[07:38] Join: Fluffy joined #corewars [07:39] :-) [08:00] boobies [08:00] Hello Fluffy [08:00] Hi datagram [08:00] How goe sit [08:00] goes it, even [08:01] Don't know. Waiting for sth. to happen [08:28] Join: Mizcu joined #corewars [08:30] Hi Mizcu! [08:34] sheesh, you know day is going to start well as you wake up from a dream of 'puter getting busted [08:35] ouch ... dreaming of computers ;-) [08:59] Heh [09:00] yes? [09:00] no? [09:00] Tuesday! [09:01] Canada! [09:01] Join: Nenad joined #corewars [09:01] Hi [09:01] No, that will soon be invaded by the US. There is no need to mention it any longer, Mizcu [09:01] Hi Nenad! [09:02] :) I just began running the rf25 battles. The results will probably come soon, as well :) [09:03] I've just started to update my pages about the history of 94nop. I've already downloaded the data from JM. If anybody has the time to get the data, I'd be very interested in all challenges after HullabaTwo became 316 (Jan 13, 2006). Just mail me :-) [09:03] My computer was "fixed" two days ago - they erased the disk and reinstalled the os. :( without making backup of the files on the disk.. Hopefully, I had copies of most of the files, but - I lost a lot of data, as well [09:03] that's bad! [09:04] Do you have all the challenges? JM had told me before that he didn't have some, and I offered to give them to him, but he forgot to mention which ones [09:04] Never let someone else fix your computer [09:05] well, I've learned that lesson, now... [09:05] Nenad: John and I think, that we have all the data before Jan 13, 2006, that is available. [09:05] Nenad: John has already lots of data after that date, but some parts are missing. [09:06] yes, but do you know which parts? [09:06] Nenad: The first gap is HullabaTwo@age 316 (16 Jan) - HullabaTwo@age 363 (Feb 1) [09:08] Next gap: HullabaTwo@410 (Feb 17) - HullabaTwo@413 (Feb 17) [09:08] Oh...there was a rf... [09:09] I didn't even know! [09:09] well, I can probably send you those reports, although I'll send some garbage, as well, cause I'm too lazy to separate them from the reports from other hills [09:09] Nenad: I have no problem with that :-) [09:10] I'm finishing up all my class projects for all my courses this week...after that I can upload the spring tourney page [09:10] spring tourney? [09:10] If anybody has challenges after Feb 17, I really want them :-) There is no need to sort the data. I'll delete everything, that's not a challenge immediately [09:11] datagram: spring tourney? [09:11] Fluffy, I'm hosting a tournament [09:11] heh, its been in the topic for a few weeks [09:12] datagram: I know, but can you give specific infos? [09:12] As far as? [09:12] Think of it like the 2005 tournament...but in spring. [09:12] datagram: The usual. Who will win? Whom to bribe? [09:12] datagram: Spring is almost over. I'm already heavily in summer ;-) [09:13] *shrug* [09:13] Its always summer here [09:13] but that doesn't bother me [09:13] : ) [09:13] Where is that "here" again? [09:14] west coast USA [09:14] los angeles [09:14] hmm ... that would be too much sun for me [09:14] Its alright [09:14] Fluffy, You can just stay inside : ) [09:15] I could, but I won't [09:15] err... Jens, do you have any idea how I should copy those messages and send them to you? I'm using Outlook, and it stores all the messages in one huge file... and it will take ages to copy them one by one [09:16] Nenad: Is that huge file some kind of text file? [09:16] *.DBX [09:16] Fluffy, Nenad If I went through the trouble of making a web interface for submitting warriors for tournaments would it be used...? [09:16] datagram: yes [09:16] datagram: no [09:16] :) [09:16] (i wouldnt..) [09:16] Roy enabled php for me, so I might fancy actually using it [09:17] I don't think I have any db though, so just putting stuff into text files seems like is what's gonna happen [09:17] Nenad: I've just found a way to read such dbx files :-) [09:18] Jens: ok. just wait a sec then [09:18] Nenad: Thanks! :-) [09:21] datagram: I think, that there are too few people out there. Writing