[02:47] Join: Waknuk_ joined #corewars [02:48] MSG: Ping timeout: 255 seconds [02:52] MSG: Quit: Ping Timeout [02:54] Join: Waknuk joined #corewars [02:55] MSG: Ping timeout: 255 seconds [03:11] MSG: Quit: leaving [04:49] MSG: Read error: Operation timed out [06:45] Join: johnkw joined #corewars [09:16] Miz: wanna play pimp my warrior? [10:02] ares: i can [10:05] i fear, i lost the source... was in my test.red file... i've been working on the parser and the file contains some test-for-loops :-( [10:12] Join: fiveop joined #corewars [10:14] i dont understand, why people use "L M FOR 0 \n ROF" to create a label at the start of the source... L points to the first instruction, so what is the FOR good? [10:18] dunno, which warrior uses such? [10:18] some of koenigstuhl [10:19] give me ones name [10:19] dawn.red [10:20] ah, thats the really ugly qscan [10:20] lol [10:22] because i've seen this FOR 0 construct in some more warriors, i suspect it to be useful [10:25] brb, switching computer [10:40] there [10:49] MSG: Read error: Operation timed out [10:49] Join: asw joined #corewars [10:50] damnit, i forgot to bring my Yace-soup with me [10:50] well, no big deal [11:40] silly ares went and lost the file so i have little to do now.. [11:41] hm... i'll look, if i can help [11:51] Join: Core_old joined #corewars [13:10] Release-candidate 1 of the stone-article is now available [13:10] same url? [13:10] http://koti.mbnet.fi/mizcu/stonearticle.txt (yes, same url, remember to push refresh) [13:11] And i immediately noticed a typo/bad english expression.. [14:55] Join: Fluffy joined #corewars [14:56] fraggin fun [14:56] :) [14:56] now i just have to write every warrior in the benchmark of 50 into an .ini -file [14:56] write a short script to do it [14:56] i dont do perl, and i dont have other scriptlanguages [14:57] Then it is about time to learn one [14:57] with exception of the command-script which i an not comfortable with [14:57] oh well, time to crack quickbasic open [14:58] urgh! [15:29] * Fluffy waves [15:29] MSG: Quit: fluffy.i < 1, # 42 [15:32] oh shit no [15:32] i want to beat someone up [15:32] pmars refuses to run with core less than 80 for some fucking reason [15:33] well, less than 100 [15:34] well, i still have older version for backup [15:42] there is a serious bug in the pmars parser... i already mentioned it... shall i post it to rgc? will anyone fix it? *G* [15:43] do so [15:59] oohkay.. [16:08] so i got Species to ignore Markovs, but now i have to make it ignore operand-frequency tables [16:11] or figure out why the tablemaker doesnt like me [16:16] its lovely how Will is always away when i curse at he's program [16:24] or rather, how i always start using it the day he is gone.. [18:09] greetings humans! [18:09] h bvowk [18:09] I hope you all had a nice new years eve [18:09] well this certainly is interesting, yace churned out sort-of maybe-working imp-almost-spiral launcher [18:10] scores kinda meh, though [18:43] "too many labels for a declaration" ?? [18:44] i start reconsidering my opinion of pmars as a standard [19:09] someone give me a debarrier [19:10] ok, it wont be of any help, but its always nice to think of the possibility [19:22] Join: willvarfa joined #corewars [19:23] Join: _willvarf joined #corewars [19:23] MSG: <_willvarf> Remote host closed the connection [19:25] hi willey [19:26] hi miz [19:26] happy new year [19:27] generating alpha:1 .. CSpeciesError roulette off end of C [19:27] OperandTable::COperandTable::suggest [19:28] but looks like i got lucky, and it generated a lucky warrior on first try [19:28] so it fails when retrying to make a warrior [19:29] i squeezed 50-warrior markov and table for test, but species still dislikes me [19:31] hmm, that is a nasty problem [19:32] it is claiming that it tried to generate a random number between two bounds, but failed [19:32] it could be a simple off-by-one bug though [19:32] interesting for it to constantly repeat itself [19:32] hmm bad seeding? [19:33] have you tried deleting all the .dat files in the directory? [19:33] yes [19:33] (always do each run in a separate directory) [19:33] (it stores the seed in the .dat