_ joined #corewars [16:48] MSG: Read error: Connection reset by peer [17:06] Join: Leniad__ joined #corewars [17:06] MSG: Read error: Connection reset by peer [18:08] "Just give me a moment, ill be right back. And try not to look too rich." [18:08] Nick Change: Mizcu changed nick to Alec_Miz [18:08] damnit [18:08] Nick Change: Alec_Miz changed nick to Mizcu [20:45] Join: Leniad_ joined #corewars [20:48] MSG: Ping timeout: 240 seconds [20:51] Nick Change: Leniad_ changed nick to Leniad [20:51] damn connection [21:30] doing a sd-clear is damn hard [21:31] depends [21:31] it's hard if I try to do it >_< [21:40] there, it works ^__^ [21:42] hmmmmmmmm [21:42] one problem is that if the djn train hits a one, it goes into coreclear, and that makes the warrior suck [21:50] whoa, after some ironing, it's scoring 110 on wilkies! that's 10 points more than I ever had [22:18] mizcu, you suggested I added a light imp spiral How'd I implement that? [22:19] there are couple different ways to launch an imp-spiral [22:19] oh, and my warrior can't win against an imp spiral too [22:19] can you wait a while? i am slightly busy atm [22:19] sure, no worries [22:31] MSG: Read error: Operation timed out [22:31] Join: Mizcu joined #corewars [22:36] MSG: Read error: Connection reset by peer [22:41] Join: Mizcu joined #corewars [22:42] Join: elkauka joined #corewars [22:42] hi folks [22:45] heya [22:50] Leniad: there are couple ways to launch an imp-spiral, [22:50] Binary launch (fast, but really huge, first type of launch) [22:51] Jmp/add -launch (can fit in small space, relatively easy to code, but creates processes that need to be killed after the spiral is launched) [22:51] Vector -launch (Fast, but the spiral is limited in size to how big table you make) [22:52] Imp-pump (there are two versions; big and small.) [22:52] I guess I shouldn't launch the spiral until after I bombed, so it has to be a small launch [22:52] hmm, was your bomber the add mov mov jmp -one? [22:53] nop, my bomber has only one mov per loop [22:53] let me place the latest code on the pastebin [22:53] spl mov add djn -one, yes [22:53] yes [22:53] I took the djn train out, because if it hit an 1 it'd go into coreclear before finishing the bomb run [22:54] big imp-pumps are generally not used, because they are big. But they generate really heavy and nasty spiral by time [22:54] small pumps can be fit in 3-4 instructions (plus the imp), but they are slightly unoptimal in speed [22:55] you are better off with a vector launch, i think [22:55] http://pastebin.ca/495865 [22:55] when should I launch the spiral? after the clear, during the clear, don't clear and just launch? [22:56] you should start the spiral-launch and the stone at same time [22:56] but if possible, give the stone little more processes when launching the two [22:56] but then I have the risk of hitting the spiral with a bomb and slowing myself down? [22:57] the balancing of processes is a delicate thing, it took me hours to get it right in Song of the blue sea [22:57] it is possible to get the spiral in such position that it is never hit during the run [22:58] that'd take lots of finetuning I bet [22:58] not always, but yes [22:59] easiest way to finetune it is to change the distance of imp-launcher to stone [23:01] http://homepages.paradise.net.nz/~anton/cw/corewar-faq.html [23:01] there is one vector-launch at the bottom of the page [23:01] you could add one more spl 1 to the top, and continue the table to dat #0, imp+6*sz [23:02] no, imp+7*sz [23:03] I was looking at http://corewar.co.uk/94war/94war12.txt [23:04] the large vector there has a weakness in speed, but it should be good enough for a try [23:04] anyways, I won't add it to my warrior until I understand perfectly how it works [23:04] so it might take a while ^___^ [23:05] yeah, and i wont be helping for a while; i need some sleep now [23:05] feel free to pester elkauka [23:05] :P [23:06] you are very generous with other people's time :D [23:06] watching southpark here ... i'm going to check the chat now and then. [23:08] oh, a #number is counted as zero for jumping, but if adressed indirectly from another function, it's counted as it's number instead of zero? [23:10] yes [23:16] i'm actually surprised you could understand my question, I phrased it horribly :P [23:17] well, the implaunch is working, but my warrior has so many processes bombing that the spiral is horribly slow [23:20] it aint supposed to be fast, it's just for backup [23:21] to tie in case your bomber dies [23:24] oh [23:24] well [23:25] I score 8 points more in wilkies now [23:25] so it must be working [23:26] ^^ [23:26] yes [23:27] after some testing against a warrior that used to beat me, now it can't beat me [23:27] ^^ [23:29] how is 126 at wilkies? [23:29] I mean, how good is it? [23:29] dunno, send it to a hill and compare the score [23:29] guess wilkies is kinda outdated, not sure [23:30] try: http://www.corewar.info/optimax/dload/fsh94nop0.3.zip [23:31] MSG: [23:37] that is surely going to destroy my poor little warrior :P [23:38] ^^^ [23:39] oh, that has lots of warriors in subclasses [23:39] should I benchmark against all of them at the same time? [23:39] btw: digitalis is in the "clr" folder. you may want to check it. it's a d-clear with imps [23:40] bah, I score _less_ on the begginner hill with the imp spiral [23:40] 5 points less to be precise :P [23:41] all warriors in one folder are following more or less the same strategy. therefore benchmark them separatly to see how you perform against the current strategy [23:42] benchmark it against all of them to get a good overall score [23:47] it got 90 overall [23:48] its prettey good afterall [23:51] hah [23:51] it got 97 without the spiral [23:51] meh [23:51] I don't get this :P [23:51] brb, food