[00:33] MSG: Quit: Lost terminal [11:02] Join: fiveop joined #corewars [11:46] MSG: Quit: Lost terminal [12:25] http://tools.ietf.org/html/rfc3092 [12:46] Join: Fluffy joined #corewars [12:46] foo! [12:49] I'll foo back later to bar a little bit more ... [12:49] MSG: Client Quit [13:30] ahoi [13:31] Join: fiveop joined #corewars [13:31] Join: Roy joined #corewars [13:31] okay, as i said 1 minute ago: ahoi [13:31] Fluffy...? [13:31] Hi :) [13:32] nope [13:32] left 13:50 CET [13:32] Doh! [13:32] Just wanted to ask him something.. but well, don't have time to do something with it anyway ;-) [13:33] corewar destroyed my sleep rhytm *G* [13:34] Hehe sorry.. [13:35] Fluffy made a very nice/good mars that uses 90% C and some Python to have a pluggable interface [13:35] I was wondering how he connects Python and C/C++ [13:35] i have had a look to these sources [13:35] a short look [13:36] If he uses a framework like SWIG I could use it to make a Java interface easily (then I'll definitly make a evolver!!) [13:36] using java?? [13:40] I work with Java 8 hours a day, so then its easy to write a evolver in that language [13:40] And if I have to switch and learn another language in the evening I don't have time to do that :( [13:43] ic [13:43] but what about claculation time.... especially on an evolver? [13:43] btw. what do you do 8 hrs a day using java? what kind fo job is it [13:48] MSG: Quit: Lost terminal [13:49] I'm a Java programmer, making JEE applications (Web) [13:51] Join: Fluffy joined #corewars [13:51] re :) [13:51] :) [13:52] Roy: Your prediction has fullfilled itself: I've got a nice cold :( [13:55] Err...? [13:55] CNS? [13:55] yes, but RNS (running ...) is a better description [13:56] Calculation time isn't so bad in Java, Java 6 is pretty fast [13:57] calculating what? [13:57] See logs, you most [13:57] *must [13:57] it this Yoda-talk now? [13:57] but it's still an interpreter, right? have any idea of how much faster the same code is, when written in c, assuming an "average" program [13:58] (min, max, before you are gonig to rip me apart for such a question) [13:58] ares: In my experience 90% of all code is fast enough with any interpreted language [13:59] fast enough? *g* [13:59] and if your code belongs to the remaining 10% you'll know it soon enough [13:59] LOL [13:59] yes, fast enough [13:59] when thinking of a mars, for example, you'd probably be in the 10%, right? [13:59] no [14:00] Java is (with the new JVM etc) in most cases as good as C [14:00] not probably, but definitely [14:00] and only if you have to do a lot of benchmarking [14:00] really? wow! i know, its a pcode interpreter, but thats amazing [14:01] we had a talk about evolvers written in java [14:01] http://www.idiom.com/~zilla/Computer/javaCbenchmark.html [14:02] Roy: About your earlier question in the logs: I don't use SWIG, but the Python's C-interface, which is very easy to use [14:03] Java is pretty slow with String building etc, but thats a mess in C too (you'll need Python for things like that) but there isn't much string handling with warriors when you make object out of them [14:03] if you think that way, i am really going to implement an optimizier within ares [14:03] Fluffy, how hard would it be for a newb C/Python programmer to create a working Java interface? :) [14:04] Roy: but you can easily use PyCorewar's code for your Java-projects, because the MARS iself uses nothing but C [14:04] (and not Python-interface) [14:04] Roy: Maybe you should ask someone, who's already created a working C/Java-Interface :) [14:04] Hehe, you where working on a new version right..? [14:05] yes, am working right now [14:06] I could already build the back-end in Java maybe that way it'll drive me to finish it someday instead of waiting for the right time/moment when I could easily connect a fast mars to Java [14:08] Do whatever you want [14:09] But who needs a Java-interface? [14:09] Hehe, I really want to use PyCorewar though [14:09] * Roy needs it! [14:09] for what? [14:09] Like I said, when working 8 hours a day with Java its much easier to drive home and continue using Java for my hobby projects instead of switching language every evening [14:10] ever tried to write a php script, returning javascript, thats creates a page with java script that creates a page in html? *Gg* [14:10] And when using it day in day out you get faster to write something good ;-) [14:11] you cant keep such a program