files, so that it is repeatable) [19:34] Join: Roy joined #corewars [19:35] happy new year roy! long time no hear; everything well? [19:35] http://koti.mbnet.fi/mizcu/species2.zip contains what i have remaining in the species-folder after clearing to the minimum [19:35] hmm, Mizcu, email me your ini file and any markov and insn files and I'll try [19:35] Happy new year! [19:35] Hi all [19:35] crikey, what I had in mind Mizcu :-) [19:36] Where have you been lately will? [19:37] baby and house renovations and busy at work [19:37] between it all, the world is conspiring to stopping me hobby coding [19:37] yourself roy? [19:37] recode that conspiracy *g* [19:38] Thank god no baby, still no house, but I am very busy with work too ;-) [19:38] she's the boss ares [19:38] ah you graduated? [19:38] Yes this spring! [19:38] And I already found a very good job (Java developer) [19:39] what kind of industry? [19:40] Its a big dutch software development company [19:40] so making software for sale? [19:41] Currently I'm working on a program for a daughter company of DHL (package transportation) [19:44] And some recreation of course http://www.youtube.com/watch?v=pbTWVDH4A04 [19:52] Mizcu: waiting for download from sf... :-( [19:53] =/ [19:55] well, i have sources too if sf get really, really cranky [19:56] Join: Core_old_ joined #corewars [19:57] So how is species will? [19:57] just got the latest cvs tarball from sf [19:58] From sf..? [19:58] eh, still being used apparently, Roy :-) [19:58] sourceforge [19:58] Ah ok [19:58] * Roy was thinking of a girly sf :) [19:59] MSG: Ping timeout: 255 seconds [19:59] ? [20:01] Never mind ;-) [20:03] Mizcu: I lied, it timed out [20:03] strange that firefox didn't report that in the download manager, instead claiming it was done [20:06] so how did your thesis go, Roy ? [20:07] Everything went great :) Never finished the evolver though :( [20:07] thats a shame [20:07] redcoding these days? [20:09] Not much, very busy with work etc, made some warriors this holiday [20:10] Expressed this a couple of times, but if I could call a fast mars from Java I would really like to make a Java evolver [20:11] But as always I'm lacking time.. [20:11] Mizcu: bad news :-( doesn't compile with g++ 4.0 [20:12] (whether that is that the cvs version is broken, or that g++ 4.0 is less forgiving than 3.x that species was developed with, I'm not sure) [20:13] will: many a obstacle.. [20:17] hmm fixing [20:21] Mizcu: compiled; have you got a copy of runtest_dummy.rc you can put somewhere? [20:21] currently not, but its a 4 second code [20:22] what is it? [20:23] jmp.b $2, $2 [20:23] jmp.b $1, $1 [20:23] start: jmp.b $0, $0 [20:23] add org start is required [20:25] on nano-hill its kinda hard to use a CLP as i did previously with species [20:34] hmm, there are no instructions in the operand table [20:34] its empty [20:34] CLP? [20:34] a) wtf b) Curious little program, an anti-coreclear [20:35] ah right [20:36] what seems to be happening is that the operands table code doesn't work well for nano [20:36] not good [20:36] nope not good [20:37] can you do random operands? [20:37] with a nano core, there isn't going to be much meaningful distinction in there anyway prehaps? [20:38] not really [20:38] plain randomity should be enough [20:39] well, 60% movs, 15%spl, 15% jmp/djn, 10% rest would be good enough [20:40] so Mizcu, you had fun or success using species before? [20:40] I'm flattered that some has tried to use it :-) [20:41] your memory has gone sore [20:41] i am pretty much only person to ever use species for anything :) [20:42] it was nice watching it churn 2-3 years ago, but it never built anything useful, with exception of 1-2 plain stones, or some coreclear wanting to be the new d-clear [20:43] but i think species wont do badly in nano [20:43] yeah about the luck I had too [20:43] species-blossom is a more promising idea, I think [20:43] (not coding at the moment, though) [20:43] since Mov-towers are more of a rule than oddity in nano [20:53] Roy: you want a dmars :-) [20:56] Err.. do I? [20:57] calling from java etc [20:58] Well, yeah, any mars the can be easily called would be