[00:03] Join: _jake joined #corewars [00:04] <_jake> bvowk, hey... this is w0lfpak. Where can I get pmars for OS X? [00:25] ah, thats an thing we still need to gave fixed [00:26] you can use just the *nix pmars from sourceforge, but it needs couple finds modified before it works on OS X [00:26] <_jake> yeah... [00:26] <_jake> any patches i can apply? [00:27] http://groups.google.com/group/rec.games.corewar/search?group=rec.games.corewar&q=pmars+OS+X [00:30] <_jake> Mizcu, nice! Thanks! [00:49] <_jake> i got a KOTH server version to compile... will that be enough start working on a warrior? [00:58] the server version will run the warriors, but since it lacks debug, well, it makes harder to debug warriors [00:58] <_jake> hmmmm [01:09] since server version only outputs the result, while a lot of experimentation from beginners needs to be able to see step-by-step what is happening in the warrior (and possibly outside) [01:28] jake: do you know any python? [01:44] <_jake> bvowk, i know python [01:49] <_jake> i got it to compile as a command line program (not as a server) [01:49] <_jake> i can enter the debugger now [01:52] you could try pycorewars if you like, its a pretty speedy mars for python.. [01:53] <_jake> link? [01:54] hrm. [01:54] thats a good question. [01:54] <_jake> found it [01:54] <_jake> http://corewars.jgutzeit.de/pycorewar/index.en.html [01:55] I abuse it for my evolver :) [01:55] <_jake> why not just use pmars? [01:56] faster, better integration with the code [01:56] no parsing overhead, no exec overhead [01:56] <_jake> hmmmm [01:58] <_jake> no p-space [01:58] yeah, no pspace. [01:58] pspace is a huge issue with evolving tho. [01:59] course, nano, and nop don't need it, the two hills I mainly play with [01:59] bbiaw. [02:46] <_jake> i've come up with a compromise: dosbox on os x running pMARS (0.8 ... not 0.9.x) [02:47] <_jake> at least i've got some visualization (running pMARSV) and a debugger [02:47] <_jake> we'll see if i need more than that [02:47] <_jake> thanks everyone for the help! [02:47] <_jake> i'm done for the night [02:47] MSG: <_jake> Quit: see 'ya when i see 'ya [02:57] Join: Caelian_ joined #corewars [02:59] jake: remind me tommorrow morning and I'll compile you an OSX pmars, I just found the patches, but my laptops at work. [03:01] MSG: Ping timeout: 240 seconds [04:08] MSG: Ping timeout: 240 seconds [04:08] Join: Caelian|w joined #corewars [05:04] Nick Change: Caelian_ changed nick to Caelian [05:39] MSG: Remote host closed the connection [05:39] Join: Caelian joined #corewars [10:37] Join: Core_new joined #corewars [12:58] MSG: Ping timeout: 240 seconds [14:21] interesting things one can learn [14:21] freebsd + intel card will automatically switch rx/tx lines when necessary, windows cant switch its own ass [14:23] in other words, instead of 10 meters of utp, i got 20 meters of phone cable taking dsl-box close enough to my 'puter [14:23] just switched the order of stuff in this room, and annual cleaning [14:38] ahh. Auto-mdix, why it isn't standard practice I dont' know. [14:40] well, actually, I do know why... network loops suck, and users are clueless.. but still. [14:40] (waitaminute, freebsd? thats not what i installed on that box..) [14:41] openbsd? [14:41] netbsd? [14:42] open [14:42] Opens made right here :) [15:01] now i just have to learn to have cdvd-drive and colume control on my left side [15:01] volume control, that is [15:28] Join: Core_new joined #corewars [16:51] Join: sf_ghoul joined #corewars [16:54] Hmm I'm attracting creepy messages since I put a picture on stumbleupon :( [16:55] And it's only 48px square! [16:56] hrm. [16:56] Hi Bvowk [16:56] sorry about the creepy messages, I'll stop ;) [16:57] Thanks [16:57] :P [16:57] A few of your warriors are dropping off nano now bvowk [16:57] yeah, I know. [16:57] Just before entering the HoF too :/ [16:57] hwm is doing too good. [16:59] sounds like nano is starting to become more or less a "monoculture" of warrior types [16:59] I'm tempted to redirect some evolver back to nano ;) [16:59] no, its more varied than when it started [17:00] we've got papers and imps and scanners (oh my!) [17:00] i should get back into the whole evolver thing ... though not really sure where to start out :) [17:00] in