readavble [14:11] No, never tried that, I never write PHP nor more then one line JavaScript [14:13] a question about validating a mars: would you be satisfied, when I review the code, comparing to pmars source and release the commented core function? (i could put in pmars source as comments) [14:13] And Java is one of the most populair home-brew-software-languages so if I put in on SourceForge or something a lot of people could download/see it [14:13] huh? [14:14] Hehe ares, don't do that not much people understand pmars sources :) [14:15] in the first place, i based my program on the code given in the icws94 draft [14:16] for i am not an graduaded professional, i might not know, how to validate somehting, so i belive, it should be proof enough to show both codes...? [14:16] I doubt, that this is possible with such complex code [14:16] the core isnt complex [14:17] its one single function [14:17] except maybe some FIFO handling [14:18] Just run all Koenigstuhl warriors with fixed positions on both cores, if they are the same there won't be any major bugs probably [14:18] why don't you let it fight several thousand fights and compare the results with pmars? [14:18] *both marses [14:19] i want to be 100% sure, 1000 fights is just not the same [14:19] and you wanted _proof_ [14:19] as i want [14:19] then run all possible matches from the warriors on Koenigstuhl [14:19] that's good enough for me :) [14:20] ok, for PyCorewar I additionally count the number of executed instructions and compare them with those executed by pmars [14:20] how could i tell, that every instruction is working the same? eg. ares-JMN jumped only if both values were != 0.... [14:21] You could try making test-case warriors for all instructions..? [14:22] in comparison with pmars results, this could do. would you review my test warriors, for publishing the set on some cw page? [14:26] No..? :) [14:27] i love you, too [14:27] *g* [14:27] * Roy probably doesn't have the patience to review everything [14:27] i did not find very much material for new mars creators, so i want to publish my collected stuff somewehere [14:29] I've started once with the same idea and abandoned it quickly after I realized how much work it would mean [14:29] Running a lot of fights and comparing the results is good enough for me [14:30] ic *ggg* [14:30] validating DAT was easy ;-) [14:31] no, it isn't [14:31] compare (do nothing, do nothing in another way) is not? [14:31] you have to make sure, that all addressing modes are "executed" correctly [14:32] a "dat.f < 1, } 2" still changes the core [14:32] sure... but thats another point on my list... _and_ the validation of inregister is taken from the c-source in the draft [14:32] Things like dat }0,{0 sound very error-prone [14:33] for now i am only checking the opcodes themselves... and if the draft code is correct, my mars should be, too... but thats to be done later [14:33] When I started with PyCorewar I needed to run the whole Koenigstuhl and compare the number of executed instructions just to find, that I've done register-evaluation slightly wrong [14:34] in the first place i wrote a fold() function by my own... and found, that this part was not so easy... i recoded the c program to pascal... sounds like a safe approach to me [14:35] * Roy has no idea how marses work on the inside [14:35] * ares cant belive that [14:35] yes, that's right. Roy has no idea at all ;-) [14:42] ares: Of course you can always use another approach ... release the software and wait until someone complains about a bug [14:44] * Roy is 50% brainless [14:44] why 50%? [14:44] and 45% clueless [14:44] 90% of my body doesn't contain any brain [14:45] Err I mean 50% of the time I have no idea what you guys are talking about, not even what I'm talking about :) [15:02] Ok, time to go! [15:03] On vacation! [15:03] * Roy waves and wishes everybody a happy newyear! [15:04] MSG: Quit: dat [15:16] i take it back, pmars core_is_ complex [15:36] I hope, that it isn't too complex [15:36] i am still trying to understand it.... especially, how JMZ works.... looks weird in the first place [15:37] what is so difficult about jmz? [15:38] i do a jump to PC+aPtr, if botth aAvalue and bValue are zero... pavel told me, that there might be a difference to pmars, and i find the following checks in pmars sim.c: [15:38] if (ADDR_B_AVALUE) [15:38] and [15:38] if (IR.B_value) [15:39] the behaviour depends of the modifer [15:39] i