perfect, especially Jens's PyMars! (it is in C and has Python wrappers but could be just as easily called by Java) [20:59] I've lost a bit of track of what evolvers and mars and things are available these days [21:01] Not much made, Jens (Fluffy) is still working on PyMars, currently the best and fastest mars I think, and then there is ARES working on a mars+parser [21:02] more of a monolithic multitool [21:02] Ares has inbuilt irc-client already [21:02] :| [21:02] It has!? Creeky [21:02] crikey! [21:03] pymars is faster than fmars? [21:03] It seems, and doesn't have to be recompiled [21:05] wow, webpage? info on how and why? [21:05] He has some info on his website I think, but its only a link to a older version :) [21:06] a version that wasn't faster? [21:06] Probably as fast, but it didn't support p-space yet if I am correct [21:06] webpage? [21:07] http://corewars.jgutzeit.de/pycorewar/index.en.html [21:08] Very un-updated :P Fluffy, finish the mars and update the site pleeaassee :-D [21:08] Join: Waknuk joined #corewars [21:08] crikey, the speed comparision and documentation links are not linked [21:09] Hi Miz, Will. Will! [21:09] hi [21:09] Hi Waknuk [21:09] isn't fmars fastest and examrs best (general purpose, fast etc) [21:10] Hi Roy [21:10] Have just been in email with Pavel ... has anyone been able to give him feedback re his changes? [21:10] Not any more, pycorewar is probably better, its faster (fastest) and general purpose [21:11] someone called ares the swiss army knife of core war *g* [21:11] By what/whose measure fastest? Isn't it just a wrapper around a C/C++ version if so which? [21:11] And another big pro is that the core of the mars is very optimized C code but is already supplied with Python wrappers [21:12] Fluffy has a benchmark for Marses, I tried it and it was much quicker then exmars and of course pmars [21:12] I ahve pymars installed ... lt me look [21:13] but that irc is just the PJIRC-java applet, www.pjirc.com [21:14] hmm bvowk using pymars? (that'd convince me) [21:14] it is fantastic if we have a faster one [21:14] and fascinating to know how much faster, and why [21:14] pymars is the work of fluffy, isn't it? [21:14] Waknuk: the one for download on his website now is a pretty outdated version, I think most is rewritten by Jens now [21:14] Yeah [21:15] Far from convincing but with a warrior pair chosen at random (van & wintermute) exmars=2.82s pymars=3.50s (1000 rounds) [21:17] van EQU vain [21:17] Yes but what version are you using? pymars started as the slowest mars ever made but got faster eventuelly [21:17] no, ares is the slowest mars ever [21:17] Hehe [21:17] no, head-mars is the slowest ever [21:18] The file is PyCorewar-0.2.0 [21:18] No, roy-head-mars is slower [21:18] Something I salvaged: [21:18] Averaged results [21:18] pMARS : 8.861 MIPS [21:18] PyCorewar : 33.880 MIPS (282.3 % faster than pMARS) [21:19] <--- is the one trying and failing to get onto the draft hill [21:19] What are you trying to make? [21:20] Almost everything completed is faster than pmars ... exhaust, qmars, exmars, fmars [21:20] Yeah, so it doesn't say much.. [21:20] Trying to test a new switch. Only handles two components. Hate imps. [21:20] corewin is not.. (even when graphics disabled) [21:21] (no actual testing done) [21:21] Corewin has a different niche ... friendliness for windows users [21:22] Roy are you trying to make me KOTH on 94nop ... ;) [21:22] Not really but it seems to be a nice side-effect [21:26] I still think the scores are a bit dissapointing... this 2007 warrior scores second best in my benchmark [21:26] (only hullabaloo is slightly better) [21:27] or your benchmark needs updating [21:27] Bymy benchmark my draft test warrior is top-50 khill material. But, [21:27] only one of the draft warriors is published (and it's a rip-off). [21:28] Heh yeah, and a lot of draft warriors are really nop warriors [21:28] by benchmarks Song.. was not really 'nop -material, but it turned out nice [21:29] Conclusion: our benchmarks suck ;-) [21:29] I had a new stone score 92 vs draft earlier today. (alone) [21:29] I hate benchmarking/tweaking. I put a dat 0,0 at the front of a 