[00:12] what is a q^i qscan and how is it different from a q^4.5 scan? [01:04] Join: Caelian_ joined #corewars [01:08] MSG: Ping timeout: 240 seconds [04:40] MSG: Ping timeout: 240 seconds [06:35] Join: fiveop joined #corewars [08:27] Nick Change: Caelian_ changed nick to Caelian [08:35] Join: Core_old joined #corewars [08:37] Join: Core_old_ joined #corewars [08:40] MSG: Ping timeout: 240 seconds [09:25] Join: Caelian_ joined #corewars [09:26] MSG: Ping timeout: 240 seconds [09:26] Nick Change: Caelian_ changed nick to Caelian [11:26] MSG: Ping timeout: 240 seconds [14:53] Join: John joined #corewars [14:53] Hi :-) [15:36] MSG: Quit: ChatZilla 0.9.78.1 [Firefox 2.0.0.3/2007030919] [17:16] MSG: Read error: Operation timed out [17:19] Join: fiveop joined #corewars [17:50] MSG: Quit: Ex-Chat [17:51] Join: Caelian joined #corewars [18:38] Join: Fizmo joined #corewars [18:38] hihi [18:47] h [18:47] Join: Fizmo_ joined #corewars [18:49] MSG: Ping timeout: 240 seconds [18:49] Join: Leniad joined #corewars [18:49] Nick Change: Fizmo_ changed nick to Fizmo [18:50] what is a q^i qscan and how is it different from a q^4.5 scan? :P [18:50] we dont know [18:50] You see, Inversed hasnt yet released any code about it [18:50] but it seems better than q^4 [18:51] oh, I see [18:51] that'd explain the lack of info on it :P [18:52] and now that you mention it, all the q^i on 94nop are by inversed :O [18:55] btw, what ide do you use? I like ARES myself [18:57] dos edit + pmars, or corewin [19:00] MSG: Ping timeout: 240 seconds [19:08] Join: Fizmo joined #corewars [19:09] Leniad: do you found this? [19:09] http://www.corewar.info/lexicon/qscan.htm [19:09] the q^i seems not published so far [19:10] yes :P in fact it's been my main reference for qscans [19:10] if you work with q^4.5 [19:10] the positioning of all your "hot" parts is quite important [19:10] it can improve your score by more than 3 pts [19:11] there are some "magic" locations to put the code [19:12] for example around 18 instructions above the main qscan body is a good point to put some code there [19:14] or maybe try my q^4.5b which seems to give slightly more points [19:14] you can find it in Last Judgement or Mandragora on Koenigstuhl [19:14] ok :P i'll look into it [19:15] and [19:16] yaaayyyyy, got 8th on 94nop with a funny strategy :-P [19:17] gj! you pushed me off too :P [19:20] hrmm, sorry [19:23] nah, don't be [19:23] I stole that place by posting my warrior on a convenient moment [19:23] copypasting your qscan gained me 4 points [19:23] lets see if I can optimize it [19:30] nice :) [19:31] 110.08 against wilfiz [19:31] may I ask if how you optimize? [19:31] not bad ;-) [19:31] optimize = changing the separation between program and qscan :P [19:31] 107.38 against fsh94nop [19:33] do you use optimax? :-P [19:34] nop [19:34] I still have to look into that [19:37] ok ;-) [19:57] is there any redcode highlighter around maybe? [19:57] I found one for the 86 standard [20:03] oh, I found one for the 94 std too [20:21] Join: Fizmo_ joined #corewars [20:22] MSG: Ping timeout: 240 seconds [20:22] Nick Change: Fizmo_ changed nick to Fizmo [20:23] I'm installing perl :O [20:24] I dont understand completely how the optimax commands work, though [20:27] Join: Fizmo_ joined #corewars [20:29] MSG: Ping timeout: 240 seconds [20:29] Nick Change: Fizmo_ changed nick to Fizmo [20:30] well, only one got to phase 3 :P [20:31] MSG: Client Quit [20:47] Join: Fizmo joined #corewars [20:47] Leniad: at which level you set phase 3? [20:48] actually it seems to have more tha one on phase 3 [20:48] but the first one said 1 / 1 [20:48] the second one said 2 / 2 [20:48] etc [20:49] still, I don't understand how the commands work, I just copied them from example actually :P [20:49] you have to look in the work directory, there is a html file [20:49] ohh, I see [20:49] simply speaking it is a random value generator, auto benchmarking tool [20:50] yes, I get that [20:50] also I get how !(a-b) works [20:50] I don't get how the tresholds work though [20:50] choose the constants you want to have optimized and replace the value by, let's say by: !