am currently trying to figure out, why there is not ADDR_B_AVALUE && ADDR_B_BVALUE or IR.A_value && IR.B_value or something similar [15:39] jmz.i/jmz.x/jmz.f jump to pc + aptr, if both a- and b-value are zero [15:39] sure [15:40] but what then so difficult? [15:40] i am confused about the different variable names for a and b value [15:40] i suppose there is a reason [15:40] yes [15:41] you should look at how the different addressing modes are handled [15:41] my mars does the evaluation, leaving pointers for the core. then i compare the values of the pointed-to cells [15:41] that doesn't work [15:41] better: i compare the values of the a-instr and b-instr registers [15:41] you should read about in-memory and in-register evaluation [15:42] i read that... and there is an option for inmemory/inregister evaluation in ares... ahh... that might be something... pavel told me, the difference might occur when aAddr = PC or similar... checking [15:43] you shouldn't use an option for inmemory/inregister eval [15:44] why? ICWS94 doesnt say, what to use, or am I wrong? [15:44] icws isnt the standard, pmars is [15:44] yes, that draft doesn't specify, but only inregister is used [15:45] hm... i could remove the option, sure [15:45] it would make things easier, too :) [15:45] i think, it doesnot interfere [15:46] when i started with ares, i didnt know, so i built in many switches... [15:46] http://homepages.paradise.net.nz/~anton/cw/corewar-faq.txt [15:46] look at 17th [15:47] i know [15:47] thats where there is an option, to choose [15:47] maybe there are other communities, playing with in memor eval *G* [15:47] no [15:48] you seem to be really sure [15:48] yes, if I define it not to be possible, it isn't possible [15:48] and you've said to create a tool for beginners. Therefore you shuld adhere to THE standard, which is pmars :) [15:49] with default settings, ares is (except for possible errors) set to standard behaviour [15:49] ... and a beginner surely doesn't know about inmemory/inregister eval [15:49] thats RTFM [15:50] or to be ignored [15:50] no, you can say that only and only if you have a large user base, which cw hasn't [15:50] you have care about the beginners [15:50] ... and kick everybody else ;-) [15:50] youre right... [15:50] * Fluffy kicks bvowk [15:51] can i input code to core directly with pmars? [15:51] yes [15:52] pmars -e warrior.red [15:52] e 1000 [15:52] to edit core position 1000 [15:52] argl... i have a server version here *g* [15:52] lol [15:52] then try to compile your own pmars :) [15:53] i have another version here... you did send me (?) [15:53] pmarsv [15:53] I would never do such a thing [15:53] forget pmarsv [15:53] AAAH! [15:53] just get the binary-pack from corewar.sf.net [15:53] i dont even have a c compiler for windows [15:54] okay [15:54] and make sure to have a backup available, because it contains a virus [15:54] ;) [15:54] i never do backup [15:54] *g* [15:54] real men do no backups [15:54] (in fact i do) [15:55] (sometimes) [15:55] Mizcu: I've always been irreal [15:58] i was redirected to sourcefourge, and there i dont find binaries... i suppose, tar.gz doesnt contain win executabels [15:58] no [15:58] http://sourceforge.net/project/showfiles.php?group_id=174312&package_id=214071 [15:59] thx [16:02] it refuses to run... [16:02] never saw the errormessage before [16:03] MSG: Quit: Leaving. [16:04] why not pmarsv ? [16:05] old [16:06] 0.9.2 any changes in core? [16:06] oh [16:06] not 092 [16:06] 080 [16:07] i want a computer like they have on the enterprise... "computer, create a mars, that is compatible to..." [16:07] "beep - that is not possible" [16:09] other question: pspace[0] holds "-1" on a defeat in the last round, "0" on tie, and "+1" on a win, right? [16:10] another other question: is the mars source of the draft compatible to pmars? [16:23] pspace 0 is -1 on first round, 0 on loss, 1 on win, 2 and higher in case of (multiway)tie [16:24] and pspace not cleared at all between rounds, right? [16:34] nope [16:34] only result is kept? [16:35] not cleared, result changes [16:35] so a warrior can pass eg. 500 values to itself next round [16:37] yes [16:37] however 500 is a bit extreme [16:37] fraction 1/16 coresize [16:37] *g* [16:38] you need only couple values [16:38] i can imagine [16:39] P^3 needs only one plus the result [16:40] of course, if you want to go extreme, you can try selfmodifying switcher (German/Neogryzor made one) or