a web interface might be good for a new hill, but not for such a "short" event. But if you already have that interface, you should use it [09:22] I mean I can write it and then others can use it for future things also [09:22] datagram: Then I say: "Why not?" If it is good, I might use it [09:24] Altough PyCorewar still has a long way to come, I might create a hill. I'd then use Python/PyCorewar for doing all the work and a little PHP-interface on my homepage [09:24] I see [09:24] I gotta work on these projects [09:24] bbl [09:24] bye [09:25] bye, datagram [09:25] Jens: this is strange [09:25] what? [09:26] interesting, Ibm had thoughts of resurrecting butterfly -keyboard in 2003 [09:26] Havent seen any coming [09:26] well, it seems that there were MANY challenges [09:26] Nenad: yes, I know [09:26] I can't upload that [09:26] it's 14 Mb [09:26] oh [09:26] and my connection is 2kb/sec [09:27] Yes, I know the problem [09:27] I'm @64 kbits/sec [09:27] well, it seems that I'll be able to give you those files when we meet in Holland, if Roy decides to organize that meeting :) [09:27] Nenad: Then we'll make a deal. Please don't delete that data. I'll try to get them otherwise [09:28] Jens: I never delete reports [09:28] Nenad: Wonderful :-) [09:28] Nenad: I guess, that you've already tried to compress the file [09:29] actually... :) [09:29] I haven't [09:29] Hmm ... if it is just a text file, compressing it might be a good idea [09:29] it helped, enough [09:29] 2 Mb [09:30] at which email do you want me to send it? [09:30] [09:31] Hmm ... I should try to figure out, which limits this mail address has [09:31] But 2 Mb is not a problem [09:32] hmm ... max. 20 MiB/mail [09:35] well, I'm sending it now... [09:35] :-) [09:35] Why do you only have 2 KB/sec? [09:36] a few more battles and I will start calculating the scores according to the rf25 formula... and then we'll know the winner [09:36] vodoo [09:36] I mean, is it usual to have such connections [09:36] how the hell should I know? old phone central [09:36] probably [09:36] but dial up internet is very slow here [09:37] I'll try getting adsl in a month or two [09:37] :-) [09:38] Jens: just for the fun, what is your prognosis for the final rankings in this round? [09:38] Who has submitted warriors? [09:39] the participants: me, Zul, Fizmo, Robert Macrae, Chip Wendell, inversed, Lukasz Adamovski, Roy [09:39] hmm ... 1. Roy 2 Chip 3. inversed [09:39] and of course: 0. Nenad [09:40] :) [09:40] I think that Zul will be high placed [09:40] but don't know if he'll win or not [09:40] he tried a tricky approach [09:41] which certainly gains a lot of points, but a fixed amount of points - which is not the maximum possible [09:41] I've tried one, too: Forgetting about the tournament [09:41] :-( [09:41] he submitted two dat's [09:41] :) [09:42] Did anybody submit complex warriors, say > 20 lines? [09:42] yes, sure [09:42] I did, Fizmo did, Roy did... [09:42] haven't looked at warrior code, so I'm not sure for the others [09:43] but yes - at least one entry was usually big, and the losing entry small [09:43] grrr ... I hate it to wait for (snail) mail [09:44] Nenad: not your email, but 2x256 MB RAM [09:44] I have here a complete computer. Only the RAM is missing. So I can't do anything with it :-( [09:45] Nenad: Any results yet? [09:45] :( [09:46] well, at least I didn't pay for much it [09:50] It seems, that I'm getting a big email :-) [09:58] Join: Roy joined #corewars [09:59] Hiya guys [09:59] Hi Roy [09:59] Running the tournament ey Nenad? [09:59] Hey Roy [09:59] I didn't really put much time in it, so I won't be to dissapointed ;) [09:59] Roy: you will probably be the last :) [09:59] (about one hour or something) [10:00] you misread the rules, didn't you? [10:00] Probably :) [10:00] Never tested anything :P [10:00] your warriors need to score 50:50 or close when combating each other [10:00] Too complex math stuff etc, just submitted two warriors [10:00] you submitted a paper and a stone [10:00] lol [10:00] :) [10:00] Haha [10:01] Roy, Do I have mysql on