addition to the mad mad bombers. [17:00] you don't like python right? [17:00] it's not my preferred language ... but oh well .. it's not That bad ... [17:01] it just has a few idiosyncrasies that just simply feel bad :) [17:02] because you could take http://www.math.ualberta.ca/~bvowk/files/jnsl-1.1.tgz [17:02] and use that. [17:02] there's not much to it [17:02] and it'll evolve alright out of the box [17:02] it's what I've been using on nano [17:02] if it's anywhere near as complex to get going as your old ccai stuff i am in for significant work :) [17:02] it doesn't seem to work so well on >nano, needs some tweaks [17:03] no, its only a couple hundred lines [17:03] Join: w0lfpak joined #corewars [17:03] its basicly just enough to get it going [17:03] hehehe .. i'll have a look [17:03] basicly a hill, and then a simple evolver on top [17:03] hey wolf [17:03] hey [17:04] how do i register my nick? so i don't lose it like last night [17:04] Hi wOlfpak :) [17:04] hey ghoul [17:04] this server has no registeration for nicks [17:04] shame [17:04] wOlfpak: have you learnt redcode now? [17:04] there are only about a half dozen people on this server at any give time ;) [17:04] ha! [17:05] almost [17:05] bvowk, you compiled the binary for jake/w0lfpak? [17:05] How much time have you spent on it? [17:05] mizcu: just getting to that.. [17:05] Mizcu, nah... [17:05] I'll post it when I've got it, so we don't have this anymore ;) [17:05] i ran the dos 0.8 version on dosbox for os x [17:06] ha! [17:06] once you start making more than just single-round fights, the dosbox will show just how bad it can be [17:06] I'll be on vacation for the next 10 days, so I probably won't be submitting anything to any hills before that [17:07] any binaries for os x will require X11 or ncurses... what we need is a Carbon version. [17:08] the dynamic core isnt bad, but when you are running program that eats all process power it can, the overhead gets really nasty [17:08] ah [17:09] for example, when i am testing if a warrior idea is good, i use a quick benchmark consisting of 100 fights against 12 warriors each [17:10] wow... so you're not exactly single stepping through the debugger on each warrior [17:10] and when i crank out an optimizer, it can get 5000 battles for each warrior, with maybe 1% hit chance.. [17:11] well, you do something in two different ways, both the same size and speed but different in construction, how do you know which is better? [17:11] right... gotta run a benchmark [17:12] MSG: Quit: Ex-Chat [17:17] Join: Caelian joined #corewars [17:21] bvowk: ... hmmm from python ... import pool ... gives me error ImportError: No module named Corewar [17:21] bvowk: do i somehow need to "install" the PyCoreWar stuff first ? [17:22] yeah.. [17:22] python setup install [17:24] ok ... it's going :) [17:30] bvowk: i assume the parameter to run_pool() .. is the amount of rounds to "evolve" ? [17:32] yup [17:33] doh. [17:33] SDL pmars *MOSTLY* works [17:33] ha [17:33] getting a somewhat random bus error [17:34] bvowk: hmmm .. those are always interesting [17:35] bvowk: what does your /etc/malloc.conf look like ? [17:35] on osx? [17:35] bvowk: assuming you're on FreeBSD [17:35] ok .. ignore :) [17:40] bvowk: am i right in assuming that when the score of the hill entry sent to valhalla dropping significantly during the first 10-20 runs it means that the warriors on the hill are starting to "develop defensive strategies" against one-another ? [17:50] I want to build one! http://jpbrown.i8.com/cubesolver.html [17:52] yeah [17:52] hehehe [17:52] nice! [17:53] w00t. [17:53] I've got the X11 version running nicely on OSX [17:53] did you have to use fink to get X11 installed? [17:54] no, there's an X11 package on the install CD that you install [17:54] ah, yes [17:55] http://www.math.ualberta.ca/~bvowk/files/pmars-osx-x11 [17:56] wheee ... i already got something that actually scores 35% wins against its siblings ... instead of tieing over 80% of the time :) [17:56] :) [17:59] cael: its actually amazing how fast they take up arms against eachother isn't it? ;) [17:59] bvowk: hell yeah :) [18:00] bvowk: btw ... what does the SLT instruction do again ? [18:00] Please give it a nice name before you submit it to the hill :) [18:00] sf_ghoul: hehe ... it's no where near that state yet :) [18:00] skips the next instruction if let than. [18:00] er.. less than [18:00] aah .. the next instruction [18:00] iirc [18:01] i assumed it meant Skip if Less Than ... but i was wondering what it was supposed to skip :) [18:03] bvowk: one more question .. i notice the "names" generated consist of 3 "sections" ... i assume these indicate some form of ancestry ? [18:06] the names are formed by the instructions.. the first one is the instruction like JMP etc.. the second is the modifiers, and the third is a hash of the oprands [18:06] basicly lets you uniquely identify any warrior.. [18:07] and it breeds them by similarity in the instructions.. (ie, it tries to group warriors with the same first name for breeding) [18:07] check [18:07] this is Fun :) [18:09] it's nice to see how fast the hill itself is evolving away from essentially "one shots" ... to something that starts to actually build some form of strategy :) [18:09] What's wrong with oneshots?! :( [18:10] nothing wrong with them per se :) [18:11] what i was trying to say was that essentially the initially "bred" warrior pool is little more than a bunch of instructions, operands, addresses and modes packed together ... and only after not even 30 cycles ... it starts to mold into something that actually tries to make sense :) [18:12] :) [18:16] cael: if you come up with any improvements, make sure you fire them back down the pipe ;) [18:17] (to the code, I don't want your warriors ;)) [18:18] Only three published warriors on the nano hill [18:18] bvowk: hehe i figured :) [18:20] the first warrior that actually manages to not get pushed off of valhalla directly after sending it ... is the first one i'll try sending to nano ... but something tells me that'll be a while still :) [18:20] cael: it depends, you can get lucky and get a kickass warrior right away.. [18:21] usually not what happens tho. [18:21] You could try sending one now to see how well it does [18:21] And send every 30 mins or so to compare [18:22] heh [18:22] cael: if you end up with 400k worth of them by accident, send me an email and I'll show you how to take care of that too ;) [18:23] 400k ? [18:23] how does one send to sal again ? [18:24] cat | mail ? [18:24] (don't forget to change the name!! [18:24] ) [18:24] :) [18:24] bvowk: hmmm .. Help on SAL page says [18:25] Either works [18:25] ah ok ;) [18:27] 51 23.3 67.2 9.5 Whispering Void caelian 79.3 0 [18:27] thats a bit short ;) [18:27] something tells me there's a LOT of "work" ahead :) [18:28] Nice name though [18:28] its more like wimpering void ;) [18:28] heheheh [18:28] i thought i'd start out "quiet" [18:29] I've found you'll have good success with running a couple dozen random starts for a while and then comining the best, and loading those into the initial hill. [18:31] bvowk: you mean let a couple of random hills build up for a bit ... save each hill ... dump them all in one "super hill" and let it stew from there ? [18:32] pretty much.. [18:32] there should be a quick_test script in there.. [18:32] it's there indeed [18:33] trying to see what it's supposed to do :) [18:33] so I get rid of the crap from what I've saved, (./quick_test.py save valhalla | sort -rn -k 2 | tail -n 50000 | awk '{print $1}' | perl -l0ne 'unlink("save/$_");' [18:33] ) and then use the run_pool.py script to load from the save [18:34] i see what you're going at :) [18:41] that one evolved? [18:41] er.. [18:41] sorry, wrong name [18:43] :) [18:47] ;strategy RedShift is an evolver (yes, I wrote an evolver!) [18:47] heh... inversed should come talk about it :) [18:49] hehehe [19:15] MSG: Quit: Bersirc 2.2: All the original sexiness of Bersirc, open to the world. [ http://www.bersirc.org/ - Open Source I [19:54] bvowk: i am slowly climbing :) [19:54] 51 31.5 49.6 18.9 Whimpering Void caelian 113.4 0 [19:54] nice. [19:55] it's a SPL, MOV, MOV, MUL, JMN construct [19:56] if you wait a while it'll become SPL MOV MOV DJN ;) [19:56] what does DJN mean again ? [19:59] decrement and jump [19:59] (if not zero) [19:59] dude.. I don't even write redcode! [20:02] lol [20:02] sorry ... just need to refresh up :) [20:55] MSG: Quit: Leaving [20:59] MSG: Quit: Trillian (http://www.ceruleanstudios.com