try implementing a three-dimensional array on pspace (Grabun, IIRC did that) [16:40] who wrote the SDL extension? [16:41] its common library, i think [16:41] (i cant get it running, with a windows error message, that tells you nothing) [16:41] i was asking for who updated pmars with sdl [16:41] J (Joonas Pihlaja) [16:42] http://www.cs.helsinki.fi/u/jpihlaja/cw/ [16:52] Join: Fluffy joined #corewars [16:52] :) [16:52] mizcu: thanks for the link - the pmarsw.exe found there is running [16:53] yffulF iH [16:54] foo [16:56] hehe .. that RFC, that you've posted, Mizcu, lead me again to languages such as Whitespace or Brainfuck :) [16:58] It might be time again to try to write a good program in Malbolge. Last time I've utterly failed [16:58] how do i make pmars display more than only the currently executed instruction? i'd like to see the "surrounding" cells, too like in the screenshot on koth.org [16:58] klick on a location in the pmars [16:58] and then press up or down [16:59] or type l100,200 in console [16:59] *pmars = core [17:00] its asking for a pmars.mac [17:00] then copy it into the cwd [17:00] i dont have one [17:00] it is in the pmars sources [17:00] aha [17:00] ahja [17:07] mm.. bacon-cheddar cheesecurls [17:12] is there an option to "restart/reload", or do i need to call pmars from command line again, to debug from start? [17:12] nope [17:12] thx [17:12] gotta restart [17:16] yay, 100% cpu usage when burning dvd [17:16] then try to up it to 200% :) [17:16] sorry, got only one ticker [17:20] stop using windows miz ;) [17:21] hrm [17:21] good morning all [17:21] No, don't do that, Mizcu! Windows are useful, especially because you can see through them [17:21] Morning bvowk [17:21] moorning, bvok [17:21] errr ... [17:21] foo, bvowk! [17:22] Flooffy? [17:22] * Fluffy reduces ares' oxygen supply to 50% [17:22] (BARes) [17:23] * ares activates the secret builtin ares-oxygen-generator [17:24] * bvowk installs windows on ares's oxygen generator control system and giggles as ares turns blue [17:24] LOL [17:25] *blubllbl* [17:25] Nick Change: ares changed nick to ibes [17:25] Part: ibes left #corewars [17:25] huh? [17:25] he's dead jim. [17:26] Join: ibes joined #corewars [17:26] wow - ESC to completely log out?? [17:26] strange trillian [17:26] * bvowk uses the jaws of life on ibes (anything "of life" would have to help in this situation right?) [17:26] (ibes = inactive blucode elaboration system) [17:30] bvowk: what do you think are the chances of having pvk write a MARS in x86asm? [17:30] too high [17:30] lol [17:30] Nick Change: ibes changed nick to ares [17:33] fluffy: I'll ask [17:34] [10:34] < pkhuong> why me?! [17:34] [10:34] < pkhuong> and doubtful. [17:34] [10:34] < pkhuong> We've pretty much screwed, I think. [17:35] bvowk, you can offer to write several DIFFERENT versions for each existing CPU [17:37] [10:35] < pkhuong> ask joonas, actually, he might be more up to it ;) [17:37] fluffy: do I look crazy?! [17:38] I think building deep core would be cooler.. [17:38] depends to the number of all existing red electrons [17:38] using small fpgas [17:38] yes, but they cost a little bit [17:38] you could build a whole pile of tiny redcode units into modern fpgas [17:38] not in bulk [17:38] you can get them for a couple bucks each [17:39] the problem I could see would be a big crossbar if you wanted to have the deep core unit do the scheduluing and reducing [17:39] that wouldn't be cheap/easy [17:40] but pci-express would be cheap and easy if you wanted to run them all straight onto the bus [17:40] hmm [17:40] how good is a cell-processor at integer operations? [17:40] no idea [17:40] and those arne't really cheap [17:40] Playstation? [17:41] yeah, that's a dual core amd64 ;) [17:41] huh? [17:42] playstations are about the same as a low end dual core box [17:42] price wise.. [17:42] Im betting the general computing perf is better on the dual core [17:43] no, I was talking about the playstation 3, which has 6 (or so) working cell-processor [17:43] s [17:44] or whatever you call them [17:44] http://en.wikipedia.org/wiki/Cell_%28microprocessor%29#Synergistic_Processing_Elements_.28SPE.29 [17:48] when I can buy a pci-e board with 64 of them on it and plug it into my machine, I'll buy the cell hype [17:48] until them. [17:48] er.. [17:48] then [17:48] So I'm back at optimizing PyCorewar for those slow CPUs [17:51] * bvowk cracks the whip [17:51] ok, ok, I'm back to