there? [10:01] err.. no idea, I think it has [10:01] heh [10:02] * Roy wants a nano-lp competition! [10:03] Or a six-line competition (so much more you can do with that) [10:03] Roy, Perhaps the spring tournament will have such fancies... [10:03] <_< [10:03] >_> [10:04] Thats why I'm hinting it right now ;-) [10:04] Hehe [10:04] Who woulda thought... [10:04] Planting ideas in your head [10:04] I'd have everything done by now if it wasn't for Kingdom Hearts 2 + work/school [10:04] Mostly KH2, though +_+ [10:05] hehe [10:05] Where should I announce the tournament? [10:05] Besides here and the google dealy [10:06] Roy: you might still not be last :) Fizmo did something similar, if not worse [10:06] datagram: at rgc [10:06] hehehe [10:06] data: right here [10:07] Fizmo's site, maybe even koth.org, JM's site [10:07] Slashdot.. :) [10:08] btw, I still have to think of a way to give % scores in the end [10:08] there will be people with score < 0 [10:08] * Roy [10:08] even with score < + max, I think [10:08] Nenad: then you should use wrap-around-math, e.g. -1 --> 65535 :-) [10:09] :) [10:09] yeah! [10:09] that was just the loophole I was aiming for :P [10:10] mts.exe (windows cw tournament runner) has the same problems...if I run nano with a lot of warriors and too much rounds it gives scores like -235236363 [10:10] Then don't use mts.exe :-) [10:10] btw. Fluffy, PyMars.py has a similair problem! [10:10] If a warriors name contains something like "235-13652462-723" it does crazy stuff [10:11] Roy: Could you give an example? [10:11] And did you read in the log about the slowness with nano..? Is that because of the python-launch-time? [10:11] Roy: Yes and no. [10:11] I think all bvowks crazy evolved names messed it up (his nano-warriors), had to rename them [10:12] Roy: No - because you are using pymars the wrong way. It was never intended to be started again for each fight. [10:12] err .. No --> Yes [10:12] Yes, thats what I said then [10:12] But its just handy to use just as pmars ;) [10:13] It isn't a problem, if you use it the "normal" way and create a parser and a mars once (!) [10:13] Roy: I'll soon start to rewrite the parser and then (hopefully) all problems with will be gone. [10:14] hehe [10:14] * Roy does a little dance [10:21] Is it that tight between me and Fizzer for last place..? [10:22] .. still early to tell [10:22] there are a lot of calculations to be done [10:22] Mizcu: adding a simple add makes a lot of difference, the double-loop-forward-thing gets better all the time ;-) [10:23] Yeah... but you tested stuff I think [10:23] Roy: its not a double-loop. Just single loop that extends on occasion. [10:25] Well ok, but thats not new, OoS does the same [10:25] We just made the OoS of Papers [10:25] o_O [10:26] i didnt expect that kind of comparison [10:26] But it is kinda like it right? [10:27] more like a scanner with zoom-trick [10:39] everyone was stunned by thinking what kind of behemot-of-a-code we have created? [10:46] ... I'm exhausted from performing all these calculations... [10:46] but I have the average score for 3/8 authors so far [10:46] Nenad: Doesn't your computer do all the calculations and you just wait? [10:46] no... I do the calculations and it waits :) [10:46] oh [10:46] and I have to double and triple check it [10:46] But it should be other way around [10:47] :) didn't you watch matrix [10:47] Nah, I've taken the wrong pill [10:47] anyway, Roy - I've got your average score [10:47] and mine [10:47] and Zul's [10:48] according to what I've seen so far, it is very likely that Chip will win, but we'll see [11:04] Join: fiveop joined #corewars [11:05] hi fiveop [11:07] hi [11:07] Hi fiveop [11:10] Nenad: hmm ... I can almost read your file, but I get some strange errors and some scores are impossible :-( [11:10] Nenad: I've never seen a koth with only 2 points ;-) [11:11] MSG: Quit: CGI:IRC (EOF) [11:13] :))) [11:13] how did that happen? [11:14] Probably my converter cannot handle every aspect of the dbx-format :-( [11:14] just open it with Outlook, you'll get theproper readings