(100-7900) [20:51] what kind of warrior you're working on? [20:51] q^4.5 -> bomb -> sdclear [20:51] still, I understand how to choose the constants to be replaced [20:51] longer bomb run? [20:51] what I dont understand is what the different tresholds mean [20:53] there are four stages the generated warrior have to pass through [20:53] yes [20:53] phase 1 tests if it survives running alone [20:53] phase 1 is simply a test if the warrior make suicide during a battle [20:53] yes [20:53] phase 2 tests against one warrior [20:54] phase 3 against a small set of warrior [20:54] and then phase 4 against a bigger amount [20:54] yes, to check if the basic functionallity of the warrior is still working [20:54] yes exactly [20:54] what I don't undersant is how the static and dynamic tresholds work [20:55] depending of the type of warrior more than 90% of randomly generated constants are not optimal [20:55] by using phases you save processor time to let only run the best of them against the complete benchmark [20:56] yes [20:56] the dynamic tresholds means that only a certain percent of the maximum score passes through [20:56] I understood that ^^; [20:57] it's useful if you don't have a clue about the scores of your warrior [20:57] ;optimax phase2 80 [20:57] - set a static threshold for phase 2 [20:57] ;optimax phase2 90% [20:57] - set a dynamic threshold for phase 2 [20:57] it selfadjust the treshold in this case [20:57] how do those work? [20:57] if the best randry (warrior with randomly generated constants) scores 100 pts [20:58] every other randry which has at least 90 pts will be pass the phase [20:58] ah I see [20:58] unless a better score was formed [20:58] and the minium will be 80 no matter what [20:58] right? [20:59] it passes always randries which scores at least 90% of the best scoring one [20:59] but if the best scoring one scores 79 [20:59] and ;optimax phase2 80 will pass everything higher than 80 [20:59] ah [21:00] so if the best score is 100, one that scores 80 will pass anyways? [21:00] it is static [21:00] yes it will [21:00] hmm [21:00] i'd make it work the other way :P [21:00] I usually work with static one if I have in advance a clue how much it should score [21:00] if I have no clue I use dynamic one# [21:01] if you want to optimize a stone then use a scanner for phase 2 [21:01] and maybe scn:cds:clr for phase 3 [21:02] and let's say I wanted to try all the constants that are mod 4 for step sizes [21:02] I'd have to make a txt file with all the constants? [21:02] if you use a sd-clear against paper you could use for phase 3 cds:clr:pap [21:02] yes [21:03] annoying :P [21:04] or maybe you could do 4*!(0-1999) [21:04] if you want to do all possibilities you can do 4*!{0-1999} [21:04] this will count from 0 to 1999 [21:05] while !(0-1999) will do randomly [21:05] ah [21:05] good idea :P [21:06] 4*!(1-1999) [21:06] somehow I suspect a step size of 0 won't work very well :P [21:06] hehehe, ok, you are right :-P [21:06] then 4*!(1-1999) [21:08] if you want to test each step size in a stone/imp with 3- point and 7-point imps you could use ![2667|1143] then for every randomly set of constants both imps will be generated [21:08] so there are some different possibilities to generate constants [21:08] MSG: Quit: Lost terminal [21:13] any improvements in your warrior? [21:14] umm [21:14] I'm still fooling around with this [21:15] btw, my warrior scores 87 against /scn and 127 against /pap :P [21:17] 127 is not bad against pap [21:17] but looks like space for improvements against scn [21:17] but somehow it does better against paper that against scanners [21:17] it's supposed to be a stone :P [21:22] hehehe, as I said, it seems space for improvement [21:22] ok, need to go now, quite late here [21:22] * Fizmo waves [21:22] MSG: Quit: ChatZilla 0.9.78.1 [Firefox 2.0.0.3/2007030919] [21:49] Load error: no DPMI selectors==================================== | [21:49] Returncode : 27392 [21:49] what the heck does that mean? :O [21:50] and now whenever I try to run optimax again I get Load error: can't switch mode [21:50] Returncode : 27136