work [17:54] theres a difference in JMZ between the draft and pmars [17:55] take the implementation from exhaust [17:55] it does the same as pmars, but is easier to understand [17:56] draft says: JMN.f.x.i jumps, if both Anumber and bnumber of bInst are zero [17:56] yes [17:56] huh? [17:56] JMZ.f.x.i jumps, if .. [17:56] the source shows (IRB.ANumber != 0) || (IRB.Bnumber != 0) [17:57] that's the code for JMN.f.x.i [17:57] and pmars shows !ADDRV_AVALUE) && (!IR.Bvalue) [17:57] and? [17:57] both are the same [17:58] indeed? *recalculate* [17:58] is the ! a bitwise negateion? [17:58] (! value ) is short for (value != 0) [17:59] so theres a difference between the explanation to JMZ in the draft and the shown source in the draft... pmars seems to do as the explanations says [17:59] you confuse me [17:59] explanaination says jumps, if BOTH [17:59] draft example source is (IRB.ANumber != 0) || (IRB.Bnumber != 0) [18:00] must be && [18:00] are we talking about JMZ or JMN? [18:00] JMN [18:01] JMN.f.x.i jumps only if at least one value is not zero [18:01] not in the draft explanation to JMN [18:01] i.e. if not ((avalue == 0) and (bvalue == 0)) [18:02] Join: Roy joined #corewars [18:02] JMZ.I vacation, @countdown [18:02] that's the same as [18:02] (avalue != 0) or (bvalue != 0) [18:02] Hi Roy [18:02] it says "JMN.i [...] jumps, if both [...] are non zero [18:02] ok, but that's not how it is handled [18:03] no [18:03] JMN is considered to jump, if JMZ does not [18:03] also not correct, according to the draft, it says "is not the opposite of jmz" [18:04] * Fluffy sighs [18:04] yes, of course you're right [18:05] but simply use the pmars/exhaust-implementation of JMN/JMZ, ok? [18:05] sure [18:05] :) [18:05] aah this is so confusing [18:07] yes :) [18:07] ares: Didn't you want to have a speed measurement for pmars? [18:08] not with high priority... i was just curious [18:08] for which system? [18:09] speed is not really my concern now... first i want to get my mars 100% compatible, then i should remove the whole IDE and then i'd try to optimize *G* so i will not do this before 2020 [18:10] about 9.8 RedcodeMIPS on a PIII-Mobile with 1133 MHz [18:12] Let's say 10 MIPS, if I kill some programs [18:20] kill windows and it goes to 11 [18:20] Windows? Which windows? [18:21] Hrm, just found out the stupid people on our stupid disco wheren't just doing stupid jumps... it is apperantly a new hype in Holland :| [18:22] jumps? [18:22] JMN [18:28] http://www.youtube.com/watch?v=U6vwhsxBTBA [18:28] hrm [18:29] now thats retarded [18:36] add a little bit of duch bigboom hardcore and you got prime example of idiocy people think cant happen in europe [18:36] replace(duch, dutch) [18:52] Heh, note it started in Belgium and France ;) [18:55] PyCorewar is too fast :-( [18:55] ... and I don't WHY [18:56] is that supopsed to be a problem? [18:56] no, i dont write bad, my shell connection lags [18:57] because of the high-speed warez-operation in the background [18:57] Yes, it is, because that leads to problems when optimizing the implementation of the MARS [18:57] go away bad thought! go away! i am not making an evolver! [18:58] lol [18:58] fluffy: its impossible for it to be too fast [18:58] Mizcu: Evolved '94nop-koth :-) [18:59] bvowk: Yes, but I don't know why it is so fast. That's the problem. [19:01] floff: rather on nano or tiny.. [19:02] Join: fiveop joined #corewars [19:04] Hi fiveop [19:05] Mizcu: Ok, then think of a lovely fluffy nano-warrior :) [19:05] bah, ill pop up cd of Monty Python on oven and couple Habaneros on drive and think it over [19:05] i might find my text-file'o thoughts somewhere [19:06] hehe, I'm stupid, now I know why it is that fast [19:07] 'caus your stupid [19:07] yep [19:08] hmm, now it is even faster [19:08] I don't understand that :( [19:13] now I know that I'm really stupid [19:14] the whole time I've benchmarked the wrong version, because I didn't install the "new" version, but only build it [19:18] hey fluffy.. [19:18] so I've got a sweet ass naming system I think.. [19:18] but you're not going to like it.. its all hexy.. [19:19] Does it mean, that your evolver is up, running and sending warriors to nano? [19:19] there, habaneros in the oven drying [19:19] anybosy want a poke in the eye? [19:19] think I should make a dictionary of all the possible families of warrior and then translate the sexy hexy stuff to latin name? [19:19] right now its navel gazing. [19:19]