there [11:14] haha [11:14] or make a script which handles it :) [11:15] I gues, that this is the easiest solution, but I can do it only later today [11:15] *guess [11:16] I'm that lucky to have no Windows-box at home :-) [11:21] 5/8 of the work complete... [11:21] however, I'll have to stop here and continue tonight, since I have to go to attend some lectures soon [11:21] :-( [11:22] the average top score at this point is 325.98 [11:22] I meant - top average score [11:22] the worst score? [11:22] -198.98 [11:23] I think that I'm simply going to add the worst score to all scores, so that the last player gets 0% [11:23] and calculate the rest as score/maxscore after tha [11:24] I won't give away any more details now. Don't want to spoil the surprise. :) [11:24] I'm off now [11:24] :) [11:24] * Nenad waves [11:24] * Fluffy waves [11:24] MSG: [11:39] Join: Roy joined #corewars [11:39] bac2 [11:39] *back [12:16] MSG: Quit: CGI:IRC (Ping timeout) [12:35] Part: Fluffy left #corewars [14:04] MSG: [14:10] MSG: Quit: humhum [14:11] Join: fiveop joined #corewars [14:13] Join: Roy joined #corewars [14:16] Mizcu is probably going to be surprised... just send a warrior to the lp hill, lets wait and see! [14:18] Yay 6th place :) [14:44] MSG: Quit: CGI:IRC (Ping timeout) [16:03] Join: Core29 joined #corewars [16:34] Join: Fluffy joined #corewars [16:57] Join: Roy joined #corewars [17:31] MSG: Ping timeout: 252 seconds [17:51] Join: sf[2] joined #corewars [17:51] Hello [17:52] Join: sf joined #corewars [17:52] hi :) [17:52] Hello [17:52] Hi Roy [17:56] sf/sf[2]: What are you playing here? [17:56] MSG: Read error: Connection reset by peer [17:56] heh, sf is playing with herself :P [17:56] I've disconnected on sf[2] [17:56] Am I still here [17:56] I see [17:56] I think I mean sf disconnected [17:57] Nick Change: Roy changed nick to sf [17:57] Hi! [17:58] * sf[2] kicks sf [17:58] Nick Change: sf changed nick to Metcalf [17:58] Nick Change: Metcalf changed nick to Jens [17:58] Nick Change: Jens changed nick to Coughy [17:58] Nick Change: Coughy changed nick to Roy [17:58] * Fluffy sighs [17:59] Coughy? [17:59] I'm having problems again [18:00] I can't see anything in the web client, but fortunately I can read the logs [18:01] Heh, sf hold on, I got a nice webirc thingy today [18:01] http://pc-intern.com/artikel-3864.html [18:01] I especially like the [18:01] 'dutch soccer voodoo doll' [18:02] Join: sf[i] joined #corewars [18:03] Nice, it works quickly [18:03] Yes, works almost instantly [18:03] * Roy uses it at work [18:03] :) [18:09] Join: Neogryzor joined #corewars [18:09] Fluff? [18:09] sf: I've replied to every of your /msg [18:09] Hi Neo :) [18:09] Hi all [18:10] hi sf, what's up today? [18:10] Replies don't seem to be coming through to either of me :( [18:11] ? [18:11] hi neo [18:11] hi Roy [18:12] i saw a new name at nano. Anyone knows who could be? [18:12] Neogryzor: bad connection :( [18:12] hmmm.... :( [18:14] I've no idea who hwm or raptor are probably both new [18:15] That's a nice oneshot for a newbie. Not sure what raptor's warriors are [18:16] terrafirma... that sounds like a name from an italian guy... [18:17] Well, a good idea on nano can bring you pretty far [18:19] btw, congrats Roy, you did a nice work there. Wich evolver did you use? [18:19] My head :P [18:19] MSG: Quit: CGI:IRC [18:19] Handwritten so? [18:20] Its mostly handcoded, but constants are bruteforced [18:20] (against the whole koenigstuhl nano) [18:20] Good, welcome to the anti-machines club :) [18:20] * Roy is a cyborg [18:20] heh [18:20] Part machine part handcoded :) [18:21] * Neogryzor is an alien [18:21] But my oldest and best warrior is still evolved.. [18:21] Human first, machine after that is ok [18:21] Now I'm completely in LP [18:22] Mizcu gave me a crazy idea I'm working on, different kind of replication [18:23] Oh, you are going to force me to work again in LP... :-/ [18:23] Heh, against the top20 koenigstuhl it doesn't score as much as Digital (yet) [18:24] Why didn't you give digital rain a qscan btw..? [18:25] btw