99 line warrior and the scroe dropped 1.5 [21:31] voodoo, the strongest tool one can (try to) control [21:32] Join: Fizmo joined #corewars [21:32] hi hi [21:32] ahoi Fizzie [21:32] hi fizmo [21:32] hi all :-) [21:33] crikey, getting full in here [21:33] hello fix [21:33] fiz [21:33] Hi Fizzer [21:33] hohohooo, really a lot of traffic here [21:33] :-P [21:33] Hi Fizmo Master [21:33] Hiiii [21:33] haven't seen you for so long time [21:35] Fiz: http://koti.mbnet.fi/mizcu/bench.txt [21:38] hmm, i havent added Poke in the plain stone -category [21:42] The Core Clear isn't a oneshot [21:42] maybe Sanny Boy [21:42] I know quite clearly. Check the Scanny Boy's position if Koen =/ [21:43] all other warriors are around 400-500 [21:43] yes, seems a good area for medium warriors [21:46] Any interest in asking KOTH to increase the number of rounds run? [21:49] oh, and i am very low on ideas on the easy-bench [21:55] how come everyone silenced? [21:57] Mizcu: you get species running now? [21:57] will: ? [21:58] Mizcu: you were going to try random operands? [21:59] actually, havent touched species in a moment [22:02] heck, i dont know what to insert in the .ini [22:03] inst_gen was easy to check as weighted random, but the operand hasnt hit my eyes with some hunting [22:03] * willvarfa looks in the cvs browser [22:11] Mizcu: operands=distribution [22:11] another error.. [22:12] (operands: _num_points expected) [22:12] operand_num_points = nnnn [22:12] operand_0 = nnn [22:12] ... [22:13] ah [22:14] make sense? [22:14] yes [22:14] phew [22:14] code was really obscure [22:19] ok, i got species to work [22:19] wahaoo [22:19] though i have to play with the operand_x -table a bit [22:19] nano is a coresize of 8? or 80? [22:20] 80, warriorsize 5 [22:20] 1/8 [22:20] so prehaps buckets for 1,2,3,4,5,6,12,24,80? [22:21] * Roy waves to all, have to work again tomorrow (yuck!) [22:21] MSG: [22:22] MSG: [22:28] bah, the actual meaning for the operand_x=value escapes me [22:31] there is probably a better way to do it [22:31] I can explain it if you like; it can be very powerful [22:31] I'd imagine it would make a major impact upon the time to find a useful warrior [22:32] i have to say that almost nothing in the source of the species actually makes sense to me [22:33] Mostly because i dont know any more C than i did 3 years ago, but Species certainly is not written to be idiotproof like Yace [22:33] ah, rather, species was written by an idiot [22:36] so if i could get a quick tutorial to the meaning of the operand_x [22:38] imagine three points in the distribution [22:38] hmm [22:38] I'd have to look at the code :-) [22:39] i certainly hope Blossom will be coded by less of an idiot ;) [22:40] http://redcoder.cvs.sourceforge.net/redcoder/species/distribution.cpp?revision=1.4&view=markup [22:40] the function at the very bottom is the one that does it [22:41] _size is the operands_num_points you specified [22:42] so pick a number between 1 and _num_points (we, zero-based, but ignore) [22:42] and then make a random number between that point and the next [22:42] so: [22:42] operands_0 = 10 [22:42] operands_1 = 20 [22:42] operands_2 = 80 [22:43] means that there is an equal chance of a number less than 10, or less than 20 [22:43] rec.games.corewar is listed in the blacklist of "cyber patrol"... tags: Quest/Illegal/Gamble Violence/Profanity Intol [22:43] but numbers between 20 and 80 will be much rarer [22:44] 1,2,3,4,5,6,12,24,80 would benefit small operands (e.g. those likely to be self-referencing the warrior itself), and then a point for a gate, and then less in the rest of the core [22:46] and running, say, 1 2 3 4 5 6 7 would give same chance for all? [22:46] silly me [22:46] would make an operand that was beyond 7 really rare [22:46] (down to mutation alone) [22:46] 10 20 30 40 50 60 70 80 [22:46] yeah [22:47] but would you want that? [22:48] (logic bug in there, never going to choose a number less than operand_0 is my bet, so prehaps you'd not find many operands that are 0 :-( [22:51] time to let this one soak overnight [22:52] good night all [22:52] * willvarfa waves [22:52] MSG: Remote host closed the connection