sf: I went to Radiohead this thuesday :)))) [18:25] uh? that's not my warrior... [18:25] Probably because he didn't write it? [18:25] Err oops no, thats inv. :) hehe sorry [18:25] You where the king dude :P [18:26] wich warrior are you asking about? [18:27] You made the King, inversed made Digital Rain... [18:27] (switched that in my head, I remember you as the LP-king) [18:28] Join: Mizcu joined #corewars [18:28] Oh, i forgot that one... it was probably a bit too big to include a Qscan [18:28] hi Miz [18:28] hi NEo [18:29] Mizcu: . [18:30] . [18:31] Mizcu: Maybe we should start to extent that way to communicate ;-) [18:31] Mizcu: dir /p [18:31] I answered straight away, and I have just answered straight away to this on (at 19:31) [18:31] sf[2]: Something's really wrong with you connection ;-) [18:32] Something is very wrong, but I can read the logs ok :) [18:33] sf[2]: I don' think, that this is the proper way to use IRC ;-) [18:33] sf: ... and you can send us your comments via nano warrior names :) [18:34] When I enter something it mostly appears straight away [18:34] Neogryzor: Unknown command or bad file name [18:34] dis appears..? [18:34] But anything anyone else writes takes ages to get through to me :( [18:35] And I can't access my email [18:35] Mizcu: format c: [18:35] Nope, wont work [18:35] I want to throw the computer outside then set it on fire [18:36] sf: Do it [18:36] sf[2]: I don't think, that this will improve your situation. [18:36] Neogryzor: then I wouldn't be able to take to you :( [18:37] sf: sell the pc and buy a homing pigeon! [18:37] haha, i was about to say exactly the same! [18:37] hehe [18:37] Roy: The ping times are awful, but data transfer rate is ok [18:38] Roy: will bvowk accept nano entries by homing pigeon? [18:38] I think he will...but you won't get your pigeon back.. [18:38] sf[2]: Are you sure, that your pigeons will reach Canada? [18:38] Knowing bvowk he'll just want to have a massive warehouse full of nanotyping pigeons [18:38] The'll take a flight... :p [18:43] sf[2]: The logs have failed and we've revealed all the secrets of corewar, but won't do it again ;-) [18:44] heh [18:45] You don't know the secrets of Corewar fluffy! [18:47] * Fluffy will never say a word again [18:47] Do you know the secret of Reepicheep's second silk? I bet not :P [18:48] do you, sf? [18:48] * Roy knows! [18:48] sf[2]: It's name is Bob! [18:49] Do you know which hill contains an unpublished self-repairing warrior? [18:49] nano! [18:49] sf[2]: Yes. [18:49] Keep it to yourself then ;) [18:49] Roy: Be polite and always answer the question(s)! [18:50] * Roy will [18:52] you are making me feel like a newbie... [18:52] Heh, only n00bs say that.. noob.. :P [18:52] (just kidding :)) [18:52] I have no idea which hill has a self-repairing warrior... [18:53] Please, sf, tell us [18:59] Neogryzor: you should set up a new webpage [18:59] i know, i know... :'-( [18:59] Neogryzor: I use atspace.com and it works fine, no problems, no ads, it's free and you are allowed to put your own ads on there [19:01] Unfortunately i don't have much free time lately.. [19:02] :( [19:02] ok, here is something Mizcu came with yesterday and I refined: [19:02] rStart spl @4072 , >rEnd+1 [19:02] sub.x rEnd , rStart [19:02] rLoop mov.i mov.i }6716 , {6633 [19:02] add.f rEnd , -1 [19:02] jmn.b rLoop , rStart [19:02] rEnd djn.f rStart , >1761 [19:02] I wanted to rebuid all from the ground in the PC and once finished upload all [19:02] A new replicator thingy! [19:02] (start at rLoop btw) [19:03] This version bombs like a stone (adding and bombing) [19:03] Does not work in a similar way to usual LP papers? [19:04] No, it copies in the loop below (while attacking) [19:05] Its kind of like OoS (IMO) it loops attacking and when is has copied the warrior it splits there and resets everything and continues [19:05] i see you don't need the mov.ab #,# in the head :) [19:06] Nope, using the sub :) [19:07] That replicator is impressive compared to the other LP replicators I've looked at [19:08] Is it on the hill yet? [19:08] Yup, 6th (but with a qscan and imps) [19:11] It also beats 94nop papers :P [19:11] :-O [19:12] But normal LP papers do that too I think.. [19:12] they do [19:12] i didn't know that [19:13] * sf[2] waves [19:13] * Roy waves [19:13] see you [19:13] * Fluffy waves [19:13] * Neogryzor waves [19:13] MSG: Quit: Physical Wreck IRC webscript: http://www.pswclan.com/cgi-bin/irc/irc.cgi [19:14] Join: sf joined #corewars [19:14] That didn't work very well :( [19:14] * sf tries again [19:14] MSG: Client Quit [19:20] i have to go too [19:20] bye neo [19:20] * Neogryzor waves [19:20] * Fluffy waves [19:20] MSG: [19:27] Join: johnkw joined #corewars [19:29] Hi johnkw [19:46] hey fluff [19:47] Hi bvowky :-) [19:47] How is your pigeonfarm? [19:47] lol [19:53] pigeonfarm? [19:54] fiveop: Read the logs! It is about a speedy way to transfer data ;-) [19:55] I remember [19:57] bleh. [19:58] that's not an answer, bvowk [19:58] It is..! [19:58] * Roy answers with 'bleh.' all the time! [19:58] eh? [19:59] Bleh in dutch is like uhm.. 'yuk!' [19:59] So if you don't like something, we just say "bleh!" [20:00] Roy: No, I don't say that. I usually avoid it then [20:00] bleh.. [20:01] * Roy is bored [20:01] and tired too [20:02] Then start to plug out your left eyes. That's always fun! [20:02] *eye [20:03] Roy: And? [20:03] i trhrenk i'nb blinw! [20:04] Do you have only one eye? [20:05] Only my two left eyes [20:09] * Roy waves [20:09] MSG: [20:09] * Fluffy waves [20:09] ... too late [20:20] * bvowk stabs fluffy in the eye with a mechanical pencil and then beats him about the head and shoulders with a steel industrial slide rule [20:21] I see. [20:21] thats for not having pycorewar done for me : [20:21] i have a vague feeling that bvowk might have issues with anger management [20:22] its the violent computer games.. [20:22] I'm a slave to my highly honed killer instinct now.. [20:22] consider a career in army [20:23] bvowk: Try stroking fluffy bunnies. That might help [20:23] Mizcu: Have you every heard of the Canadian Army? [20:23] *ever [20:23] I used to butcher and eat fluffy bunnies.. they're great with shake'n'bake [20:23] lol [20:24] Well, shocktroops are at high demand in usa [20:24] hey, the canadian army exists, they put a base next to my techincal skool and the bastards kept hogging the pinball machines [20:24] if I was going to be a gun toting killer, I'd at least freelance.. [20:24] no reason to be getting shot at for government wages [20:25] What a strange thing to say? [20:25] * ? --> ! [20:25] what part? [20:26] bvowk: "no reason to be getting shot at for government wages" [20:26] Do you really want to be shot for more money? [20:27] not really.. [20:27] but if I was going to be getting shot at, I'd at least negotiate my own contract! [20:27] I would usually run away [20:27] Then join carnivale as assistant [20:28] I dunno, I might make a good long range sniper :) [20:28] So far it worked [20:28] although I'd have to curb my habit of shouting "w00t!" [20:28] Does than mean you like laying down in same position for days? [20:29] sitting does not count [20:29] well, lets just make it clear, I'm not going to be out shooting people any time soon.. so thats kinda a moot point :) [20:30] s/any time soon/at all ever/ [20:30] Is everything nicely monitored there in Canada, too ;-) [20:32] why not, if you can do all of america for pennies, why not throw in canada for free? [20:32] although, if I ever find out about it, we're loading up the pickup truck and we're going to lynch the prime minister on the front lawn of the parliment building [20:33] Mizcu: He definitely has a problem! [20:33] Atleast not a big as with full-time techsupporters [20:34] hmm ... that reminds me, that I haven't read userfriendly today! [20:37] Join: pkhuong joined #corewars [20:37] pk! [20:37] bv! [20:37] Hey pkhuong! [20:37] go write code! [20:37] * bvowk cracks the whip! [20:38] :) [20:38] heh. I got lit to read, and I don't understand the big picture [20:38] I need code to run, so you two have to get to work :) [20:39] bvowk: I could give you lots of code to run! [20:39] oh? [20:39] got another set of patches? [20:39] bvowk: MARSes should rawk on T1 [20:39] yes they should [20:39] I should sit down and really assimilate the standard, but it's got lots of really scary symbols. [20:40] get some code and he who must not be mentioned will start stuffing those up [20:41] he won't start coding until you've gotten started, he's a theoretical guy until there's code :) [20:41] yeah, but I can't write code when I don't know how i'm supposed to pipe the s-boxes, lin tran and shufflers together [20:41] :p [20:41] its true! [20:41] once you show him something thats code.. he switches from math mode to uber-l33t coder mode.. [20:42] you've seen it happen! [20:42] true, true :) [20:44] sounds like Metcalf [20:44] metcalf is uber leet coder all the way :) [20:45] he who must not be mentioned was trained as a mathematician first.. so he really does have two modes.. [20:45] its like a switch that you can flip on and off.. [20:45] heh, same here. I'm a coder first, but my dad wanted me to be a mathematician ;) [20:45] MSG: Quit: Trillian (http://www.ceruleanstudios.com [20:45] obv, he failed. bwahaha :D [20:46] I'm tempted to go get a math training too, partly because I work here, and partly because it would utterly annihilate my parents minds.. [20:46] but work-wise, I think maths is a better choice. [20:47] yeah, you don't have to write code, you just need to prove theorem x is possible under a certain edge case... [20:47] no PoC :) [20:47] that's why I study maths and not cs [20:47] heh, double major :D [20:47] * bvowk is *SO* going to hear it for that [20:47] bvowk: You usually prove, that exists at least one way to do it ;-) [20:47] or you disprove that there isn't any way [20:48] fricken non constructive proofs. [20:48] pkhuong: starting to study cs in addtion in october [20:48] heh. Officially, I started studying CS this semester ;) [20:48] bvowk: read "The Haskell Road to Maths, Logic and Programming" as a start [20:48] fiveop: I did :) [20:48] I was being a shit disturber with that comment... [20:48] knowing the number of math people who read the logs [20:48] hm? [20:49] erg.. [20:49] they just killed my ac.. [20:49] mitzi gonna die again? [20:49] maybe [20:49] bbiab [20:51] stupid compiler! [20:52] yeah, too bad compilers aren't even sufficiently smart. [20:52] Let me guess ... there isn't another compiler for Linux apart from GCC and ICC, right? [20:53] language? [20:53] C [20:53] there's something based on wacom [20:53] *watcom [20:53] Is it any good? [20:53] watcom was good in its days. [20:53] That means no? [20:53] That means I have nfi. [20:54] You should spell it out. This channel needs to be banned everywhere. We'll get more new Redcoders that way ;-) [20:54] banned? [20:54] who are you referring to? :) [20:55] the guvmint [20:55] I'm talking about using bad words like "j*****" [20:55] :) [20:56] there's something called mgcc in the ubuntu package system [20:56] hm [20:56] what's your problem? [20:56] and what does it do? [20:56] perhaps it was some other name starting with m ;) [20:56] My MARS is too slow! [20:57] that's the problem and it will always be that problem [20:57] do you need a compiler that produces faster binaries? [20:57] mm.. looked at the asm output to see the problem? [20:57] yes, but there is only one obvious answer [20:57] code everything in asm :-( [20:57] no, just the part that isn't right. [20:58] yes, everything [20:58] got code? [20:58] oh wait, can't, bv won't be happy ;) [20:58] bvowk will be happ [20:58] y [20:59] my 88 MARS is already 10 % faster than fmars, but the 94nop is a little tricky bastard [20:59] :) [20:59] oh, btw, the branch conversion, the "problem" w/ the non optimising optimisation is that you wrap around v. infrequently [20:59] *re the [20:59] yes? [21:00] so the very rare branch mispredict was better on average than the constant cost of masking and shifting. [21:01] which means, that there is no better code? [21:01] not likely :/ [21:01] constant coresize is important, though. It lets the compiler do shift/sub reduction on the divisions for modulo. [21:02] what does vtune say, btw? [21:02] hmm ... maybe I'll add some classes for the most common settings [21:02] Nothing, because I haven't used it so far [21:03] Sascha was so kind to find better algorithms for MARS and I only needed to code them [21:03] *for the [21:03] because if mispredict on the opcode dispatch hurt a lot, it might be worth it to run 2 marses in parallel (and dispatch on both marses at the same time) ;) [21:04] you mean in one thread or in two threads? [21:04] in 1 thread [21:04] nah, won't work [21:04] won't fit in L1? [21:05] won't fit into L2 [21:05] our code is already too big for L1 [21:05] might wanna uninline some of the stuff ;) [21:06] no, we've partially unrolled the loop and added several switches, which was a very good idea [21:06] unroll the loop? [21:06] do ... while (--cycle) [21:07] in this loop there is only a big switch (on isntruction) [21:07] and we've simple added some more switches [21:07] and some checks to abort properly [21:07] k. [21:07] why the unrolling? [21:07] speed! [21:07] erhm... how? [21:08] ok, that's a long story: [21:08] it is all about the process queue [21:09] there is a data structrure (pqinfo), which contains pointers to head/tail of process queue + number of processes [21:09] so we need to access pqinfo->head to get the next instruction [21:09] yes. [21:09] execute the instruction [21:10] and the switch to the other warrior [21:10] k. [21:10] by adding using two big switches, we already know, that the first switch is executed only with instructions for the first warrior ... [21:10] (you know what I mean) [21:10] yeah [21:11] and the second switch is only used by second warrior [21:11] but if it's a jump tabled switch... [21:11] so we've instroduced head1,head2,tail1,tail2 [21:11] and uses theses values directly [21:11] k. [21:11] seems like a v. small gain to me. [21:12] on my celeron that got me from 9 to 12 MIPS [21:12] on bvowk's ned (AMD Athlon X2) it got me from 44 to 57 MIPS [21:12] ok, that version uses 4 switches [21:13] 2 switches war 55 MIPS [21:13] *was [21:13] err ... I've forgot one thing [21:13] we've also changed the process queue [21:14] we have interleaved the process queues of the different warriors [21:14] that way we only need one head pointer [21:15] Do you know, what I mean? [21:15] yes, again, don't see the gain here :) [21:15] but re 4 switches, I guess short cycles in instructions help w/ perf. [21:16] It seems so [21:17] yay, at least that part of our random ramblings may have paid off :) [21:18] At the moment we have 16 different MARS [21:19] and you can pick the one, which performs best on your machine [21:19] Fortunately I only have to put big chunks of C code together to get the different versions [21:20] That will introduce the art of optimizing the MARS :-) [21:22] heh, get bv to evolve the best set [21:23] MSG: Quit: humhum [21:23] yes, but before I'll do that I'll add some new options. 16 MARS are way to few to evolve ;-) [21:23] ;) [21:23] Might be a good idea anyway. I could demand money for telling you the best set of options for your CPU ;-) [21:25] Evolving MARS:es.. [21:26] hmm ... I know, that they've already tried to evolve small pieces of asm-code for things like sorting, but I don't think, that evolving whole MARS will be doable soon [21:26] In any case I've learned, that we have way too many instructions. [21:27] Only some 300-400 different instructions are used for all warriors on Koenigstuhl [21:27] oh, no, it's risc all over again. [21:29] No need to worry. I'll add new instructions soon :-) [21:29] Then we only need a hill, which uses them [21:29] ok, ok. risc+vector insturctions :/ [21:30] no, we'll have "spz", "spn", "ijn" and "fluffy" [21:30] I've added the last instruction only to annoy bvowk [21:41] Time to get some sleep. [21:41] * Fluffy waves [21:41] MSG: Quit: Leaving. [21:45] Join: Sylv-- joined #corewars [21:45] Nick Change: Sylv-- changed nick to pkhuong- [21:47] MSG: Ping timeout: 252 seconds [21:53] MSG: