00:00:04 wow 00:00:10 AnMaster: welcome bac 00:00:11 k 00:00:12 radio went out with public message 00:00:14 over all channels 00:00:16 wut 00:00:19 what's happening 00:00:22 aliens? 00:00:28 AnMaster: aliens‽‽ 00:00:28 ehird, major gas leak in big city 00:00:37 yikes 00:00:41 AnMaster: of the deadly kind, I assume? 00:00:49 ehird, unknown 00:00:52 ehird, far from hear 00:00:54 here* 00:01:10 i'd interject with something like "that's what you get for being a dirty swede!" but it seems inappropriate 00:01:25 * ehird waits for the BBC to pick it up 00:01:41 ehird, it cut off in the middle of the news. Right in the middle of a sentence. 00:01:57 AnMaster: ... what caused that? The gas? 00:02:01 oh you mean 00:02:03 the broadcast 00:02:04 ehird, the message yeah 00:02:16 AnMaster: must be pretty bad tthen 00:02:26 ehird, "please go indoor and close all windows door and ventilation" 00:02:29 yeah 00:02:39 AnMaster: Yikes. 00:02:46 AnMaster: Severe, then. 00:02:57 ehird, from some industry there it seems. 00:03:06 * ehird F5 F5 F5 F5 on bbc.co.uk 00:03:09 (well okay, Cmd-R) 00:03:21 ehird, heh 00:04:24 ehird, there is nothing on the news anywhere about it yet... 00:04:28 not even on sr.se 00:04:29 yep 00:04:34 (radio's website) 00:04:35 AnMaster: must be very early report then 00:04:38 no time to write anything about it 00:04:39 ehird, yeah 00:04:46 which implies that the leak is very serious 00:06:13 AnMaster: you've got me all jumpy now :P 00:06:33 ehird, not that Swedish radio is very nocturnal. One channel shuts down after 01:30 until 06:00, one sends classical music streamed from BBC(!), the remaining two sends the same thing during the night, pop music. 00:07:06 mm 00:09:18 * ehird F5s sr.se too for good measure 00:10:27 AnMaster: also, wait 00:10:28 shut all windows? 00:10:30 it's 2am isn't it 00:10:34 ? 00:10:44 01:10 here 00:10:48 well, it seems like there wouldn't be many windows open at 2am :-P 00:10:49 well, 1am too 00:11:13 ehird, um, I slept with open window since mid-April 00:11:16 fresher air 00:11:23 AnMaster: in sweden? 00:11:24 ehird: even if you're up at 2am? 00:11:27 must freeze to death 00:11:35 ah 00:11:38 that's what you meant 00:11:43 oklodok: well, the windows in this house generally get closed when it gets cold at night 00:11:44 ehird, nah. I have a secret weapon 00:11:46 at least here it's been summer for weeks 00:11:58 oklodok: but it never really gets warm in .fi/.se does it :-P 00:12:08 ehird, anyway, no ventilation in this house, it is too old 00:12:16 built 1907 00:12:27 this is a victorian sort of house I think 00:12:30 well it's a flat 00:12:31 mhm 00:12:33 so half of a victorian house 00:12:35 well, it's sort of a flat 00:12:40 the bathroom is up a flight of stairs 00:12:43 it's rather silly 00:12:49 free standing 00:13:00 ehird: i don't know about your standards, but warm enough that i might keep a window open if i ever did 00:13:07 oklodok: what celsius? 00:13:24 AnMaster: waddya mean be freestanding 00:14:23 ehird, like http://sv.wikipedia.org/wiki/Fil:Villa-nyköping-ostra4.JPG 00:14:35 AnMaster: term is "detached house" 00:14:44 ehird, kay. in Swedish it is "villa" 00:14:48 I live in a flat in a terrace, so not nearly as free as that :-P 00:15:01 flat in a terrace? 00:15:13 AnMaster: a http://en.wikipedia.org/wiki/Terraced_house containing flats. 00:15:30 aka apartment 00:15:34 if you're oh so very american 00:15:36 :P 00:15:46 I'm not 00:15:54 ehird: i don't really know. i saw a 14 a few mornings ago, but it was very early. 00:16:00 may have been night 00:16:06 oklodok: 14C is what it's like here at night :-P 00:16:20 10C atm outside and it's midnight 00:16:22 ehird, I have a secret weapon 00:16:25 AnMaster: oh? 00:16:27 a BFG? 00:16:28 10C outside here too. 00:16:32 ehird, hot water bottle! 00:16:32 :P 00:16:39 http://outside.hut.fi/ is the temperature sensor at our university, but it gets direct sunlight, like you can see from the graphs. 00:16:41 fizzie: yeah but it's sweden. their celsi-ii is colder. 00:16:52 err 00:16:53 finland 00:16:57 ehird, actually, Finland is colder than Sweden 00:17:02 i said finland 00:17:03 it was a mistake 00:17:10 I meant that fizzie's 10C is colder than my 10C 00:17:11 :-P 00:17:17 uhu? 00:17:21 yes 00:17:24 it's a law of finnishology 00:17:28 fizzie: wow that graph is pretty 00:17:40 oklodok: it's like a sine wave after a beating 00:17:48 There's that five-years-or-so big graph at http://outside.hut.fi/five.html which might average out the worst sunlight-freak-o-temperatures. 00:18:03 -!- coppro has quit (Read error: 110 (Connection timed out)). 00:18:06 Since it's 24h average, or so they say. 00:18:32 The radiolab or some other people had a real weather-station readings, but I can never remember the URL for that. 00:19:02 ehird, also, slightly open, not wide open 00:19:14 http://www.yr.no says it is 8 C here 00:19:47 Oh, there it is. http://radio.tkk.fi/en/weather/ has some real data. 00:19:50 ehird, rained today, so colder than yesterday 00:20:40 still nothing on bbc or that sr.se thing 00:28:26 "Ammoniakutsläpp i Stenungsund" 00:28:30 http://www.svd.se/nyheter/inrikes/artikel_2917959.svd 00:28:31 there 00:28:35 damn 00:28:37 that is BAD 00:28:42 AnMaster: /me runs google translate 00:28:48 AnMaster: how bad 00:28:51 AnMaster: national crisis bad? 00:28:54 http://en.wikipedia.org/wiki/Ammonia 00:28:55 that bad 00:29:09 AnMaster: holy. fucking. shit. 00:29:22 says it was unclear how large it was 00:29:24 at midnight 00:29:52 Sorry, we are unable to translate the page you requested. 00:29:52 http://www.svd.se/nyheter/inrikes/artikel_2917959.svd 00:29:58 "Men enligt polisens länskommunikationscentral hade inte räddningstjänsten vid midnatt begärt någon hjälp av polisen angående avspärrningar eller eventuell evakuering." <-- emergency service had not requested any help with evacuation at the time of writing. Well I guess that changed since 00:30:07 the last is my comment. 00:30:24 AnMaster: could you translate the tagline and the two paragraphs? 00:30:30 kay 00:30:46 Ammoniakutsläpp i Stenungsund <-- Leak of ammonia in Stenungsund 00:30:56 Ett utsläpp av ammoniak har skett vid Axo Nobels anläggning i Stenungsund. Räddningstjänsten har i radion gått ut med ett så kallat Viktigt meddelande där allmänheten uppmanas att stanna inne, stänga fönster och ventilation. 00:32:08 A leak of ammonia has happened at Axo Nobel's plant/facility in Stenungsund. The Emergency service has in radio sent a so called Important message where people are asked to stay indoors, close windows and ventilation 00:32:15 that is a semi-broken grammar 00:32:19 in the translation 00:32:31 (couldn't be arsed to try to fix it) 00:32:36 Det var vid midnatt oklart hur stort utsläppet är. 00:32:36 Men enligt polisens länskommunikationscentral hade inte räddningstjänsten vid midnatt begärt någon hjälp av polisen angående avspärrningar eller eventuell evakuering. 00:32:50 AnMaster: "a so called Important" :-D 00:32:53 It is at midnight it unclear how large the leak is. 00:33:27 ehird, I think that is an official name "Important message" 00:33:32 eyah 00:33:33 yeah 00:33:34 it's just funny 00:33:35 it was with upper case in Swedish 00:33:39 since "so called X" means doubtfully-X 00:33:40 Men enligt polisens länskommunikationscentral hade inte räddningstjänsten vid midnatt begärt någon hjälp av polisen angående avspärrningar eller eventuell evakuering. 00:33:41 like 00:33:49 ehird, it isn't like that in Swedish 00:33:53 yeah 00:33:53 :P 00:33:56 anyway 00:33:59 Men enligt polisens länskommunikationscentral hade inte räddningstjänsten vid midnatt begärt någon hjälp av polisen angående avspärrningar eller eventuell evakuering. 00:35:07 But (according to the police's communication central) the emergency service had not (at midnight) requested any help with barricades or evacuation. 00:35:10 AnMaster: Men enligt polisens länskommunikationscentral hade inte räddningstjänsten vid midnatt begärt någon hjälp av polisen angående avspärrningar eller eventuell evakuering. 00:35:44 who cares about some silly catastrophy, let's just play the nick pasting game 00:35:55 oklodok, no 00:35:58 00:35 oklodok: AnMaster: Men enligt polisens länskommunikationscentral hade inte räddningstjänsten vid midnatt begärt någon hjälp av polisen angående avspärrningar eller eventuell evakuering. 00:36:01 I was trying to get back on topic 00:36:02 ... 00:36:04 :P 00:36:05 AnMaster: gotta agree with him here, this is #esoteric 00:36:07 going to bed now 00:36:11 night! 00:36:12 bye 00:37:21 5 pages to go, and i'll be done, and my next exam isn't in long times! 00:37:25 http://www.aftonbladet.se/senastenytt/ttnyheter/inrikes/article5177106.ab 00:37:27 same text 00:37:28 heh 00:37:33 ah yeah TT 00:37:36 AnMaster: it'll come from a news agency 00:37:38 it's like Reuters 00:37:39 yeah 00:37:40 like Reuters or AP 00:37:40 right 00:37:41 news agency 00:38:34 ehird are you still benchmarking bbc response time? 00:38:46 :DD 00:38:54 oklodok: can I hire you to be funny at me all day? 00:38:55 news: 00:38:59 google maps works without js 00:39:08 yes really 00:39:08 AnMaster: that can't be very pleasant 00:39:20 click. wait. click. wait. click. wait. gee, if only I could drag 00:39:46 ehird: i don't think i'd work well paid by hour 00:39:58 oklodok: flat monthly rate? 00:40:02 ehird, lim_(x -> 0) x = google response time 00:40:03 :P 00:40:18 ehird: well i'm all for a steady income :P 00:40:21 AnMaster: well sure but :P 00:40:29 http://maps.google.com/maps?q=Stenungsund&oe=UTF-8&ie=UTF-8&split=0&ei=fvERSvCXIcnDsgaFoOSJDg&output=html&zoom=3&zp=OOOOOO 00:40:31 that is where 00:40:43 AnMaster: you can't even click! 00:40:43 which interface is it 00:40:47 ehird, what? 00:40:55 you can't click on the map itself 00:40:57 to "drag" 00:40:59 ehird, I think it is for mobile phones 00:41:02 and such 00:41:03 ah 00:41:05 AnMaster: nah 00:41:10 too wide fram efor that 00:41:15 true 00:41:25 ehird, for links -g ? 00:41:30 for every length of time there is a piece of news google caught on before said time had elapsed 00:41:32 AnMaster: heh 00:42:10 but I think this is new 00:42:16 I remember trying without JS before 00:42:18 no worky 00:42:19 Joke site idea: silentcarreview.com 00:42:31 ehird, sounds like a good idea 00:42:32 even 00:42:37 (yeah seriously) 00:42:44 AnMaster: you'd have to silence all the other motorists' cars too :-P 00:42:52 ehird, oh? 00:42:54 ah 00:42:56 hm 00:43:01 trickier 00:43:06 it won't make much difference if you're the only silent car on the road will it 00:43:20 ehird, yeah but make everyone buy one 00:43:35 well if we're going down that route make it mandatory that everyone own a kitten 00:43:38 morale will improve instantly 00:45:09 ehird, allergic to them. Can't 00:45:19 plus cats are evil 00:45:32 AnMaster: if I was allergic to kittens I'd never get out of bed 00:45:38 i'd just lie there thinking about how worthless my life is 00:45:56 :P 00:46:08 ehird, allergic to dogs, cats and some sorts of pollen. 00:46:17 well dogs is okay i don't like dogs 00:46:21 and pollen isn't cute and fluffy 00:46:25 but kittens... 00:46:56 AnMaster: weren't you supposed to sleep?!?!?!?!?!?!?!? 00:48:58 oklodok, see above 00:49:02 think 00:50:35 ehird, updates 00:50:42 ooh 00:50:50 ehird, "situation under control" 00:50:52 AnMaster: hmm no, my thinking cannot explain it 00:50:54 "no wind" 00:51:07 AnMaster: boring, i was hoping for like, zombie outbreak 00:51:19 yeah those are always interesting for a while 00:51:24 ehird, wouldn't be nice. I know people in that area. 00:51:37 yes, but, zombies kind of outweigh that 00:51:46 ehird, that isn't the effect 00:51:53 the effect is just dead people 00:51:59 dead people turn into zombies via magic 00:52:03 Phngh. Finally got graded those AI course project-works. Meh, I've been making snide comments about our TCS lab professors who reply to emails in the middle of the night; but now I self sent those grades out at 02:30 am or so. 00:52:22 :D 00:52:58 ehird, they still don't know how much leaked. But the leak is "probably" stopped 00:53:14 they suspect that the tank that leaked wasn't full 00:53:23 AnMaster: how comforting! 00:53:45 http://www.svd.se/nyheter/inrikes/artikel_2917959.svd 00:53:59 brb learning to read swedish 00:54:09 ehird, copy paste to google translate 00:54:09 works 00:54:14 :P 00:54:31 -!- pikhq has quit (Read error: 54 (Connection reset by peer)). 00:54:34 ehird, was it better we had that huge fire just a bit south of here? Last year. 00:54:44 - Situation is under control. We do not yet know how much has been leaked. But the weather is favorable, there is no wind that can spread a possible gasmoln. It smells of nothing, which suggests that it is not so much leaked, "said Bengt Carlsson. 00:54:49 ehird, 8 fire brigades worked on it at max 00:54:49 it smells of nothing 00:54:52 what a way to detect 00:54:56 just sniff! 00:55:02 It is a tank when it is filled capacity of 3 000 cubic meters of ammonia for some reason leaked. 00:55:07 3,000 cubic meters? ouch 00:55:18 ehird, could have been bad yeah 00:55:21 if it had been full 00:55:27 ah 00:55:31 the translation obscured the meaning 00:55:55 ehird, "It is a tank which when filled has a capacity of 3 000 ..." 00:55:57 the real reason i was scared is that sweden is a province of finland 00:56:02 one of the 3 people living there could have got hurt 00:56:23 the real reason i was scared is that sweden is a province of finland <--- technically wrong 00:56:39 no no totally true 00:57:11 ehird, but was it 00:57:30 ehird, yes or no 00:57:39 finland used to be a province of sweden 00:57:42 ehird, was it better we had that huge fire just a bit south of here? Last year. ehird, 8 fire brigades worked on it at max 00:57:43 that 00:57:48 but then time-space warped and Finland-space started 00:57:50 AnMaster: "better" how 00:57:55 wind was blowing the smoke my way 00:58:01 + toxic smoke 00:58:02 ah 00:58:06 cool 00:58:08 (plastics) 00:58:13 ehird, I mentioned it in here then 00:58:14 ... 00:58:22 night really 00:58:23 i don't recall last year 00:58:24 :-P 00:58:25 bye 00:58:44 night! 01:00:00 He's probably lulling us to a false sense of security; he'll jump out of hiding and attack any minute now. 01:00:07 noooooooooooo 01:00:07 3000 cubic meters... could probably be used as a blue whale aquarium 01:00:26 (i want one!) 01:00:43 oklodok: yes store them in ammonia 01:01:29 3000 cubic meters is still just a cube with approximately 14.422 meter sides. That's not so big. 01:01:38 true 01:02:09 it's quite big for an aquarium 01:02:24 in fact i don't think i've seen one as big 01:03:54 The boat-building department here at TKK has a pool of 7865 cubic meters (5.5 meters deep, 11 meters wide, 130 meters long), but that's just plain old water. 01:04:17 :O 01:04:26 how many blue whales is that? 01:04:27 fizzie: omg awesome 01:04:35 -!- lifthrasiir has joined. 01:04:46 They also advertise the "world's largest 40 x 40 meter ice-model-test-pool where you can perform experiments with 70 mm thick model ice", but they don't say how deep that thing is. 01:05:16 http://www.tkk.fi/Units/Ship/General/ and then the "ship laboratory"; it's a silly javascript thing that'll toggle the content open. 01:05:48 The English version doesn't say "world's largest"; only the Finnish one has that. Heh. 01:06:07 they don't want to reveal their secrets 01:06:10 -!- inurinternet has quit (Read error: 113 (No route to host)). 01:06:26 Hopefully they build better ships than web pages. 01:07:40 There's wave-making machinery in both pools (except it's called a "basin", that sounds more officious); I'm *very* certain the students use them as huge-ass swimming pools (don't move the dash again) when no-one's looking. 01:07:59 fizzie: huge ass-swimming pools 01:08:09 Maybe that, too. 01:08:18 huge ass swimming po-ols 01:08:21 it's moving!! 01:10:27 -!- inurinternet has joined. 01:16:02 oooooooooooooooooooooooooooooooooooooooooooooooooooo 01:16:13 ooooooooooooooooooooooooooooooooooooo 01:16:40 free as a bird am i now 01:16:47 i don't have a care in the world 01:17:21 and i am green 01:18:46 why so 01:19:23 oklodok: well because glo-fold 01:23:15 ! 01:23:28 oklodok: y eah! 01:25:47 Metaplace is open beta! 01:26:23 Pavitra's on MP 01:26:29 ? 01:29:17 w00t, it's possilble to go into the MP Central waters! 01:29:29 -!- Gracenotes has quit (Read error: 113 (No route to host)). 01:35:56 ehird: grep is twice as fast with hyperthreading 01:36:08 bsmntbombdood: grep on what 01:36:09 /dev/random? 01:36:13 yeah 01:36:25 bsmntbombdood: that's such a real-world operation. 01:36:28 not 01:38:13 i remember reading an article detailing how someone used DFAs to grep at like 12 gb/s 01:38:30 and i was pretty impressed 01:38:35 but this processor can do 9 01:38:48 bsmntbombdood: grep is singlethreaded 01:38:51 how can it be faster w/ ht 01:39:05 it is trivial to parelellize manually 01:39:36 bsmntbombdood: you said "grep" 01:39:40 not "my parallel grep" 01:39:45 which skews the results 01:39:52 ... 01:40:02 no it doesn't 01:40:14 meh :p 01:41:18 bsmntbombdood: try fgrep 01:41:21 I bet you could hit 12gb/s 01:41:39 naw, grep uses the same algorithm on fixed strings 01:41:50 bsmntbombdood: but with less processing itme 01:41:54 due to not having to interpret special chars 01:41:58 boyer-moore judging from the fact that you get better speed on longer strings 01:42:02 fine, one sec 01:43:27 -!- pikhq has joined. 01:43:44 linux's /dev/urandom is really slow 01:43:51 bsmntbombdood: cat it to a file 01:43:52 then fgrep 01:43:55 no shit 01:43:56 (ssd, ofc) 01:44:02 no, ramdisk ofc 01:44:07 lol 01:44:12 true 01:44:21 bsmntbombdood: you have more ram than my first HD size 01:45:28 there's some bad locking too 01:45:35 bsmntbombdood: What, you expect it to create entropy? 01:45:42 pikhq: it's urandom 01:45:44 it's meant to make shit up 01:45:46 i am running 8 dd's of /dev/urandom and each cpu is only at 35% 01:45:48 random is the one that freezes 01:45:51 Ah, right. 01:45:52 pikhq: yes, i do 01:46:12 bsmntbombdood: It's not going to make entropy. 01:46:18 ... 01:46:19 It's going to pretend to have entropy. 01:46:20 ;) 01:46:21 yes 01:46:35 it output is indistinguishable from real entropy 01:46:44 err no 01:47:09 with a computationally bounded attacker, yes 01:47:14 -!- Corun has quit ("Leaving..."). 01:47:56 Well if you specify "computationally bounded"... So are primes, for some notion of computationaly bounded. :p 01:48:11 "computationally bounded" = real 01:48:35 Depends on the psuedorandom number generator. 01:48:52 obviously 01:49:01 /dev/urandom is purported to be cryptographically secure 01:49:26 Granted. 01:50:19 $ time echo ran* ran* ran* ran* ran* ran*| xargs -n 1 -P 8 grep foobarbaz 01:50:19 real 0m2.355s 01:50:19 user 0m9.897s 01:50:19 sys 0m6.420s 01:50:53 10 gigabytes/second 01:51:29 bsmntbombdood: your rig is amazing. challenged only by my future one ;-) 01:51:34 fgrep takes exactly the same time 01:52:02 ehird: God, if I had $10,000 in disposable income. 01:52:12 pikhq: not that much! bsmnt's cost $1,7000 01:52:12 and check this out- 01:52:13 w 01:52:14 er 01:52:16 $ time echo ran* ran* ran* ran* ran* ran*| xargs -n 1 -P 8 fgrep foobarbadsfkj09i3209ilksajdfkljaslfjlasfjdlkajsfdlkjalsfjdksajdfkljaslllfkdkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkkazffffffffffassssssssssssssssssssssssssssssssssssssssssssssss 01:52:16 real 0m1.611s 01:52:17 $1,700 01:52:27 ehird: Our uberbuild. 01:52:30 mine will probably cost about $4,000 01:52:32 pikhq: oyeah :-D 01:52:33 bye → 01:52:59 the 6 seconds of system time worries me 01:53:06 i'm thinking i can do better 01:58:50 -!- coppro has joined. 01:59:12 bsmntbombdood: Uh, you do realise that file access consists of system time, right? 01:59:28 yes 02:06:41 ahahaha 02:06:57 http://www.google.com/url?sa=t&source=web&ct=res&cd=3&url=http%3A%2F%2Fwww.arl.wustl.edu%2F~sarang%2Fjsac_cameraready.pdf&ei=mAUSSpCGG8OHtgfOoeWACA&usg=AFQjCNGdR0vyT1iYMrr6xRtEeUuwGhjSAg&sig2=tdY3YBP1h0x85tTt4DtNPw 02:07:04 that's the paper i was talking about 02:07:22 a measly core i7 is way faster than their specialized hardware 02:23:00 wc(1) is like a zillion times slower than grep 03:04:10 -!- Gracenotes has joined. 03:04:34 bsmntbombdood: wc(1) counts shit. Grep does regexps. 03:04:34 Whaddya expect? 03:12:19 counting is faster than regexps 03:13:03 actually wc -w takes about the same time as grep 03:13:07 wc -w is what is slow 03:17:00 Mmm. 03:17:36 wc -c is of course constant time 03:18:16 Well, yeah. 03:52:20 I love Allegiance 03:52:21 http://www.youtube.com/watch?v=EZ__4qWLO7E 04:05:49 i need a 64 bit rolling checksum 04:26:51 -!- bsmntbombdood_ has joined. 04:27:37 -!- bsmntbombdood has quit (Read error: 113 (No route to host)). 04:53:29 -!- pikhq has quit (Read error: 104 (Connection reset by peer)). 05:09:47 -!- oklodok has quit (Read error: 60 (Operation timed out)). 05:44:31 -!- pikhq has joined. 06:17:00 -!- asiekierka has joined. 06:17:01 hi 06:19:36 hi 06:21:35 Heh, I had a scoring idea for BF Joust 06:21:45 you battle on each possible tape size 06:21:57 draws are 0, wins are +1 and loses are -1 06:22:13 So for example if a program wins with the bot on 20 tape sizes but loses on one, it has 19 points 06:22:22 Now, it battles with every program in the list other than itself 06:22:43 so if there are 10 programs and the program we're talking about is not one of these 06:22:47 it can score up to 200 points 06:22:55 or fail at -200 points 06:30:47 -!- upyr[emacs] has quit (Read error: 113 (No route to host)). 06:33:41 -!- asiekierka has quit. 07:15:37 -!- kar8nga has joined. 07:38:53 -!- olsner has joined. 07:50:16 -!- kar8nga has quit (Read error: 104 (Connection reset by peer)). 07:59:59 -!- clog has quit (ended). 08:00:00 -!- clog has joined. 08:43:06 -!- GreaseMonkey has joined. 09:01:13 -!- Sgeo has quit ("Leaving"). 09:02:06 -!- lereah_ has joined. 09:08:41 -!- Judofyr has joined. 09:42:39 -!- sebbu has quit (Read error: 104 (Connection reset by peer)). 09:43:12 -!- sebbu has joined. 10:16:53 -!- GreaseMonkey has quit ("YES -> thor-ainor.it <- THIS IS *DELICIOUS*!"). 10:33:23 -!- oerjan has joined. 11:27:54 -!- tombom has joined. 11:39:37 AnMaster: omg Darths and Droids 11:39:53 *& 11:42:34 oerjan, agreed 11:42:47 It *so* has not been ten years. Soon you'll be telling me it's been over ten years since the Matrix. 11:43:03 no idea if it was. 11:43:38 fizzie, when was Matrix released 11:43:42 fizzie: well it's been ten years according to our clocks. the simulation of course runs much faster. 11:43:48 probably a couple hours 11:43:51 hehe 11:44:00 March 1999, apparently. 11:44:23 then it is 10 years + some months I guess. 11:44:48 what about Terminator? 11:45:12 25 11:45:21 wow really? 11:45:23 heh 11:45:32 1984; I had just managed to be born the year before. 11:46:05 I was born -5 years before. 11:46:31 18 for #2 11:46:59 So even Terminator 2 can now drink alcoholic beverages. (Well, depending on the local age limits and so on; it's 18 around here.) 11:47:10 I never seen Terminator, only Terminator 2 11:47:18 s/seen/watched/ 11:47:35 -!- sebbu has quit (Read error: 104 (Connection reset by peer)). 11:48:04 -!- sebbu has joined. 11:48:43 -!- coppro has quit (Read error: 104 (Connection reset by peer)). 11:49:01 -!- pikhq has quit (Read error: 104 (Connection reset by peer)). 11:49:04 -!- Halph has joined. 11:49:10 -!- Halph has changed nick to coppro. 11:56:51 -!- KingOfKarlsruhe has joined. 12:15:47 -!- tombom has quit ("Peace and Protection 4.22.2"). 12:24:44 -!- jix has joined. 12:26:04 -!- MizardX has quit ("What are you sinking about?"). 12:42:48 -!- MizardX has joined. 13:17:30 -!- oerjan has quit ("leaving"). 13:20:21 -!- oerjan has joined. 13:20:27 !show slashes 13:20:27 perl (sending via DCC) 13:22:23 * oerjan realizes debugging in the inner loop may not exactly help performance 13:23:00 !delinterp slashes 13:23:01 Interpreter slashes deleted. 13:23:19 !addinterp slashes http://oerjan.nvg.org/esoteric/slashes/slashes.pl 13:23:19 Interpreter http___oerjan_nvg_org_esoteric_slashes_slashes_pl does not exist! 13:23:25 !addinterp slashes perl http://oerjan.nvg.org/esoteric/slashes/slashes.pl 13:23:28 Interpreter slashes installed. 13:23:40 Ooh, now that !show + DCC thing is *fancy*. 13:23:59 fungot: Silly bot, you don't even know how to do *anything* with DCC. 13:24:00 fizzie: the chip select which joystick port. this port, there are characters waiting in the table that starts at the sprite in the individual terms by performing the indicated memory location. when you turn it back on. 13:24:06 !slashes /#/1,2,3/test # 13:24:06 test 1,2,3 13:29:04 !slashes http://oerjan.nvg.org/esoteric/slashes/bct.sss 13:30:44 ok that didn't really help any 13:45:41 -!- jix has quit (Read error: 60 (Operation timed out)). 14:20:09 -!- jix has joined. 14:26:55 -!- oerjan has quit ("leaving"). 14:27:33 -!- KingOfKarlsruhe has quit (Remote closed the connection). 15:00:18 -!- pikhq has joined. 15:34:30 -!- FireFly has joined. 15:39:01 -!- upyr[emacs] has joined. 15:39:01 -!- BeholdMyGlory has joined. 15:42:36 .ssssssssssssssss 15:42:40 source code for snakes 15:50:27 Like python? :o 15:52:23 no. 15:52:33 thats source code made from snakes. 15:52:35 totally different thing. 16:06:44 -!- lereah_ has quit ("Leaving"). 16:10:23 heh 16:10:39 is this BF loop balanced or not: [>[-]+++[>++>++<<],++.<+] 16:10:47 counting just <> it is. 16:10:48 but 16:10:58 it is also equal to [>[-]+++[>++>++<<]] 16:11:03 which is unbalanced 16:11:32 (this is because the inner loop is trivially provable to be infinite and will always run (also trivial to prove) 16:11:35 ) 16:13:19 -!- Slereah has quit (Read error: 60 (Operation timed out)). 16:13:28 -!- Slereah_ has joined. 16:13:47 wot 16:13:51 psygnisfive, ? 16:13:59 what do you mean balances 16:14:03 balanced 16:14:20 psygnisfive, "current cell is same before and after loop" 16:14:48 oh ok. 16:14:49 if it is you can do all sorts of optimisations in BF that you can't otherwise 16:16:04 -!- ais523 has joined. 16:16:25 hi ais523 16:16:27 ais523, is [>[-]+++[>++>++<<],++.<+] balanced 16:16:38 AnMaster: yes 16:16:40 ais523, but it is equal to [>[-]+++[>++>++<<]]! 16:16:42 which isn't 16:16:49 I meant, balanced square brackets 16:16:56 ais523, balanced moves 16:17:01 which is the usual meaning 16:17:06 yes, balanced moves 16:17:17 and given that the outer loop never finishes 16:17:23 it doesn't matter whether that loop's balanced or not 16:17:27 hm indeed. 16:17:40 p++; while(1); p--; leaves p the same 16:17:43 p++; while(1); doesn't 16:17:45 meh have to rewrite part of the calculator 16:17:47 yet they are the same 16:17:58 for loop balance 16:18:28 hm in fact 16:18:37 this in theory opens up lots of ways to optimise 16:18:45 isnt there some sort of compiler you could write for this? :P 16:19:05 since you know that after that outer loop anything done in it will be void. 16:19:09 yes 16:19:10 psygnisfive, err what 16:19:18 psygnisfive, I'm writing a BF compiler... 16:19:31 oh, is that what youre doing /right now/? 16:19:50 psygnisfive, well right now I'm writing on irc 16:19:54 :P 16:19:57 and writing code at the same time 16:20:09 (single core multi tasking style) 16:20:14 ok 16:20:20 im going to go shower, then run off. bye. :P 16:20:25 cya 16:27:28 -!- coppro has quit (Read error: 110 (Connection timed out)). 16:37:39 -!- inurinternet has quit (Connection timed out). 17:00:23 00:52 bsmntbombdood: the 6 seconds of system time worries me 17:00:23 00:53 bsmntbombdood: i'm thinking i can do better 17:00:24 coppro has joined (n=coppro@unaffiliated/coppro) 17:00:26 00:59 pikhq: bsmntbombdood: Uh, you do realise that file access consists of system time, right? 17:00:28 00:59 bsmntbombdood: yes 17:00:30 align the ssd bitch 17:00:59 ehird: He was feeding from /dev/urandom. 17:01:09 irrelevant 17:01:13 it'd still help 17:01:15 it always helps 17:01:16 :P 17:01:19 No it won't. 17:01:21 <_< 17:01:27 pikhq: whaddya think about his grep being faster w/ HT? 17:01:28 it's odd 17:01:48 Very odd, considering that grep is single-threaded. 17:02:15 a multithreaded grep isn't theoretically impossible 17:02:16 pikhq: no, he ran multiple greps 17:02:18 iirc 17:02:25 and they went twice as fast 17:02:29 ais523: Yes, but it *is* single-threaded right now. 17:02:31 GregorR: GregorR-L: there? 17:02:36 Somebody mailed me a copy of SICP. Now what would they mean by that...? —Guido van Rossum 17:02:40 HAVE YOU READ YOUR SICP TODAY? 17:02:42 ehird: I'd imagine that grep doesn't touch the cache much. 17:03:03 A better test would be, say, compilation. 17:03:16 i want to give whoever sent gvr sicp $1m 17:03:34 ehird: you don't have $1m 17:03:36 also, who is gvr? 17:03:41 ais523: creator of python 17:03:41 1 megadollar? 17:03:45 ais523: Guido van Rossum. 17:03:48 aha 17:03:55 it's ehird's answer that's more useful to me 17:03:56 and Benevolent Functional-Programming-Ignorant Dictator for Life 17:04:08 obviously some /prog/ger sent him SICP 17:04:09 ehird: He's better than some. 17:04:12 well, python has lambda... 17:04:32 that's enough to implement functional programming from first principles, if you really want to 17:04:36 "hur hur tail calls = tail recursion and can always be replaced trivially with a while loop what do you mean tailcalling another function" 17:04:37 tbh, any TC lang is 17:04:43 "I am dropping reduce() from the base language because IT HURTS MY BRAIN" 17:04:50 "I HAVE TO TURN IT INTO A FOR LOOP this is obviously reduce's f ault" 17:04:57 ... *Wha*? 17:05:00 "also don't use map() it gives you rabies" 17:05:06 pikhq: reduce = fold, he removed it from py3k 17:05:12 because he can only think imperatively, apparently 17:05:15 Even *Tcl* does tail call recursion these days. 17:05:23 is the name reduce or fold more common? 17:05:25 does it do general tail call optimization? 17:05:30 that's more useful than tail recursion 17:05:31 and Perl has an /operator/ for tail call optimisation 17:05:41 ais523: reduce is what the scripting langs use, fold is what the academic languages use 17:05:44 so you have to do it by hand, but it supports TCO too 17:05:49 (granted, you do it via "tailcall" instead of calling your function, but that's because they're just a tiny bit lazy) 17:05:50 scheme has fold, fold-right in SRFI-1 17:05:53 haskell is even more specific 17:05:57 foldl, foldr, foldl1 17:05:59 foldl' 17:06:00 etc 17:06:03 ais523: They've got a proc to optimise it in Tcl 8.6. 17:06:24 In current Tcl, you just have to rely on the stack not growing too large. 17:06:24 Or implement it yourself. 17:06:24 * ais523 vaguely wonders if gcc-bf does tail-call optimisation 17:06:31 it has a stack, after all 17:06:44 my guess is it's probably capable of using gcc's built-in TCO 17:06:47 but I've never tried 17:06:48 (Tcl is rather ridiculously modifiable, after all) 17:06:56 anyway, what I'm saying is, some hero from /prog/ gave gvr sicp as a joke. 17:07:05 ~fin~ 17:07:07 ehird: Which is totally awesome. 17:07:10 Yes. 17:07:29 I only hope they included the /prog/snake. Live. 17:07:34 *Alive 17:08:04 anyway: http://www.groklaw.net/article.php?story=20090518204959409 17:08:07 Hmm. Funny thing is, the only Python programmer I know well is accused of programming like Python's a Lisp... 17:08:19 I mention Alpha's terms of use on Groklaw, and then this happens 17:08:25 and I have no idea whether or not it's coincidence 17:09:05 haha, in the /r/programming/ comments on the link to the tweet people are trying to understand it :-) 17:09:22 "I think it can be a reference to MIT's changes in curriculum rather than the discussion on tail recursion." 17:10:33 well, it could be 17:10:47 it's totally a /prog/ger trolling him. 17:11:00 SICP's a nice high-brow way to troll someone 17:11:07 although I've never actually read it 17:11:16 should I, I wonder? 17:11:22 not now because I have unrelated exams 17:11:32 ais523: if you can avoid testing the Scheme code on 5 different broken implementations to check it's portable, then yes 17:11:46 I'd recommend MIT Scheme to go through with it, since it's a very traditionalist R5RS sort of implementation. 17:11:55 what's the exact subject of the book? 17:12:02 ais523: Achieving satori. 17:12:12 I already know quite a bit of functional programming, although I'm self-taught in it 17:12:12 But, uh, it's about the structure and interpretation of computer programs. 17:12:31 It's about the fundamental interactions between interpreters, compilers, languages, metacircular things, homoiconicism, Scheme, ... 17:13:19 ais523: But I'm not sure SICP can be called high-brow in the context of /prog/ 17:13:56 ais523: Wikipedia sums it up well: 17:14:04 [[Using a dialect of the Lisp programming language known as Scheme, the book explains core computer science concepts, including abstraction, recursion, interpreters and metalinguistic abstraction, and teaches modular programming. 17:14:05 17:14:07 The program also introduces a practical implementation of the register machine concept, defining and developing an assembler for such a construct, which is used as a virtual machine for the implementation of interpreters and compilers in the book, and as a testbed for illustrating the implementation and effect of modifications to the evaluation mechanism. Working Scheme systems based on the design described in this book are quite common student projects. 17:14:12 ]] 17:14:21 It's a very academic sort of book. 17:14:50 yes 17:15:03 hmm... I wonder if it's like UNIX 17:15:15 in that it is possible to make something better than UNIX, but you'll almost certainly fail unless you understand UNIX first 17:15:28 I think SICP is pretty much perfect. 17:15:37 mit's website is really pretty: http://mit.edu/ 17:15:40 I mean, in its descriptions of how to make an interpreter 17:15:53 ais523: it doesn't really describe that 17:15:59 ugh, that picture in the middle is really glaring on the eyes 17:16:01 it describes the theory of interpreters 17:16:01 it's just executable theory 17:16:06 what? 17:16:10 your brightness is set too high 17:16:16 quite possibly 17:16:22 it handles black on white just fine, though 17:16:26 and I have terminals set to grey on black 17:16:33 i just like the logo and lowercase title, pretty much :P 17:16:34 so seeing random white-on-black things overglares 17:16:48 ais523: i said yesterday that I want an lcd whose off state is white, not black 17:16:56 so that white would be soft on the eyes, not black 17:17:23 there are non-LCD screen architectures which manage that, but unfortunately they're impractical atm due to pixels taking almost a second to change colour 17:17:31 when that's fixed, though, you may have your wish 17:17:46 ais523: but I like LCDs! 17:17:55 what about them do you like? 17:17:56 also, I wonder if OLED can do it? 17:17:58 in particular/ 17:17:58 and I dunno 17:18:02 they're very sharp 17:18:07 pixels are very defined, not blurry 17:18:21 unlike, say, plasma displays, from what I know 17:18:47 also, I like subpixel rendering 17:18:59 I saw a picture where a bunch of text moved 1/3 of a pixel left each line 17:19:01 er, right 17:19:11 by using subpixel rendering 17:19:43 ais523: so what technologies are you talking about? 17:19:56 those flexible paper-based ones 17:20:05 ais523: those aren't bright enough 17:20:10 and don't have good enough colour 17:20:15 yet 17:20:21 isn't it inherent 17:20:35 no, the colour's physically inside the piece of paper 17:20:39 and gets moved around 17:20:49 wait, which displays are we talking about here? 17:20:52 Amazon Kindle-style? 17:20:56 yes 17:21:06 they use real paper? 17:21:07 are you sure? 17:21:08 they're about good enough to use for something like the Kindle atm 17:21:19 and I'm not quite sure if it's chemically the same as paper 17:21:24 but it has similar physical properties 17:21:26 maybe a bit stronger 17:21:41 Hmmmmmmmmmmmmwell. 17:21:48 I dunno. 17:21:51 What about OLEDs? 17:21:53 They use "organic" light. 17:22:05 Light that's fairly traded and not using chemicals, I suppose :-P 17:22:22 ehird: they use organic chemicals, rather than inorganic chemicals 17:22:37 I was joking 17:22:53 where organic = based around carbon chains 17:23:22 The Optimus Maximus keyboard developed by the Art. Lebedev Studio and released early 2008 uses 113 48×48-pixel OLEDs (10.1×10.1 mm) for its keys. 17:23:24 didn't know that 17:23:32 too bad it's a shit keyboard 17:23:38 yes 17:23:44 it seems engineered to be as expensive as possible 17:23:46 rather than actually useful 17:23:48 who looks at their keys, anyway? 17:23:50 I sure don't 17:24:04 I do on occasion to hit the home row 17:24:13 it's faster than feeling for it if I took my hands off the keyboard for any reason 17:24:18 my style is too weird for things like home rows :-) 17:24:19 and I am now, because you got me thinking about it... 17:24:25 ehird: so's mine 17:24:27 ais523: on a similar note, you're breathing 17:24:29 also, blinking 17:24:31 but my fingers need to start somewhere 17:24:33 have fun 17:24:42 ehird: blinking doesn't work on me, although the breathing thing does 17:25:06 ais523: also, your heart is beating. 17:25:17 ehird: I'm incapable of making my heart beat manually 17:25:25 well that's your problem! 17:25:25 so it'll just continue on auto, as always 17:27:57 ehird: do you have your heart on manual? 17:28:00 -!- inurinternet has joined. 17:28:09 ais523: no, I'm fairly sure that would be an emergency 17:28:16 can _anyone_ do that? 17:28:20 so why is it a problem if I don't? 17:28:34 because I told you it's beating so it's stopped now! 17:28:36 and although it's not quite on manual, some people have learnt to be able to control the speed with mental effort 17:28:38 if it's still beating, well, quantum immortality. 17:29:02 including stopping it for a few seconds 17:29:12 that's one skill I'm happy not to have 17:29:20 I'm not sure how useful that is, but presumably there's some reason they learnt it 17:29:24 that's more important than party tricks 17:32:59 ehird: everyone has some voluntary control over their blood flow 17:33:12 I don't that I'm aware of 17:33:16 because the heart doesn't push hard enough in order to get the blood all the way back up your legs 17:33:33 therefore, you have to keep moving your legs or feet (although only a few centimeters every few minutes is enough) 17:33:37 to stop all the blood getting stuck there 17:33:46 people occasionally faint from standing too still as a result 17:34:02 heh 17:34:08 ais523: that's not what I'd call voluntary 17:34:10 that's a side-effect 17:34:25 it's sort of like saying putting your head in a paper bag is a way to voluntarily stop breathing 17:34:46 heh 17:35:32 Somebody mailed me a copy of SICP. Now what would they mean by that...? —Guido van Rossum 17:35:32 HAVE YOU READ YOUR SICP TODAY? 17:35:33 awesome 17:37:48 ais523: reduce is what the scripting langs use, fold is what the academic languages use <-- which category do you put erlang in. 17:37:55 AnMaster: I don't know erlang 17:37:59 AnMaster: does it use reduce or fold? 17:38:07 also, I suspect it counts as academic, given its history 17:38:12 although company-academic, not university-academic 17:38:38 ais523, fold 17:38:59 ais523: yes, it's "company research" 17:39:08 rather than "university department of language research" 17:40:34 I think we're all agreeing, that's nice to see 17:40:45 if a little unusual 17:40:50 heh I found a bug in this wireless telephone. 17:40:57 -!- Judofyr has quit (Read error: 104 (Connection reset by peer)). 17:41:03 AnMaster: wut 17:41:14 After you held down the "end call" button for 10 seconds it goes into a "mode" where every button you press cause the screen backlighting to flicker and an odd short noise to be emitted. Loudly. 17:41:14 ais523: technically I think that means we died and went to heaven 17:41:24 Like a mix between beep and static noise 17:41:35 AnMaster: oh, that's probably a feature to stop you redialing after an argument, or something. I bet a marketroid came up with it. :-P 17:41:38 only way to make it exit the mode is take out the batteries and reinsert them 17:41:52 ehird, "end call" not "begin call" 17:42:08 ehird: it does sound like the definition of some features I've heard, but they wouldn't do it like that 17:42:19 I can't see it being a "feature"... 17:42:22 17:40 Ami: Antibiotic culture charged with failure to comply with the anticivic reportioning and reclamation decrees. 17:42:23 17:41 Ami: Jevanic polysyndicates post-charged for post-modern pseudosymbolism and monoreference. 17:42:24 17:41 Ami: Citizens, reminder. Inaction is conspiracy. Report counterbehaviour to your nearest stabilization delegate or workforce intake hub for full ration reward. 17:42:26 possibly a test mode. 17:42:27 ↑ this is basically what Sine is like all the time. or rather not, but it feels like it sometimes 17:42:28 That is possible 17:42:30 AnMaster: I was joking 17:42:32 many cars, for instance, have a feature that if you hold down the lock button on the keys for long enough it sets the alarm off as loud as it can and there's no way to stop it for several minutes 17:42:37 so you can use your car as a sort of rape alarm 17:42:42 which is a bit ridiculous 17:42:44 car rape alarm 17:42:50 three words I never thought I'd see consecutively 17:43:18 ehird: I'm trying to figure out what Ami is saying there, but failing 17:43:33 ais523: Ami is a bot, although I think those lines are either from a preset list, or entered manually by an operator 17:43:41 instead of, say, a markov chain 17:43:42 although the third line reminds me a bit of Paranoia XP 17:43:46 that's the general 17:43:47 theme 17:43:55 16:59 Cyclic bulletin file is missing. 17:43:55 16:59 Civic presence at 11 opt-ins and 16 opt-outs on 3 zoning enforcercement spars. 17:43:56 16:59 1 stabilization delegate(s) in position. 17:43:58 16:59 19 biotic cultures 17:44:01 16:59 Controlling therapy for 5 civic opt-ins with 1 zoning enforcement spars. 17:44:02 16:59 Local civic biomass: 5 Max: 9 17:44:04 16:59 Civic mass: 27 Max: 30 17:44:06 from the connection process 17:44:13 it could be a computer game, I suppose 17:44:16 it isn't, but it could be 17:44:18 although most of the time Ami just hands out awards for having been the last person to said something for ages 17:44:21 I found a test mode on a synth recently. It displayed TEST on the LCD. Then it played all tones, did a stepless "gliding" of a pure tone from ~40 Hz to well out of hearing range. After that it tested all LED segments on the display one by one. Then it shut down. 17:44:35 how to trigger it: Hold the setting button down while you powered it on 17:44:48 i've always wanted a speaker that I could play tunes out of the range of hearing on 17:45:01 at least they couldn't be discordant 17:45:15 ehird, hm? I tested range with a microphone 17:45:26 and it registered tones long after I didn't hear anything 17:48:02 * TaskManager will be replaced by TaskFreak!, dotProject and Trac 17:48:02 (tickets). 17:48:03 * DocManager will be replaced by MediaWiki and Trac (wiki). 17:48:05 * Discussion Forums will be replaced by phpBB. 17:48:07 * Diary and Notes will be replaced by WordPress. 17:48:11 AnMaster: well, SF are removing the awful forums, but replacing them with ... phpBB. 17:48:17 oh my 17:48:22 they moved into 2000! 17:48:23 I don't know which is worst 17:48:31 phpBB is excellent feature-wise, but truly awful code-wise 17:48:31 so 17:48:42 they will remove the bug tracker for any projects using it? 17:48:45 ais523: the features are mostly useless bloat 17:48:46 or is taskmanager something else 17:48:53 AnMaster: "tickets" 17:48:55 so bug tracker will disappear 17:49:04 and tbh, just replacing the whole thing with trac would probably be an improvement 17:49:08 ehird, and any projects using it? What will happen then 17:49:09 even though it doesn't have a forum 17:49:14 Dunno. Migration? 17:49:19 ehird, mhm hope so 17:49:22 -!- tombom has joined. 17:49:26 ais523: moving to trac being an improvement is a sign of a truly awful situation :) 17:49:47 trac isn't bad, it's just bland 17:49:55 it's good enough, but not much better 17:49:55 trac eats ram. 17:50:01 200 MB for the fcgi process 17:50:03 AnMaster: I didn't know that 17:50:03 I dunno, I'm not a fan of its UI and general concept-y stuff. 17:50:07 It's very mediocre. 17:50:12 ehird: mediocre is a good word 17:50:19 but then, I don't consider mediocre to be bad 17:50:22 ais523: it's perfectly cromulent 17:50:44 ais523: and if you think that's not actually a real real word, wiktionary lists it under "English": http://en.wiktionary.org/wiki/cromulent 17:51:00 ehird, cromulent? 17:51:01 cromulent is a real word 17:51:04 AnMaster: Cromulent. 17:51:17 aspell doesn't think it exists. Must be pretty rare 17:51:26 AnMaster: it's a neologism, but one that caught on 17:51:26 Wiktionary confirms it. 17:51:30 It's like Netcraft 17:51:30 it means roughly "makes sense" 17:51:38 Principal Skinner: He's embiggened that role with his cromulent performance. 17:51:40 ehird: that Wiktionary article has a big "do not trust this" box at the top 17:51:41 ais523: nope, it means "fine" 17:51:50 so 17:51:51 also, I'm selectively blind. 17:51:54 ehird: well, as a neologism, I wouldn't expect it to have a consistent meaning 17:52:05 ais523: in the same episode that coined it? 17:52:10 "it's a perfectly fine word" works too 17:52:18 well, as in "legitimate" 17:52:30 perfectly legitimate performance? 17:52:34 that doesn't really imply "good" 17:52:39 I suppose you could translate "cromulent" to "there's nothing wrong with it" 17:52:46 the implication to me is not "good", but "acceptable" 17:52:56 I think it's like a weak "good" 17:53:03 if you embiggen a role, obviously you did quite well 17:53:18 btw, reality check: we're arguing about the definition of a word coined on The Simpsons 17:56:05 Deewiant: link to that list of video cards ordered by awesome? 17:56:10 finnish thing 17:56:19 jathardware.com 17:56:33 thanksinyou 17:57:04 Deewiant: it lists the GeForce 9800GTX+ as below the 4770, but it's actually competitive-and-sometimes-better with the 4850 17:57:05 odd 17:57:06 The links that say "nopeusjärjestys" ("awesome order") are the ones you want 17:57:27 ehird: Well, they're in the same category 17:57:32 true 17:57:33 So they're clearly competitive 17:57:43 but the 4770 is a lot worse than both the 4850 and the 9800GTX+ from what I've seen 17:57:52 The ???? means he doesn't know whether the 4770 is better or worse than the 4850 17:57:58 ah 17:58:05 Deewiant: wait, does it really translate as awesome order? 17:58:08 I was joking 17:58:08 And the space between 4770 and 9800 means he thinks the 9800 is a bit worse 17:58:13 ehird: No, it doesn't :-P 17:58:24 17:58 Deewiant: And the space between 4770 and 9800 means he thinks the 9800 is a bit worse ← ati fan? :p 17:58:45 ehird: ------ is the only "true" separator 17:58:49 Sure. 17:58:52 Everything else is kinda variable 17:58:56 Deewiant: no, wrong number of -s 17:58:58 it should be ---- 17:59:03 or -- for a sig separator 17:59:07 Actually ----- 17:59:10 On that page 17:59:22 well, they're wrong 17:59:24 they should have used 4 17:59:35 Why's that 18:00:29 Deewiant: It lists the 9800 GTX+ one category below the 8800 GTX. What I've read suggests the opposite. 18:00:35 odd. 18:00:42 Deewiant: because 4 is standard! 18:00:48 What standard 18:01:20 ehird: The list is mostly about the hardware used 18:01:23 The Alex Smith standard. 18:01:28 Deewiant: Ah. 18:01:46 I.e. if one has more MHz than another but is otherwise identical it'll be higher, regardless of what tests say 18:01:53 Deewiant: ... but the 9800 GTX+ is a G92 and the 8800 GTX is a G90. 18:01:56 (Although in that case I'd be surprised if tests said it should be the other way) 18:02:00 One came out last year, the other in late 2006. 18:02:09 I'm incredibly surprised that it would be considered better by any measure. 18:03:22 It seems that the 8800 GTX has 86.4 GB/s bandwidth whereas the 9800 GTX+ has 70.4 18:03:43 Other than that, the latter is indeed advertised as better. 18:03:47 ... it is? 18:04:05 Higher clock, fill rate, etc. 18:04:18 Based on a quick browse of nvidia.com, anyway 18:04:21 I'm amused that people can get so involved in high-ending their computers 18:04:33 ais523: this is mid-range 18:04:33 the graphics card on here's an Intel 915 18:04:39 which isn't even powerful enough to run Vista, really 18:04:46 and which I suspect is obsolete nowadays 18:04:49 and I want to play certain games, certain games that a lesser card would give me bad FPS on; I have researched this :-P 18:05:16 but god I hate the naming systems 18:05:29 9800 GTX+ < 8800 GTX, yes, a higher number and a + means it's worse! 18:05:31 agreed, graphics card naming is almost impossible to fathom 18:05:54 * ais523 vaguely wonders if AnMaster even has a 3D graphics card 18:05:56 4770? that's better than the 4830! why? because of the technology used inside, not the actual market, duh! 18:06:02 ais523: he has some GeForce 7780 or something 18:06:04 Both ATI and nVidia have a 9600 GT, at least 18:06:11 Deewiant: they do? 18:06:12 haha 18:06:14 ais523, yes... 18:06:16 ATI's is some two-three years old 18:06:20 nVidia's is fairly new 18:06:21 GeForce 7600 GS 18:06:29 Actually, older than that, I think 18:06:30 -!- oerjan has joined. 18:06:31 ais523: And CPU naming is getting crazy, too. 18:06:34 But anyway. 18:06:38 I blame AMD for purchasing ATi. 18:06:39 pikhq: well, it used to work on IC names 18:06:44 now, if 8800 is better than 9800, you'd expect 7600 to be even better right? 18:06:45 noooooooooooooooooooooooooo! 18:06:46 which also don't make a whole lot of sense, normally 18:06:54 as they were originally designed just to be unique, not ordered 18:07:35 Deewiant: I'm googling, everyone's saying the 8800 GTX is good but not as good as the 9800 GTX, let alone the 9800 GTX+ 18:07:35 now, if 8800 is better than 9800, you'd expect 7600 to be even better right? <-- In that case my old GeForce 3 Ti 200 would be AWESOME 18:07:49 sadly it is no longer functional 18:07:52 AnMaster: shit man, Voodoo cards! 18:07:54 which is why I changed to this card I have 18:08:02 ehird, what was before that 18:08:06 anyone remember games installing an extra menu item for Glide? 18:08:08 I doooooo 18:08:30 ehird, Glide? I remember seeing something about Glide in mupen64... but that is all 18:08:45 AnMaster: Glide is the 3dfx Voodoo cards' hardware-based implementation of some OpenGL calls 18:08:51 heh 18:09:31 theory: someone should write a generic graphics card 18:09:41 which automatically optimises graphicsy stuff 18:09:53 so you can use pure-software implementations of the renderer, and it runs really fast anyway 18:09:59 that way it's future-proof to standards improvements 18:10:01 ais523: err, graphics cards ARE generic 18:10:04 to a large degree 18:10:08 ehird: I mean, even more than that 18:10:10 there's a reason they're called -PUs 18:12:09 Deewiant: http://i7.techpowerup.com/reviews/Zotac/GeForce_9800_GTX_Plus_Amp_Edition/images/crysis_1920_1200.gif 18:12:14 9800 GTX+ is 22.6 FPS 18:12:15 and, wait for it 18:12:18 the 8800 GTX is 18:12:20 22.6 FPS 18:12:22 Amazing. 18:12:35 Hey, it has about 200MB more memory with which to be at the same FPS, though. 18:13:22 hmm... does crysis automatically try to run at the fastest FPS possible? 18:13:35 DNA Maze runs at 20 FPS, for instance, even if the hardware is capable of more 18:13:48 which it is, considering how simple the graphics are 18:14:41 Most games don't have FPS limiters. 18:15:17 well, there's nothing animated in DNA Maze 18:15:18 Personally, I'm in favor of limiting FPS to the monitor's refresh rate. 18:15:28 Updating in the middle of a refresh just looks like shit. 18:15:30 nothing changes faster than once every 5ms 18:15:37 pikhq: Which means what, exactly, on LCDs 18:15:58 Deewiant: LCDs still have a 'refresh rate'. 18:15:59 so limiting to 20 FPS means that updates can happen exactly when they need to 18:16:02 rather than waiting until the next frame 18:16:09 They can only update the LCD so often. 18:16:48 * pikhq notes that FPS might be a better unit than hertz for this purpose, but alas, LCDs themselves are still statted in Hz refresh rate. 18:17:49 That's 60 Hz for just about all LCDs, isn't it 18:18:27 Not all LCDs. 18:18:33 Though that is the norm. 18:18:39 08:10:39 is this BF loop balanced or not: [>[-]+++[>++>++<<],++.<+] 18:18:46 08:10:58 it is also equal to [>[-]+++[>++>++<<]] 18:18:59 oerjan, it is balanced 18:19:16 i'd say the first is clearly balanced, and the second is then essentially balanced 18:19:23 oerjan, interesting. 18:19:45 the second could be phrased as: a loop that never finishes is balanced for most purposes 18:20:19 it's balanced for the purpose of optimising, anyway 18:20:31 the first is syntactically balanced 18:20:31 if you get past the loop, the tape pointer didn't move 18:20:53 the second is logically balanced if you define that as "ends up in the same place _if_ it finishes" 18:21:01 yes 18:21:05 which is a good definition 18:22:11 Especially for optimisation purposes. 18:22:14 btw the second relies on cell size being even :D 18:22:46 there is nothing that says every bf variant _needs_ to use a power of 2 >:) 18:23:10 presumably this one does, though 18:23:15 yeah 18:25:29 i vaguely recall redcode uses a prime, at least for memory size, maybe FukYourBrane could use the same 18:26:11 oerjan: FukYorBrane has a prime number of possibilities for each tape element, I think 18:26:22 whereas BF Joust uses 256, partially because it needs to be even 18:26:33 (as i recall, because this prevents you avoiding your own code just by placement with a fixed step scan) 18:26:54 wait, no, redcode has a power of 2 for memory size 18:27:02 no, wrong again 18:27:07 it's round but not a power of 2 18:27:08 such as 8000 18:27:59 if it were a multiple of 256, say, you could just move a little bit off your code, and then move right 256, kill & destroy, repeat 18:28:34 and you would only hit enemies 18:28:44 oerjan: that's a common tactic 18:28:47 but using small numbers like 4 18:28:51 rather than large numbers like 256 18:28:57 256 is likely to miss your enemies too 18:29:01 yeah i just realized that 18:29:06 ais523, btw, have you noticed google maps work without js? 18:29:16 not sure if you were here when it was discussed 18:29:36 it's balanced for the purpose of optimising, anyway <-- not exactly. It is unbalanced for some optimisations 18:29:54 AnMaster: which? 18:29:54 some of those working inside the loop 18:30:02 18:13 ais523: hmm... does crysis automatically try to run at the fastest FPS possible? 18:30:04 yes 18:30:13 but most games have vsync (= limit to my LCD's fps) 18:30:56 oerjan: The second doesn't rely on cell size being even, assuming cells wrap. 18:30:59 ok maybe they decided that was fun and made it a reason for _not_ using a prime 18:31:05 It relies on cell size being *greater than 2*. 18:31:14 Erm. Greater than 3. 18:31:25 pikhq: by cell size i mean number of possible values, not bits 18:31:43 -!- asiekierka has joined. 18:31:47 if it's odd and wrapping, then that _will_ eventually reach 0 again 18:31:48 Hi 18:31:54 ais5523: How's the scorelist 18:32:03 did anyone write their own BF Joust stuff 18:32:03 oerjan: [>++>++<<] 18:32:06 btw the second relies on cell size being even :D <-- where does it do that? 18:32:08 No, it won't.' 18:32:17 pikhq: yes it will 18:32:27 oerjan, index_diff = unchanged 18:32:28 It doesn't do anything to the cell that's being tested. 18:32:30 look at it 18:32:43 pikhq: oh, i misread 18:32:44 What are you talking about? 18:32:52 bah 18:33:14 The cell values needs to not be 3 for it to run. 18:33:17 -!- Corun has joined. 18:33:21 -!- BeholdMyGlory has quit (Remote closed the connection). 18:33:22 thought it did like [++>++<] 18:33:31 -!- kar8nga has joined. 18:33:31 "8800 GTX XXX Edition" 18:33:34 -!- BeholdMyGlory has joined. 18:33:45 * pikhq shan't behold. 18:33:48 oerjan, well that one I can't yet detect as infinite, haven't implemented that type of optimisation of it yet. 18:33:58 * oerjan swats some glory -----### 18:34:04 XCorp 8800 GTX XXXXX Xdition 18:34:15 AnMaster: ah 18:34:21 oerjan, I see some code for it in esotope-bfc, but it messes up my head trying to work it out 18:34:25 you could do it trivially 18:34:40 it is mathy stuff ;P 18:35:00 it's just greatest common divisor 18:35:03 AnMaster: actually that one _isn't_ infinite for all initial values, regardless 18:35:10 only for odd ones 18:35:15 Everyone should know GCD 18:35:19 oerjan, yes 18:35:19 oerjan, http://rafb.net/p/6pRr7k45.html 18:35:20 so it needs context 18:35:45 oerjan, I have context: I know the index cell is a constant already 18:35:50 the constant propagation pass told me 18:36:14 ehird, I know how gcd is defined 18:36:48 yeah i noticed 18:37:14 oerjan, otherwise I wouldn't have been able to remove the stuff after the infinite loop above. 18:37:50 You know, I wnnt to make a Boolf**k machine 18:37:53 I want* 18:38:08 AnMaster: that paste also requires you to know the Euclidean algorithm with coefficients 18:38:36 (for let u * m + v * w = gcd(m,w)) 18:38:36 I only wonder how will I do the I/O 18:38:53 *extended Euclidean algorithm 18:39:11 oerjan, I'm not familiar with "extended gcd" 18:39:20 * AnMaster googles 18:39:24 AnMaster: it's not the gcd that is extended 18:39:43 it's the algorithm for finding it that is extended so you get the other coefficients in that equation 18:39:52 mhm 18:40:29 which is very useful for solving linear integer equations 18:40:55 * ehird compares the 8800 Ultra w/ the 9800 GTX+ 18:41:10 oerjan, *reads on wikipedia* 18:41:30 asiekierka: Serial line. 18:41:36 Assuming it's on an FPGA. 18:41:39 Not hard to do. 18:41:51 The worst thing about picking a graphics card is that gamers don't know the meaning of grammar. 18:41:57 "9800GTX+ is faster than 8800 Ultra"←kay 18:42:10 ehird, that seems fine? 18:42:18 AnMaster: that was a separate thing 18:42:23 pikhq: It's using 7400 series and the like 18:42:27 two different lines 18:42:30 ehird, you might want to add some "the" though 18:42:32 but ok 18:42:33 And I wanted to do a LED 18:42:37 thought it was an example 18:42:47 possibly two buttons to do < and > manually 18:42:59 -!- inurinternet has quit (Read error: 60 (Operation timed out)). 18:43:31 asiekierka: I suggest a bank of 8 toggle switches. 18:43:56 Oh, and a button to enter it in. 18:44:04 Output via 8 LEDs. 18:44:20 ...Why 8 LEDs? 18:44:30 A byte. 18:44:31 A char. 18:44:35 A byte. 18:44:36 I know 18:44:39 AnMaster: btw that is the basis for several of the algorithms in [[Brainfuck constants]] on the wiki 18:44:41 but I'm doing Boolf**k. 18:44:41 A char can be any size > 8 bits, pikhq. 18:44:48 asiekierka: Boolfuck outputs in bytes. 18:45:01 oerjan, hm 18:45:07 Outputs the bit under the pointer to the output stream. The bits get output in little-endian order, the same order in which they would be input. If the total number of bits output is not a multiple of eight at the end of the program, the last character of output gets padded with zeros on the more significant end. If the end-of-file character has been input, outputs a zero to the bit under the pointer. 18:45:08 It just builds those bytes up 8 at a time. 18:45:18 s/8/bits/ 18:45:19 (for finding multiplicative inverses mod 256) 18:45:52 oerjan, I see the algorithm on wikipedia for it. I understand it. But I don't understand why it is useful :/ 18:47:31 ehird: Ehh? 18:47:34 AnMaster: as i said, it's useful for finding the integer solutions to linear equations 18:47:37 So I am not implementing Boolf**k 18:47:49 I am implementing a variation of BF operating on bits, also in I/O 18:48:01 AnMaster: or do you mean useful for brainfuck? 18:48:15 oerjan, both :/ 18:48:36 but actually 18:48:44 is there anything better (or easier) to implement 18:48:49 better as in more useful 18:48:54 and no not Befunge 18:48:56 that calls for FPGAs 18:49:26 see, if you have something like [-]+++[---->++++++<] 18:49:59 Well, I think i will make an OISC 18:50:05 asiekierka: As in more useful? 18:50:06 Brainfuck proper. 18:50:11 As in easier? OISC. 18:50:24 pikhq: But both of these have been done! 18:50:32 The latter on 7400 series, too 18:50:46 As in cooler? Malbolge. 18:51:14 oerjan, How does it help for the compiler wanting to find if loop is finite or infinite.. That is the bit I don't understand. 18:51:27 pikhq: I don't have a house big enough to store wirewrapped 7400 series chips for Malbolge! 18:51:34 then that is essentially solving the problem: is there a y such that 3+4*y = 256*k ? 18:51:40 ... *Wirewrapped* 18:51:42 Oh. 18:51:45 well 18:51:48 not necessairly 18:51:50 maybe proto'd 18:51:54 but I'm trying to NOT solder 18:51:54 oerjan, ah. I see 18:52:06 Fine, then. Create your own architecture. 18:52:12 Port a C compiler to it. 18:52:16 and if so, the result is 5* the smallest such y 18:52:27 meh, seems like my i7 965 XE and the 9800 GTX+ will suck up >430W at load, so I need a diff power supply 18:52:37 pikhq: I don't have a year to waste (See: BMOW) 18:52:38 If you're feeling especially motivated, port Minix to it. 18:52:44 oerjan, 5* the smallest such? 18:52:47 Magic-1. :) 18:52:49 err 18:52:50 what 18:52:53 AnMaster: er, 6 18:52:54 I don't have time, either (See: Magic-1) 18:52:56 http://www.homebrewcpu.com/ 18:53:00 i miscounted my +'s 18:53:02 Ah. 18:53:03 ehird: get a nehalem xeon 18:53:04 I also don't want something very hard (See: both) 18:53:12 asiekierka: Be hard-core! 18:53:21 so i don't know what to do, really 18:53:21 see, if you have something like [-]+++[---->++++++<] 18:53:26 I want something simple but useful 18:53:27 oerjan, hm. The fact that I'm representing - as "add 255" makes this slightly complex I think... 18:53:29 bsmntbombdood_: Why? All that buys me is ECC (= slower ram), another $500 out of my pocket, having to buy a server motherboard, and lameness. 18:53:33 I'm entirely sure you can write a C compiler for OISC 18:53:36 bsmntbombdood_: Also, I doubt it'll suck up LESS watts... 18:53:48 ehird: [two] xeons 18:54:03 bsmntbombdood_: Make that another $2,000 out of my pocket vs an i7. 18:54:07 And even more wattage. 18:54:07 AnMaster: oh. well in principle multiples of 256 don't matter (you could do most of this with (mod 256) arithmetic) 18:54:11 And even more on the mobo. 18:54:15 asiekierka: Oh, I'm sure you could. 18:54:22 Maybe even get LLVM to target it. 18:54:49 also, only a 9800? 18:54:58 bsmntbombdood_: no, not only a 9800 18:54:59 oerjan, it makes stuff easier elsewhere to not have any "add -1" or similar. Since erlang is all bignum it is a pain to make sure the range of merged nodes have the proper value if negative. Much easier if the code only needs to deal with positive values. 18:55:11 bsmntbombdood_: a 9800GTX+, which is competitive-with-and-sometimes-better than a Radeon HD 4850 18:55:24 ehird: it's like a year old 18:55:24 9800GTX+ >>>>>>> 9800GTX >>>> 9800GT 18:55:28 bsmntbombdood_: yes. Yes it is. 18:55:33 So's the 4870 and it still screams. 18:55:35 AnMaster: oh. well adding 255 might actually fit this equation slightly better 18:55:44 oerjan, really? hm 18:55:49 bsmntbombdood_: And I want something I can cool passively with the Accelero S1 Rev2. 18:56:05 since you then only have a few variables to worry about being negative 18:56:18 bsmntbombdood_: Besides, the 9800GTX+ manages Crysis @ 1920x1080 at 41fps (with some settings bumped up, IIRC) 18:56:23 So it's not exactly crappy. 18:56:29 9800GTX+ >>>>>>> 9800GTX >>>> 9800GT <-- can be simplified into a + But that is a dead store. So generated code could be int main(void) { return 0; } 18:56:31 * AnMaster runs 18:56:31 i dunno what crysis is 18:56:36 AnMaster: wait, did i say +4? i meant -4 18:56:44 bsmntbombdood_: An incredibly GPU-intensive game. 18:56:58 Even the 4870X2 (two 4870s stuck together) can't max out all of the settings to the very max and still be playable at 1920x1080. 18:56:58 AnMaster: but you could make that 4*255*y to fit your framework perfectly 18:57:14 maybe GTX 295 in 3xSLI could do it 18:57:28 AnMaster: did you literally run or just joking? 18:57:32 oerjan, hm... only thing that can be negative is offset, and mov nodes (< is {mov, -1} basically) 18:57:50 AnMaster: did you literally run or just joking? <-- ehird's line? Joking. 18:57:59 and yes joking 18:58:02 about the run 18:58:19 i assume so, unless you have a very portable irc client 18:58:27 -_- 18:58:30 which people might, nowadays 18:58:53 bsmntbombdood_: So I don't know what you meant by the 9800GTX+ not being good. 18:58:53 http://www.arctic-cooling.com/catalog/product_info.php?cPath=2_&mID=101. 18:58:59 now that's a big cooler 18:59:04 ehird: it's just old 18:59:14 bsmntbombdood_: Dude, there aren't many graphics cards much newer than it. 18:59:21 bsmntbombdood_: Heck, the i7 came out last year! 18:59:34 but there's nothing newer than it 18:59:43 bsmntbombdood_: The i7 came out *18 months ago* 18:59:46 AnMaster: so you are not having anything negative, which means you do all loop checks by checking if things are divisible by 256, i assume 18:59:47 And newer != better. 19:00:00 The 9800GTX+ is a powerful card, and the most powerful you can cool with the S1 Rev.2 apart from a 4870. 19:00:05 oerjan, um? 19:00:13 bsmntbombdood_: that is not a big cooler 19:00:16 bsmntbombdood_: http://www.arctic-cooling.com/catalog/product_info.php?cPath=2_&mID=105 this is a big cooler 19:00:20 ehird: why do you want it passive? 19:00:29 (in ctx: http://www.ngohq.com/attachments/news/769d1198274397-arctic-cooling-launches-accelero-s1-rev-2-100_1050.jpg) 19:00:31 oerjan, I have a cleanup pass that runs after all the node merging passes which adjusts out of range values 19:00:32 AnMaster: so 3 + (4*255)*y = 256*k fits what you would do with [-]+++[---->++++++<] 19:00:43 oh damn, that is a big cooler 19:00:43 optimise([#bfn{ ins = add, val = Val } = H|T], Result) 19:00:43 when Val > 255 -> 19:00:43 optimise(T, [H#bfn{val = Val rem 256 } |Result]); 19:00:45 like that 19:00:53 bsmntbombdood_: For silence. Plus, I don't really need more performance than the 9800GTX+. Barely anyone except HARDKOR LEET GAMERZ do. 19:01:02 AnMaster: oh well. anyway that just becomes an additive constant to k, which we don't care about 19:01:03 and irc client messed up indention 19:01:03 So since I _can_ cool it passively... 19:01:07 And yeah. 19:01:10 It is gigantic. 19:01:14 ehird: and hardcore elite scientists 19:01:20 I want to make a 16-command no-parameter set 19:01:22 bsmntbombdood_: they use PS3 clusters ;-) 19:01:24 Probably sort of extended BF 19:01:58 AnMaster: btw of course that equation has no integer solution 19:01:58 oerjan, and you lost me. Best way is to try to replicate what esotope-bfc did here. Since it is in python it also has bignums iirc. 19:02:19 not sure if it uses -1 or +255 though 19:02:20 AnMaster: was that paste from esotope-bfc? 19:02:24 AnMaster: do you know anything about mathematics? :) 19:02:29 oerjan, the comment? yes 19:02:32 I suggest addition to an address, subtraction from an address, and conditional jumping. 19:02:43 Only marginally easier than Brainfuck! 19:02:55 ehird, some, but the Swedish school system sucks. Ending high school without knowing about linear algebra 19:02:55 :( 19:03:10 AnMaster: Can't suck as much as the UK system. 19:03:16 oh well 19:03:20 But tbh, I'm rather maths-retarded myself; mostly learned from WP. 19:03:20 ehird, I'd say it is equally bad 19:03:27 AnMaster: One can finish university without knowing about linear algebra. 19:03:31 Well, I currently have NOP, NOP, <, >, +, -, M->A, A->M 19:03:52 * oerjan goes back to web browsing 19:04:03 and SKP - Skip the next command 19:04:05 pikhq: are there any instruction sets that don't have zero params on all instructions that have a constant length? 19:04:07 that is, nop is 00 00 19:04:08 pikhq, depends on what you study. Lawyer? Certainly. CS? No. 19:04:11 to pad out for the argument, say 19:04:18 AnMaster: CS involves linear algebra? 19:04:20 well, I use 4 bits per command 19:04:24 ehird, easier for you to learn from WP. It is your native language. 19:04:27 AnMaster: Coulda fooled me 19:04:35 also, isn't there a wikipedia in Swedish? 19:04:38 :P 19:04:40 ehird, at least CS here has linear algebra 19:04:43 ehird, yes but it sucks 19:04:43 ehird: It is mandatory for most CS degrees. 19:04:43 well 19:04:46 pikhq: well yeah 19:04:50 I guess I'm just cynical 19:04:54 since there are a lot of idiots doing CS 19:04:55 ehird, 1) very incomplete 2) should be called "Simple Swedish" 19:05:07 19:04 ehird: pikhq: are there any instruction sets that don't have zero params on all instructions that have a constant length? 19:05:07 19:04 ehird: that is, nop is 00 00 19:05:46 ehird, what is the context... 19:05:58 assuming time is UTC I can't find it 19:05:59 AnMaster: none 19:06:07 pikhq: AnMaster: One can finish university without knowing about linear algebra. 19:06:07 19:03 asiekierka: Well, I currently have NOP, NOP, <, >, +, -, M->A, A->M 19:06:08 19:03 oerjan goes back to web browsing 19:06:10 19:04 asiekierka: and SKP - Skip the next command 19:06:12 19:04 ehird: pikhq: are there any instruction sets that don't have zero params on all instructions that have a constant length? 19:06:15 ehird, today? 19:06:15 19:04 ehird: that is, nop is 00 00 19:06:17 19:04 AnMaster: pikhq, depends on what you study. Lawyer? Certainly. CS? No. 19:06:19 19:04 ehird: to pad out for the argument, say 19:06:21 ... 19:06:23 well, no 19:06:23 A few fucking seconds ago! 19:06:25 no SKP 19:06:27 now there's 19:06:29 It's 19:06 19:06:41 ah right 19:06:42 there 19:06:45 hmm 19:06:50 ehird, I thought it was UTC 19:06:54 what are the addresses of eax,etc on x86? are they constant? 19:07:03 SNZ (Skip if A != 0), SKZ (Skip if A = 0), JMP (to set address), SKP (just skip), SAH (A->high byte of addr), SAL (A->low byte of addr) 19:07:09 I have 0000-1101 occupied 19:07:10 ehird, addresses of eax? 19:07:14 cuz i have 2 nops currently 19:07:20 AnMaster: addresses of eax, etc. 19:07:21 ehird, it is a register. You can't have a pointer to a register 19:07:23 do you know what etc. means? 19:07:25 so 2 commands more 19:07:27 what can I add? 19:07:32 AnMaster: not neccessarily 19:07:33 oh 19:07:36 Add A to cell and Subtract A from cell 19:07:37 you could have a ptr-to-register just fine 19:07:42 ehird, memory mapped registers yeah 19:07:42 by reserving the first N bytes of memory-space 19:07:48 but iirc x86 doesn't have that 19:08:00 it seems like a logical design decision 19:08:00 ehird, also that would mess up NULL pointers badly 19:08:03 instead of having 3 instructions 19:08:05 constant, memory and register 19:08:07 well it could work with non-zero null 19:08:08 you can just have 19:08:12 but that would be messy 19:08:12 constant, memoyr 19:08:14 *memory 19:08:17 AnMaster: last N bytes, then 19:08:35 ehird, of the possible address space, not real memory installed right? 19:08:40 AnMaster: right 19:08:53 could work. But iirc x86 doesn't do it like that 19:08:58 x86 is in fact a mess 19:09:07 So the opcodes are NOP, MVL, MVR, ADM, SBM, PLA, PHA, SNZ, SKZ, JMP, SKP, SAH, SAL, ADA, SBA 19:10:01 AnMaster: actually, all opcodes being two bytes wouldn't work 19:10:07 you need it to be >machine word 19:10:10 for load-into-X 19:10:10 and yes, setting an address is just setting them in 2 cells and then you can just do PLA, SAL, MVR, PLA, SAH, if you store low-high 19:10:31 if high-low, swap SAH and SAL 19:11:18 so to set a cell to 3 you do ADM ADM ADM 19:11:45 to clear the cell you do (assuming $0000 is command 1, $0001 is command 2...) 19:11:56 or nah 19:11:58 this is a mess 19:12:00 and won't work 19:12:03 kay, instruction = 5 bytes 19:12:18 pikhq: think I should align at a boundary more than "byte"? :P 19:13:01 ehird: Kilobyte. 19:13:07 pikhq: per instruction‽‽‽‽ 19:13:11 hmm, i'd rather 3 bytes tbh... 19:13:25 pikhq: any idea how to do a load-word-into-X when I can only pass a pointer? 19:13:27 not a value :-P 19:13:29 * pikhq creates an architecture with 32-bit aligned instructions. 19:13:34 And 8 bits. 19:13:39 ooh, maybe: 19:13:39 Erm.\ 19:13:42 8-bit addressing. 19:13:44 :p 19:13:49 a nop is 00 XXXX, where the XXXX is just ignored 19:13:51 now 19:13:55 LOAD looks at the next instruction 19:13:55 so 19:13:59 LOAD ptr; NOP value; 19:14:00 >:D 19:14:17 pikhq: discuss my evility 19:14:19 ehird, x86 has very varying instruction size 19:14:25 AnMaster: yep 19:14:27 -!- MizardX- has joined. 19:14:30 oh not talking about x86= 19:14:31 ? 19:14:32 I'm trying to make the cpu not have to guess about it 19:14:35 i'm designing an instruction set 19:14:42 where all instructions are exactly the same length 19:14:47 but that's OP XXXX 19:14:47 ehird, you could do all loads from memory :D 19:14:52 and that's just enough to store one pointer 19:14:53 that would work, but be silly 19:14:58 wait hm 19:14:59 no 19:14:59 and you need to load _something_ somehow 19:15:01 so 19:15:02 still machine word 19:15:05 LOAD ptr; NOP value 19:15:07 ehird, what about "increment" 19:15:08 :D 19:15:10 since it has to be NOP XXXX 19:15:14 and set zero 19:15:14 the XXXX can just be ignored 19:15:16 and LOAD can look at it 19:15:17 with those you can load 19:15:18 :D 19:15:20 AnMaster: no 19:15:24 that's not interesting, that's boring and slow 19:15:34 ehird, yes of course 19:15:58 -!- MizardX has quit (Nick collision from services.). 19:16:01 otoh, LOAD looking at the next instruction is Evil(TM) 19:16:21 ehird, what about "load and skip next instruction"? 19:16:23 -!- MizardX- has changed nick to MizardX. 19:16:30 ehird: As is using NOP for storage. :) 19:16:33 then have the next instruction as an operand 19:16:40 AnMaster: that's just variable instruction size 19:16:47 ehird, yeah. But with a different name 19:16:48 :D 19:16:56 variable instruction size is so inelegant :P 19:17:25 ehird, are there any major architectures that use fixed size? Possibly Itanium I guess... 19:17:35 keyword being Very Long instruction word 19:17:37 AnMaster: RISC architectures. 19:17:46 pikhq, so what about the load one there 19:17:56 pikhq, loading a constant into a register. 19:18:03 or referencing a constant memory cell 19:18:07 err 19:18:08 byte 19:18:13 i guess they just have OP XXXX YYYY and ignore Y when it isn't applicable 19:18:15 which is teh ugly 19:18:16 you need it larger than word size 19:18:30 ehird, that is variable instruction size basically 19:18:34 no 19:18:37 the instructions are all the same length 19:18:41 just parameters are often ignored 19:18:46 hm 19:18:51 still 19:18:52 uuuuuugly 19:18:55 more ugly than variable length 19:18:56 ehird, ah you mean HUGE instruction? 19:18:57 right 19:19:02 yeah that is uglt 19:19:05 OP XXXX YYYY 19:19:06 ugly* 19:19:08 two words plus a byte 19:19:19 ehird, that is a horribly uneven size to load 19:19:23 not efficient 19:19:36 ehird, btw, are you planning to implement some hardware with this? 19:19:40 AnMaster: maybe 19:19:44 prolly not FPGA if I do 19:19:47 that's so cheating 19:19:52 Logic circuits! 19:19:54 ehird, how is FPGA cheating 19:19:57 ah hm 19:20:02 AnMaster: it's too high-level ;) 19:20:09 ehird, ASIC? 19:20:10 :P 19:20:20 mmmmmmmmmmmnope 19:20:25 ehird, then what 19:20:28 TTL 19:20:34 Time To Live 19:20:36 err 19:20:39 http://en.wikipedia.org/wiki/Transistor-transistor_logic 19:20:51 http://www.homebrewcpu.com/ is made entirely with 'em 19:20:59 "It doesn't use an off-the-shelf microprocessor, but instead has a custom CPU made out of 74 Series TTL chips." 19:21:03 what asiekierka was talking about a bit befor 19:21:03 e 19:21:05 ehird, hm that have long signal paths 19:21:05 7400 19:21:16 it will be even slower than a FPGA... 19:21:49 so? 19:21:54 the magic-1 runs at 4mhz 19:22:00 that's faster than an apple IIe 19:22:04 and faster than a C64 19:22:04 hehe 19:22:08 and faster than a PDP-11 19:22:12 and faster then the IBM PC/XT 19:22:18 BMOW doesn't use only 7400's, it also uses GAL's 19:22:21 ehird, the same speed as a PIC12F629 iirc 19:22:22 but it's still great 19:22:24 and a little slower than the Macintosh 512 19:22:33 so... yeah, it's not exactly slow 19:22:40 for a homebrew type thing 19:22:41 ehird, which is a computer in a package with 8 legs. 19:22:44 it outputs image to VGA, has sound, multitasking, cool stuff 19:22:47 ehird, agreed 19:22:53 and it will be there (with Magic-1) at the Maker Faire or something 19:23:00 asiekierka: BMOW? 19:23:10 http://hackaday.com/2009/02/27/bmow-a-home-made-cpu/ ah. 19:23:50 "I was pleasantly surprised to find that Magic-1 has been quite solid at 4 Mhz." 19:23:54 Hardcore overclocking 19:24:10 BMOW runs at ~2MHz 19:24:19 -!- MizardX has quit ("Proclamation of invalidity!"). 19:24:34 -!- jix has quit (Connection timed out). 19:25:00 "Magic-1 continues to run rock solid at 3.75 Mhz with the new (and last - I hope) wiring modifications. There is only one hardware item left that causes me some minor concern: Magic-1 can't stand the heat. It is located in the master bedroom of my house (which, because I live in a very mild coastal climate, has no air conditioning). Sometimes, though, the temperature in the room gets up to 80 F. When this happens, Magic-1 starts failing. A typical 19:25:03 heat-related failure is when running the Original Adventure program and it aborts at start-up with a vocabulary initialization failure. Once the room cools down in the evening, though, correct behavior resumes. " 19:25:07 Then, the invention of the noisy computer fan. 19:25:14 FFFFFFFUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUUU— 19:26:53 Well, just for jollies I tried running the old Whetstone benchmark on Magic-1 to see how my floating point performance is. I knew it would be really, really poor - and it is. I had to do a bit of modification to the benchmark source because my score was so low that it didn't register in the normal "Millions of Whetstones per Second" metric. As I predicted, it would be better for me to report Magic-1 in the "Dozens of Floating Point Operations per seco 19:26:55 nd". 19:26:57 Anyway, I estimate that a full run of the benchmark will take somewhat more than 20 hours. My shortened run ended up at around 0.000134 MWIPS, or 11.2 DWIPS. 19:27:00 I'll probably go ahead and do the full 20-hour run this weekend and officially submit my results :-). 19:27:07 11 dozen floating point operations a SECOND! WOW! 19:27:25 ehird, http://www.homebrewcpu.com/Pictures/P5150118.JPG is cool 19:27:34 AnMaster: yes, it's like WOPR 19:27:41 in fact, a computer where that is meaningful is cool 19:27:49 but at 4 MHz 19:27:50 hm 19:27:57 I don't think you could read it fast enough 19:28:03 (http://www.pldos.pl/bogus/hardware/komputery/imsai/pics/wargames_wopr.jpg) 19:28:07 (from the movie Wargames) 19:28:09 (in case you didn't know) 19:28:23 ehird, never heard of that movie 19:28:32 it's great. 19:29:21 ehird, why does Magic-1 have a switch labled "DMA Req" 19:29:31 it seems so out of place 19:29:36 with that style of technology 19:29:40 why? 19:30:04 ehird, anything with lamps on the front to show registers is by definition way too old to have DMA in my mind. 19:30:33 AnMaster: it runs minix! 19:30:43 ehird, yeah. Collision 19:31:04 AnMaster: modern computers have lights that flicker when the hard disk is used 19:31:07 equally as silly 19:31:20 ehird, well yeah. 19:31:27 ehird, macs doesn't iirc? 19:31:32 *don't 19:31:35 And no, they don't. 19:31:39 Not even the Mac Pro. 19:32:03 -!- inurinternet has joined. 19:32:21 wb inurinternet 19:33:04 AnMaster: any more ammonia news? 19:33:37 AnMaster died of ammonia poisoning 19:33:43 ah 19:33:57 nah 19:34:19 AnMaster: you seem dead to me 19:34:23 ehird, and iirc they contained it 19:34:27 how boring 19:35:13 * pikhq suggests a system architecture 19:35:29 pikhq: ? 19:35:38 There are two instructions: NOP and HCF. 19:35:49 Ah. 19:35:53 http://www.homebrewcpu.com/Magic1.pdf <-- seems like it has CF slot on the back 19:36:18 And it has a 1-bit addressing scheme. 19:36:18 or side or whatever 19:36:27 -!- MizardX has joined. 19:36:32 pikhq: No, I think four would be best. hcf, sfi (set fire intensity), sfg (set fire geometry), str (stutter). 19:36:35 Sorry, it has exactly one bit of memory. 19:36:48 Which can be either 0 or 1. 19:36:52 0 is a noop. 19:36:53 1 is HCF. 19:36:54 You fiddle about with them all the time to perform the computation, you see. 19:36:58 Also, it has an infinite tape. 19:37:04 To modify it, make the fire burn it the right way. 19:37:12 Therefore, it is turing complete. 19:37:18 Jumps? Just burn the bits on the IP. 19:37:39 It can be implemented by a light switch wired up to a power source and a spark plug in a barrel of gunpowder. 19:37:50 pikhq, :D 19:37:51 pikhq: Mine's better because it is TC. 19:37:57 what's HCF 19:38:04 asiekierka: halt-and-catch-fire 19:38:05 asiekierka: Halt and Catch Fire. 19:38:10 oh 19:38:31 ehird: Your idea sounds like an interesting Turing tarpit. 19:38:35 pikhq: yep 19:39:35 Turing tarpits are the cool thing around here doncha know 19:39:40 ais523: how much do you think it'd cost to get an fpga/vhdl environment? yeah it's so uncool but :P 19:39:41 They're like the cha cha cha 19:39:47 pikhq: Turing firepit, rather 19:39:53 ais523: just a simple slow thing 19:40:15 ehird: not a ridiculous amount if you just want low-end stuff and demonstration boards 19:40:28 ais523: will the software run on linux? 19:40:45 -!- Sgeo has joined. 19:41:11 ehird: I don't know; I've only ever seen Windows software for that 19:41:16 WINE? 19:41:18 but IME, such software normally works well on WINE 19:41:19 well 19:41:21 what do you use? 19:41:39 well, at home I use VHDL, which is Linux, but it's just a simulator 19:41:40 Windows software for what? 19:41:43 -!- kar8nga has quit (Read error: 60 (Operation timed out)). 19:41:50 Sgeo: vhdl 19:41:55 *GHDL 19:42:00 at University, we have a choice of programs 19:42:06 any of which may or may not work in any given situation 19:42:15 I just want non-shitty tools for cheap and an fpga to run stuff on :-) 19:42:17 but the one you'll likely have to use is the low-end one that comes free with the demo board 19:42:26 as the others will tend to be price on request 19:42:27 Hardware stuff goes over my head 19:42:32 ais523: crappy, I assume? 19:42:39 not bad, but generally limited 19:42:42 Mostly because I'm cheap 19:42:47 sort-of, shareware-style 19:42:50 and I doubt there's a buzzing pirate scene around FPGA stuff :-P 19:42:53 -!- kar8nga has joined. 19:43:19 ehird: there's http://opencores.org which is not piracy, but open-source hardware is a great concept anyway 19:43:29 is it any good? 19:44:02 I haven't used it much 19:44:10 but from what I've heard, the stuff there's good and high quality 19:44:20 but you won't find anything amazingly big and complex 19:44:25 small processors are about the best you can hope for 19:44:38 Can an fpga be made into a video card? 19:44:40 ais523: wait, that's not fpga software? 19:44:41 you implied it was 19:44:43 Sgeo: yes 19:44:47 see caustic graphics 19:44:49 ehird: it's VHDL and Verilog code snippets 19:44:56 Sgeo: an FPGA can be made into anything 19:44:56 they are going to sell an fpga that does realtime raytracing 19:44:59 ~5fps 19:45:05 w/ accompanying software 19:45:08 very cool 19:45:20 the only difference between an FPGA and a dedicated chip are that the FPGA's slightly slower, and the dedicated chip is cheaper in bulk amounts 19:45:20 ah, do sell, it seems 19:45:44 ........so once I have fpgas, it's like being able to download hardware? 19:45:50 Sgeo: Yes. 19:45:51 Yeah. 19:45:52 yes 19:45:55 although, only digital hardware 19:45:59 **Sgeo's mind blows** 19:46:04 ehird, yes, it did 19:46:06 and you'll need extra components in order to do anything but digital inputs and outputs 19:46:25 I suggest doing CMOS analog circuitry. :p[ 19:46:33 i suggest nanotechnology 19:46:39 after all, Intel say they'll do it by 2015 19:46:42 pikhq: but then the whole thing would be type B, and you'd get massive distortion 19:46:45 for processors 19:46:53 you'd probably want to tweak CMOS somewhat to make a type AB circuit 19:46:58 pure-B is only good for digital stuff 19:47:04 ehird: Nano-robots? Forward the singularity! 19:47:08 "The 11 nanometer (11 nm) node is the technology node following 16 nm node. The exact naming of this technology node comes from the International Technology Roadmap for Semiconductors (ITRS). According to the 2007 edition of this roadmap, by the year 2022, the half-pitch (i.e., half the distance between identical features in an array) for a DRAM should be 11 nm, although Intel's "Architecture and Silicon Cadence Model" places it closer to the year 2015. 19:47:10 Intel's Pat Gelsinger claims that Intel sees a 'clear way' towards the 11 nm node.[1][2] " 19:47:11 ais523: It'd be crappy, yes. 19:47:12 11nm = nanotechnology 19:47:15 ais523: But it can be done. 19:47:17 pikhq: totally 19:47:22 I've been fantasizing about open-source hardware, but only when imagining 3d printing that can do electronics 19:47:34 Sgeo: that exists 19:47:37 it's called a fabricator 19:47:43 do you have millions of dollars? 19:47:44 pikhq: you'd have to be fighting the intrinsic digitalness of type-B circuits every step of the way 19:47:49 ehird: $200. 19:47:58 pikhq: $200 silicon fabricator? 19:48:00 WAT. 19:48:05 ehird: Reprap can print circuitry now. 19:48:08 Sgeo: an FPGA's more like a single chip which is generic enough to be configured however you want 19:48:09 Not silicon. 19:48:17 pikhq: Reprap isn't really the same thing... 19:48:20 I'm thinking "circuit board" type printing. 19:48:21 ais523, right, I understand 19:48:30 still 19:48:34 *fantasizes* 19:48:45 Sgeo: you can buy fpgas pretty cheap, I think 19:48:46 ehird: That + FPGA = ... Well, print your own motherboard, anyone? 19:48:54 ais523: how much is a demo kit? $200 or so right? 19:48:59 or was it £200 19:49:04 I'm not sure offhand 19:49:19 but there are only two main manufacturers, more or less like with x86 chips and with graphics cards 19:49:26 so it would be simple enough to just check both 19:49:31 who are they? 19:49:36 Xilinx and Altira 19:49:47 also, it seems that hardware companies gravitate to a two-manufacturer system 19:49:48 I wonder why? 19:50:16 -!- jix has joined. 19:50:28 Well, I want to create a small-size computational machine, as in, a machine geared for computations 19:50:30 :P 19:50:32 but nah 19:50:33 gotta go 19:50:33 Because it's a hard market to get into, but a lucrative market to stay in? 19:50:39 (typical technology company homepage) 19:50:41 You need to upgrade your Adobe Flash Player. You also need to have javascript enabled. 19:50:44 -!- asiekierka has quit. 19:50:52 ais523: the latter's typical too 19:50:53 pikhq: yes, but, it's never 3 or 4 19:50:54 hmm... round one goes to Xilinx, I think 19:50:57 it's always 2 19:51:01 ehird: I mean in how it looks 19:51:11 ais523: also, it's Altera 19:51:24 wait, no 19:51:28 that's a different FPGA company 19:51:29 now that's confusing 19:51:54 wow, that is confusing 19:51:59 and I'm not entirely sure I got the right one now 19:52:07 ah, yes 19:52:07 http://www.hitechglobal.com/boards/v5ddr3_pcie.htm 19:52:09 is that what I want? 19:52:12 Altera are the FPGA people 19:52:14 it seems terribly bloated! 19:52:23 and no, no way you'll be able to afford a Virtex 19:52:26 those are the high-end chips 19:52:27 ais523: Altira are about energy things 19:52:29 and ah 19:52:33 ais523: $10,000? 19:52:36 sort of thing? 19:52:39 about that, yes 19:52:39 also, it says PowerPC 19:52:43 that's not very custom! 19:52:52 they have integrated microprocessors nowadays 19:52:53 ais523: http://www.xilinx.com/products/devboards/index.htm 19:52:53 the high-end ones 19:52:56 starter kit? 19:52:58 spartan or coolrunner 19:53:07 spartan is $189 19:53:11 coolrunner is $39 19:53:12 spartan's the one that the university buys for students to mess with 19:53:15 coolrunner is CPLD 19:53:22 spartan is FPGA 19:53:24 The Xilinx Spartan®-3A Starter Kit delivers instant access to Spartan-3A FPGA device features such as SUSPEND power-saving mode, high-speed I/O options, DDR2 SDRAM memory interface, commodity flash configuration support, and FPGA/IP protection using Device DNA Security. 19:53:26 and I don't even know what CPLD stands for 19:53:28 god that's so bloated 19:53:30 ddr2? 19:53:33 fuck that! 19:53:35 ehird: they're all bloated 19:53:44 ais523: i want a lean mean custom FPGAing machine :-P 19:54:16 ais523: http://www.altera.com/products/devkits/kit-dev_platforms.jsp 19:54:19 altra prices 19:54:20 seems rather unlikely, given the way the industry operates 19:54:21 MAX II is $150 19:54:26 personally, I think the FPGA industry is insane 19:54:26 argh 19:54:27 it's CPLD 19:54:30 the whole industry, that is 19:54:33 ais523: howso? 19:54:45 The low-cost Cyclone® II FPGA Starter Development Kit is ideal for evaluating Altera's high-performance, low-power, 90-nm technology. By using this RoHS-compliant starter development kit, you will see 60 percent (on average) higher performance and 50 percent (on average) lower power than competing 90-nm, low-cost FPGAs. Several reference designs and demonstrations included in the kit make for a quick, "out-of-the-box" evaluation experience. 19:54:46 they get away with all sorts of ridiculous stuff because they don't have real competition 19:54:46 * pikhq observes once again that Magic-1 is pretty awesome 19:54:47 $150 19:54:59 8-Mbyte SDRAM 19:54:59 512-Kbit SRAM 19:55:00 4-Mbyte flash 19:55:04 ok, that's more minimal like I expected 19:55:11 how many LUTs? 19:55:13 http://www.altera.com/products/devkits/altera/images/fig1-cyclone2-staterkit.jpg 19:55:14 that looks nice 19:55:17 that's the measurement of how complex an FPGA is 19:55:24 ais523: doesn't say, what does it stand for? 19:55:29 lookup tables 19:55:33 although, only digital hardware 19:55:34 hm 19:55:37 it's how they're implemented 19:55:41 ais523: http://www.altera.com/products/devkits/altera/kit-cyc2-2C20N.html; it doesn't seem to tell you 19:55:52 ais523, why not analogue FPGAs 19:55:53 maybe the reference manual? 19:55:54 ehird: look at the chip, not the development kit 19:55:58 ais523: how? 19:56:04 AnMaster: ever tried to write a truth-table for an analog circuit 19:56:17 AnMaster: Because FPGAs are a very, very digital-only design. 19:56:34 ais523: http://www.altera.com/literature/lit-cyc2.jsp lots of cyclone ii shit 19:56:36 ais523, in fact I have a vague memory of having read about such a FGPA... Something about "evolving" hardware... 19:56:38 We're talking logic gates here. 19:56:45 AnMaster: that wasn't analogu 19:56:45 e 19:56:49 that just used analogue IO 19:56:54 it evolved FPGAs 19:57:07 -!- BeholdMyGlory has quit (Remote closed the connection). 19:57:08 it was actually an attempt to use a digital combinatorial circuit in order to evaluate an analog sequential function 19:57:09 ehird, it ended up using analogue bits inside iirc. 19:57:11 which in theory is impossible 19:57:14 but it ended up working anyway 19:57:16 ais523: 20K LEs 19:57:21 ais523, ah 19:57:23 does that help? 19:57:32 it's the EP2C20 19:57:39 ais523: LEs seems to be it 19:57:43 18,752 LEs 19:57:45 ah, yes 19:57:51 Altera use a different name for them 19:58:00 ais523: the max of the cyclone IIS is 68,416 19:58:03 so how good's 20K? 19:58:21 1 LUT is about equal to one byte of machine code 19:58:22 (the min of the cyclone IIs is 4K :-P) 19:58:23 in circuit complexity 19:58:29 ais523, iirc some components were not reachable, but when taken away the FPGA stopped working? 19:58:32 ais523: ah, so it controls how advanced your machine can be? 19:58:36 or whatever it was 19:58:36 AnMaster: yes 19:58:37 ehird: yes 19:58:45 ais523, speculation about induction iirc 19:58:45 ais523: what would, say, MIPS be? 19:58:46 roughly 19:58:53 a MIPS cpu 19:58:57 nice simple RISC 19:58:58 quite a lot more than that, I imagine 19:59:00 ehird, 0 SEK. 19:59:06 AnMaster: wut 19:59:14 ais523: ok, what about a really simple embedded ARM 19:59:17 200K? 19:59:20 I've managed to use huge numbers of LUTs before just trying to do multiplication 19:59:28 ehird, My ISP sent one for free. My ADSL modem runs on MIPS it seems 19:59:32 a 32-bit multiplier costs a lot, if you don't have a separate one on the chip 19:59:36 ehird, so do quite a few other ones that can run dd-wrt 19:59:37 and sich 19:59:40 -!- BeholdMyGlory has joined. 19:59:41 such* 19:59:46 AnMaster: I didn't ask price 19:59:47 I asked LUTs 19:59:55 RAM uses a lot of LUTs too (one per /bit/ of RAM, plus a few more), but nearly all FPGAs have on-board RAM separate from the LUTs 19:59:56 ehird, oh misread 19:59:59 ais523: i'd hate to be making a totally-awesome minimalist CPU and then run into a LE limit 19:59:59 ehird, btw what is a LUT 20:00:04 AnMaster: look up table 20:00:07 ah 20:00:09 this FPGA has 20K of them 20:00:23 ais523: yeah—239,616 bits of RAM 20:00:40 I love the way FPGA RAM is measured in bits 20:00:52 because the word length depends on how you connect them up 20:00:57 ais523, at least there is no 1000 vs 1024 thing then 20:00:58 :D 20:01:00 ais523: it's 39 kilobytes! 20:01:01 er 20:01:02 29 20:01:19 ehird, how large is your byte 20:01:23 ehird: is that the dedicated RAM, or the RAM if you use all the LUTs as RAM too? 20:01:24 6 bits right? 20:01:27 AnMaster: lawl 20:01:36 some manufacturers give the second value to make the chips look better than they are 20:01:40 ais523: don't even know 20:01:42 but 20:01:45 ehird, that is quite possible! 20:01:55 wait, wait. 20:02:03 19:54 ehird: 8-Mbyte SDRAM 20:02:03 19:54 ehird: 512-Kbit SRAM 20:02:04 19:54 ehird: 4-Mbyte flash 20:02:06 ais523: 20:02:08 so it has external ram at well, at least 20:02:10 my board 20:02:20 how much do you think I could fit into 18,752 LEs? 20:03:22 I'm trying to find some data on that atm 20:03:27 thanks 20:03:57 wow, it supports 640x480 VGA 20:03:59 at 60hz! 20:04:06 it has a PS/2 connector 20:04:07 one of them 20:04:09 keyboard or mouse 20:04:09 they come with lots of useless features, normally 20:04:10 take a pick 20:04:25 ais523: like a 24-bit audio codec 20:04:34 also, 18 LEDs 20:04:36 8 green, 10 red 20:04:43 4 seven-segment displays—that's FOUR calculators! 20:04:50 ais523: a "USB-Blaster" port, whatever that is 20:04:52 an SD card connector! 20:05:03 microphone-in, line-in, line-out 20:05:05 this is truly amazing 20:05:25 those are one-digit seven-segment displays 20:05:30 ais523: ah 20:05:34 that's less than one calculator! 20:05:56 Audio CODEC The development board provides a Wolfson WM8731high-quality, 24-bit, 20:05:56 sigma-delta audio encoder/decoder (CODEC) for applications such as 20:05:57 MP3 players and recorders, PDAs, smart phones, and voice recorders. 20:06:37 ais523: so is altera verilog or vhdl? 20:06:47 ehird: almost certainly both 20:06:56 but the general rule is that the military uses vhdl, everyone else uses verilog 20:07:04 I thought you used vhdl 20:07:10 they're pretty much the same language nowadays anyway, just with different syntax 20:07:12 and yes, I was tought VHDL 20:07:14 for some reason 20:07:19 *taught 20:07:24 Verilog's syntax looks nicer 20:07:25 I suspect most talented FPGA people end up in the military... 20:07:37 ehird: Verilog is C-like, VHDL is ADA-like 20:07:41 ais523: figured out what 20K could fit? 20:07:53 not yet, I'm still looking 20:07:59 "However, using this 9-valued logic (U,X,0,1,Z,W,H,L,-) " 20:08:02 9 valued logic. 20:08:03 yes 20:08:04 Fuck VHDL 20:08:08 booleans have 9 possible values in VHDL 20:08:13 Yes :p 20:09:02 ais523: Verilog isn't that c like 20:09:05 ais523: ... 9 possible values?!? 20:09:07 if (rst) // This causes reset of the cntr 20:09:07 count <= 5'b0; 20:09:08 else 20:09:10 if (cet && cep) // Enables both true 20:09:12 begin 20:09:14 if (count == length-1) 20:09:16 count <= 5'b0; 20:09:17 HOW DOES THAT MAKE SENSE? 20:09:18 else 20:09:20 count <= count + 5'b1; // 5'b1 is 5 bits 20:09:22 end // wide and equal 20:09:24 pikhq: http://en.wikipedia.org/wiki/IEEE_1164 20:09:26 impedance, drive and shit 20:09:29 Because it's hardware. 20:09:40 Fuck hardware. 20:09:55 pikhq: true, false, true via a resistor, false via a resistor, contradiction, uninitialised, between true and false, unconnected, unknown 20:09:56 ** pikhq has disconnected 20:09:56 I'll train a bunch of rodents to do my computations. 20:10:15 http://www88.wolframalpha.com/input/?i=bacon+as+mass+of+milky+way&a=*C.bacon-_*ExpandedFood.dflt-&a=*C.milky+way-_*Astronomical.dflt-&a=*EAC.ExpandedFood.PreparedCuredPorkBacon-_**PreparedCuredPorkBacon.*Food%3ACookingMethod_Food%3APanFried--- 20:10:18 in RL, hardware can end up with a lot more values from 0 to 1 20:10:19 Bacon as mass of milky way 20:10:23 1 slace 20:10:25 *slice 20:10:27 Pan-fried 20:10:28 *than 0 and 1 20:10:42 "Assuming "milky way" is an astronomical object | Use as a food instead " 20:10:44 * ehird eats the milky way 20:10:52 ehird: I use IP via carrier pigeon. 20:10:59 (yes, I know it's a chocolate bar) 20:11:04 And I feed my pigeons coffee beans. 20:11:11 pikhq: true, false, true via a resistor, false via a resistor, contradiction, uninitialised, between true and false, unconnected, unknown <-- why is "via resistor" significant 20:11:17 pikhq: that's hardware, just wetware 20:11:32 ehird: Well, yes. Wetware, however, makes much more sense. 20:11:38 pikhq: Are you on crack? :-) 20:11:41 AnMaster: because if you connect true via resistor to false, you get false 20:11:48 but if you connect true to false, you get a contradiction 20:11:50 No, but my computer rats are. 20:11:57 ais523, I see... 20:12:01 contradictions are /bad/, they can damage the chip in theory 20:12:04 * pikhq picks up a bit 20:12:31 ais523, how do you model analogue cicuits 20:12:43 * pikhq makes sure the bit is in operating condition, gives it more crack, sets it back into RAM 20:12:53 ehird: OK, I found some details: you can do an M68000 microprocessor in about 6000 LUTs 20:12:58 depending on what other features of the chip you use 20:13:00 AnMaster: With a breadboard. 20:13:04 hm 20:13:06 ais523: that's pretty good 20:13:14 would be nicer if you could do it more like VHDL 20:13:14 that's very good, in fact 20:13:15 or such 20:13:21 AnMaster: using a language other than VHDL 20:13:27 ais523, such as? 20:13:29 ais523: i could have over 20 68000s :P 20:13:38 I'm not sure if there is one yet 20:13:46 ais523, ah... 20:13:46 ehird: Cluster! 20:13:52 ehird: also, buy yourself an anti-static wrist strap, if you're planning to mess with FPGAs 20:14:00 they're very static-sensitive, much more so than computers 20:14:00 ais523: Er. 20:14:04 Okay. :P 20:14:24 ais523: do you actually touch it much? 20:14:41 I wouldn't expect so 20:15:12 ehird: depends on what you're doing with it 20:15:24 but mostly you wouldn't, I expect 20:15:43 I'm paranoid about static, though, I've seen all sorts of good circuits ruined due to it 20:16:33 what about making static-resistant/proof chipset 20:16:35 chipsets* 20:16:38 and so on 20:16:42 would be interestig 20:16:45 interesting* 20:16:59 possible? 20:17:18 Possible, if you've got a chip fab. 20:17:20 you mean a computer? 20:17:31 pikhq, huh? 20:17:36 AnMaster: could be, but the technique of using a static-sensitive chip and putting it in an antistatic box tens to be cheaper 20:17:39 *tends 20:17:52 ehird, no I mean FPGAs, ICs and so on 20:18:00 AnMaster: not what i meant 20:18:13 ehird, but what I meant! 20:18:17 kay 20:18:25 AnMaster: it depends on the internal logic method 20:18:33 TTL-based chips are a lot less static-sensitive than CMOS, for instance 20:18:34 hm 20:18:41 but slower right? 20:18:47 not directly 20:18:53 oh? 20:18:53 but they use more power 20:18:59 more power = more heat = harder to cool 20:19:04 ais523, what method is used for current CPUs btw? 20:19:05 so you have to run TTL-based circuits slower or they overheat 20:19:12 and CMOS is pretty much standard for anything complicated 20:19:17 kay 20:19:22 simpler things are often lowpower-schottky-TTL 20:19:36 ais523, "shottky"? 20:19:37 if for some reason you don't want CMOS (say you're the military and want something that won't break in weird conditions) 20:19:54 AnMaster: it's some slightly different way to make a transistor, I think 20:20:00 ah 20:20:05 ais523: Or if you're putting the thing in orbit. 20:20:09 pikhq: yes 20:20:12 AnMaster: CMOS is used atm, but we're nearing on nanotechnology 20:20:12 thought it was a typo for "shoddy" 20:20:20 I imagine they don't use CMOS much on spacecraft 20:20:22 although I don't know 20:20:42 22nm is the last pure CMOS, iirc 20:20:47 and 11nm is real nanoelectronics; I may be wrong 20:20:53 The space shuttle still has solid-state circuitry. 20:21:10 ais523, did you mean "Schottky"? 20:21:12 Might even have some tubes. 20:21:14 err yeah 20:21:16 I typoed 20:21:18 not you 20:21:20 meh 20:21:20 AnMaster: I said "Schottky", then you misqupted me 20:21:22 *misquoted 20:21:24 "Fifteen nanometres is equivalent to 0.005906 thou, or 0.000015 millimetres, and is the width of a very very very very very thin human hair which you can't see unless your eyesight is very very very very very good." —The Inquirer 20:21:26 ais523, ah 20:21:28 indeed 20:21:32 ehird: haha 20:21:50 there are a lot of nanometres in a deci-inch, surely 20:21:58 ehird, that is a joke right? 20:22:07 ais523: it's very very very very very thin hair. 20:22:24 AnMaster: nope, 100% serious 20:22:28 (deci-inches are /not/ the industry standard for pin spaces, that's the 2.54mm standard, but the name is inevitable given that it is exactly a tenth of an inch) 20:22:34 Wow. The space shuttle's RAM... 20:22:35 the very-hair-very-eyesight measurement system is very very very common 20:22:38 ehird, you don't get human hair that thin 20:22:41 pikhq, what about it 20:22:43 *Magnetic core*. 20:22:43 AnMaster: prove it 20:22:48 :D 20:22:51 pikhq: that's brilliant 20:22:52 pikhq, WOW 20:22:54 and actually, doesn't surprise me 20:23:03 given that they must actually get loads of cosmic rays up there 20:23:03 that's ho 20:23:03 t 20:23:05 ehird, too lazy 20:23:18 I imagine the transistors in the processors are massively big by today's standards, too 20:23:18 http://en.wikipedia.org/wiki/IBM_AP-101 They've got 5 of these. 20:23:23 AnMaster: "Analyze every human hair on the planet" "too lazy" Understatement of the century 20:23:23 probably measured in microns 20:23:48 ais523: eh, 1 um was circa 1985 20:24:01 and we had 0.X ums until 2002-2003 20:24:05 w/ 90nm 20:24:06 It's a System 360. 20:24:08 (before that 0.13 um) 20:24:10 (basically) 20:24:18 ehird: big in order to avoid radiation disruption, rather than because they can't make them smaller 20:24:20 So, fucking huge. 20:24:23 ais523: I know 20:24:27 I'm just saying it's not THAT old 20:24:33 yes 20:25:07 also, µm not um 20:25:58 ais523: lazy :P 20:26:15 D'aw... 20:26:20 They upgraded the thing in 1990. 20:26:27 They use an AP-101s. 20:26:31 Has semiconductor memory. 20:26:50 ehird: it's AltGr-m 20:26:55 And a whole 1.2 MIPS. 20:28:24 ehird, you didn't say that. You said "prove it". Which could be referring to some report 20:28:29 ask Alpha about it! 20:28:35 AnMaster: such a report would have to review all human hair too 20:30:23 ehird, nah. Just prove it isn't possible. 20:30:38 how? 20:30:50 "On Monday, I start at Google. I don't know the exact details of what I'll be doing there, but I can only assume that the bright folks there have realized that the future of massively powerful data centers belongs to wire-wrapped TTL. I'll be sure to bring my wire-wrap gun." —homebrewcpu.com 20:31:44 He works on Android. 20:31:52 Still, :D. 20:32:03 yeah 20:32:05 on the java vm 20:32:45 [[Shortly after I declared Magic-1 "hardware complete", I casually mentioned to my wife that I was starting to think about Magic-2. Her response was swift, and final: 20:32:45 20:32:47 "No, there will be no Magic-2!" 20:32:49 o 20:32:49 20:32:51 I can't blame her. She was an extraordinary good sport during Magic-1's design and construction - especially during the wire-wrapping phase. For most of a year, she put up with electronic junk littering the kitchen table, wire-wrap insulation fragments on the floor and a husband often lost in concentration while the kids were hollering for attention. 20:32:56 20:32:58 She's the love of my life, the woman I plan on growing old with, mother of my children, my partner and best friend. I have to respect her wishes on this. 20:33:01 20:33:03 wire-wrapping? 20:33:03 So, there will be no Magic-2. 20:33:05 20:33:07 Instead, we'll call the follow-on project "Magic-16".]] 20:33:12 ais523: TTL thing 20:33:18 I know what wire-wrapping is 20:33:20 oh 20:33:24 just I'm slightly surprised it was being used 20:33:28 ais523: Magic-1 is a wire-wrapped TTL machine. 20:33:30 ais523: it IS a homebrew cpu... 20:33:33 :) 20:33:35 and avoids soldering 20:33:37 yes, but I would have soldered 20:33:45 ais523: have you seen it? 20:33:46 because it's easier and more reliable 20:33:47 it's quite big 20:33:52 I wouldn't say it's easier 20:33:54 soldering is a pain 20:33:56 and no, I haven't seen it 20:33:59 He wanted it to be hard-core. 20:34:12 and especially with the typical scale of TTL chips, soldering is easy 20:34:20 they're some of the easiest things to solder 20:34:26 ais523: http://www.homebrewcpu.com/Pictures/bu_3.JPG 20:34:35 annoying soldering is if you have something in a SOIC package 20:34:42 I once spent several hours soldering one of those 20:34:50 http://www.homebrewcpu.com/Pictures/P5140113.JPG ← it has a window, just like HARDKOR PCS! 20:34:53 with a magnifying glass and a special soldering iron 20:34:54 it's even blue 20:35:03 ais523: wire-wrapping is more beautiful, though 20:35:09 http://upload.wikimedia.org/wikipedia/commons/d/d1/Computerplatine_Wire-wrap_backplane_detail_Z80_Doppel-Europa-Format_1977.jpg 20:35:29 ugh no, that's just illegible 20:35:39 ais523: but pretty 20:35:46 and could be soldered in about half a second per pin for someone who knew what they were doing 20:35:54 which is faster than the wire-wrap would be 20:35:58 ais523: and who didn't care about burning themselves :) 20:36:16 ehird: with the apparent scale of that, burning oneself would be really rather unlikely 20:36:27 your hands would be nowhere near the soldering itself 20:36:41 meh. 20:36:42 Part of the idea was to be retro. 20:36:44 soldering burns are more likely when putting the soldering iron away afterwards 20:36:47 (I know this from experience) 20:36:49 And wire-wrapping is very retro. 20:36:50 hm 20:36:57 ais523: You fail. :p 20:37:00 why wire wrapping 20:37:06 ... Granted, I've done worse. 20:37:10 pikhq: no, it's nearer your hand then than at any time during the actual soldering 20:37:17 I've burnt myself about three times like that 20:37:19 (I once grabbed a soldering iron by the tip by accident) 20:37:30 ouch 20:37:43 Yeah, I got a pretty solid burn from that. 20:37:50 um 20:37:51 running your hand under the cold tap for a few minutes normally means you have no lasting effects 20:37:57 I never burned on my soldering iron 20:38:02 :) 20:38:03 pikhq: maybe you should get one of the solder gun things 20:38:04 anyway 20:38:09 how does wirewrapping work 20:38:11 which heats up in about 3 seconds, and cools down just as quickly 20:38:13 why does it stay attached 20:38:19 AnMaster: friction 20:38:23 I once touched a hot, bright lightbulb with the my hand directly. 20:38:25 ais523, err, the end? 20:38:27 In a sort of grasping motion. 20:38:36 That was not actually very damaging. 20:38:37 But damn it hurt. 20:38:38 AnMaster: you wrap the wire around a lot, tightly 20:38:52 ais523, trapping the end under the wraps? 20:39:04 no, you wrap the end of the wire around the pin 20:39:09 if you wrap tightly enough, it stays on 20:39:16 ais523, that makes no sense 20:39:20 AnMaster: The process of wire-wrapping creates a cold weld. 20:39:28 pikhq, uhu.... 20:39:34 pikhq, how does that work 20:40:08 ais523: "Wire wrap construction can produce assemblies which are more reliable than printed circuits" 20:40:12 Bringing two pieces of metal in contact with each other at enough pressure welds them together. 20:40:21 ehird: yes, I've seen some /really/ low quality printed circuits in the past 20:40:29 pikhq, interesting 20:40:34 -!- jix has quit (No route to host). 20:40:45 soldering is more reliable than wire-wrapping; but the actual printed wires on PCBs can often be awful 20:40:49 so the connections are fine, but the wires break 20:40:59 ais523, ah yes, seen that too 20:41:06 you can use stripboard if you want the reliability of wires and of soldering 20:41:07 Works only with metals that don't oxidize, IIRC. 20:41:18 but that doesn't work on high-frequency circuits 20:41:34 pikhq: most wires in use nowadays are tin-plated, though, even though they're made of copper 20:41:34 (or in a vaccum) 20:41:39 that's why they don't look red like copper does 20:42:51 * pikhq observes that wire-wrapping is kinda cool, if absurdly labor-intensive. 20:43:10 Unless you've got a handheld wire-wrapping tool or an automatic wire-wrapper. 20:43:21 (the latter is about as cheap as a chip fab) 20:43:38 i want one of those loom things 20:43:41 and emulate x86 on them 20:43:50 it'd totally be possible 20:43:54 pikhq: but solder wave baths are cool too 20:44:03 ehird: but x86 is an awful architecture 20:44:07 ais523: Indeed, they are. 20:44:10 ais523: thus making the emulation more impressive 20:44:14 run X on a loom! 20:44:20 That's just painful. 20:44:31 Now, I bet that Magic-1 could run X. 20:44:37 Not well, but it could run. 20:44:42 (assuming video hardware) 20:44:45 4mhz, Minix... 20:44:45 rather slowly, I'd imagine 20:44:49 It could run X pretty nicely. 20:44:51 4mhz isn't too bad... 20:44:52 ais523, are there good quality printed circuit boards? 20:44:56 You could run a few xterms. 20:44:57 AnMaster: yes 20:44:58 Hmm. 20:45:00 motherboards, for one 20:45:05 So why not? 20:45:13 It wouldn't be fancy stuff like we're used to, but it'd at least run X. 20:45:20 ais523, right. But none of those you actually end up soldering on yourself! 20:45:22 what's really disappointing, though, is that at University, we couldn't make our own PCBs for safety reasons 20:45:22 Not much more than Xterms and TWM, I'd imagine. 20:45:28 so we had to pay for the technicians to make them 20:45:34 ais523: ... 20:45:37 and they came out lower-quality than the ones I'd made alone at secondary school 20:45:40 That's really lame. 20:45:49 ais523, is it hard to make PCBs? 20:45:50 pikhq: very, considering I was approved to make them when at school 20:45:51 All you really need to make a PCB is a dark room. 20:45:52 and not at university 20:45:59 AnMaster: no, you just need the appropriate equipment 20:46:05 ais523, how is it unsafe then 20:46:10 pikhq: xterms and twm should be enough for anyone 20:46:13 AnMaster: it involves chemicals 20:46:23 AnMaster: The best way to do it at home is a photographic process. 20:46:25 ais523: oxygen is a chemical!111 20:46:26 some of which are moderately dangerous 20:46:26 ais523, err... So what about chemistry classes at uni 20:46:32 pikhq: you don't need a dark room 20:46:37 vs. those at school 20:46:38 I agree with the photographic process 20:46:39 ais523: It helps. 20:46:42 but it's done with ultraviolet 20:46:53 so as long as you block out natural light and are fast, you don't care about visible light around 20:47:03 ais523: that's how you make printed circuit boards? 20:47:04 ais523: The thing is, a dark room has all the equipment you need. 20:47:08 like developing a photograph with ultraviolet? 20:47:09 :-D 20:47:22 that's *awesome* 20:47:23 ehird: yes 20:47:24 I didn't say it needs to actually be dark. 20:47:37 pikhq: ah, ok 20:47:38 that's ultra mega awesome 20:47:44 ehird: Enlarger, wash basin, and some chemicals. :) 20:47:52 what do you actually develop, though? 20:47:56 not the result 20:47:56 the input 20:48:00 ehird: basically, you have PCBs coated with copper, where the copper is itself coated with photoresist 20:48:08 you print out your design on tracing paper 20:48:20 then shine ultraviolet through the tracing paper onto the photoresist 20:48:29 what does the design look like? 20:48:37 ehird: the shape of the copper you need 20:48:39 It's a negative. 20:48:45 haha that's so awesome. 20:48:46 ink where the copper is, blank where it isn't 20:48:57 so it's arguably a positive, the whole process is a double-negative 20:48:59 so this is how they make motherboards? 20:49:06 for motherboards, it's automated 20:49:18 and the boards have multiple layers, so it's a bit more complicated 20:49:19 ais523: Oh, didn't realise that it was positive-positive printing. 20:49:51 pikhq: black on the input = no ultraviolet = fixed photoresist = the acid can't get to the copper 20:49:57 so it's sort-of double-negative printing 20:50:01 I wasn't thinking. 20:50:25 That's positive printing. ;) 20:50:58 anyway, once you've shined the ultraviolet onto the photoresist, you take the whole thing over to a vat of developer 20:51:01 and submerge it in there 20:51:07 I wonder what the ideal instruction set for C is 20:51:20 ehird: PDP-11. 20:51:20 the photoresist is naturally green, but it turns purple in the developer and washes off in a sort of wispy fashion, it's rather pretty 20:51:24 pikhq: not really 20:51:29 then you dip the whole thing in a vat of acid 20:51:36 and that's the bit that people keep screwing up 20:51:38 -!- Asztal_ has joined. 20:51:48 ais523: I always thought motherboards were just like, squirting and denting silicon and copper and shit with machines. 20:51:48 But you can run the original compiler on a PDP-11. 20:51:53 Photographic motherboards is so cooler. 20:51:58 if you don't leave it in there long enough, the acid doesn't have enough time to etch away the copper 20:52:00 and you get open circuits 20:52:07 -!- jix has joined. 20:52:10 *short circuits 20:52:20 ais523: That's kinda lame. 20:52:21 if you leave it in there too long, the acid gets in under the photoresist and the wires become unreliable 20:52:37 -!- Asztal_ has quit (Client Quit). 20:52:42 Especially since the etching isn't that much harder than a photographic fixer... 20:52:48 no, it isn't 20:52:49 wow, Minix doesn't do copy-on-write when fork()ing 20:52:50 how rubbish 20:52:57 which is why it's beyond me that people keep screwing it up 20:52:57 Granted, the etch bath is a bit more caustic than a fixer. 20:53:02 I am, as you guessed, rather bitter about this 20:53:09 Don't blame you. 20:53:31 (also, a 45-degree angle or less on a PCB is known as an "acid trap", because the acid tends to get stuck in it even after you've taken the PCB out of the acid) 20:53:33 I mean, really. It's a process that's well over a hundred years old by now. 20:53:40 My *grandmother* could do it better. 20:53:57 (really; she has dark room equipment laying around) 20:54:00 pikhq: what method do you use in order to get the spare photoresist off afterwards? 20:54:19 when I used to do it, we did it by covering the whole board with ultraviolet and developing a second time 20:54:23 * Corun notes that tanenbaum is giving a lecture as his uni a week on thursday 20:54:27 ais523: For a circuit board? 20:54:32 I think that's the best method. 20:54:37 pikhq: yes, it works because there are no components 20:54:41 and agreed 20:54:48 although there are at least two methods which work at that point 20:54:50 Might be a wash bath instead, but *shrug* 20:55:10 I've seen a catalogue that recommended you used a PCB eraser to clean the lines of photoresist 20:55:17 Corun: That's pretty nice. I recommend going, he's a rather good speaker. 20:55:18 and a PCB eraser works much the same way to a pencil eraser 20:55:30 Yeah, I'll go 20:55:41 Tell him about all my next gen OS ideas :-) 20:55:41 I wonder if tanenbaum still considers linux obsolete 20:55:44 Or maybe I shouldn't... 20:55:52 * pikhq saw his address at USENIX ATC '08, after he got the USENIX Lifetime Achievement Award. Good lecture. 20:55:55 I mean he could just steal them ^_^ 20:56:07 pikhq, who? 20:56:16 ... 20:56:17 AnMaster: Tanenbaum. 20:56:18 Tanenbaum. 20:56:22 * AnMaster googles 20:56:26 * ehird boggles. 20:56:30 Corun: Feel free to tell him. 20:56:30 You DON'T KNOW WHO TANENBAUM IS? 20:56:34 ehird, isn't it German for xmas tree... 20:56:34 Stop using a computer right now. 20:56:35 iirc 20:56:44 AnMaster: Less talking, more shutting off. 20:56:48 Minix inventor, and author of like a million OS books and basically he's professor of all Operating Systems. 20:56:48 He likes talking about kernel designs. 20:57:00 AnMaster: He wrote the book on kernels. 20:57:00 ehird, did ais523 know... 20:57:09 AnMaster: I'm pretty fscking sure ais523 knows who Tanenbaum is... 20:57:15 ehird, considering he didn't know who jwz was... 20:57:21 He knew who jwz was. 20:57:26 ehird, no he didn't 20:57:27 He just didn't know he worked on Netscape. 20:57:27 AnMaster: according to lingbot, "Tanenbaum" translated from German to English is "apertium" 20:57:39 ais523: you knew who jwz was, didn't you? 20:57:41 but as far as I know, "apertium" isn't an English world 20:57:43 ais523, what is "apertium"? 20:57:43 just not that he worked on netscape 20:57:45 ah 20:58:10 ehird: my situation with Tanenbaum is similar; I know he's famous in computers/programming, but can't remember what for 20:58:18 ais523: flaming Linux 20:58:21 is the most famous thing 20:58:22 but also, Minix 20:58:29 and a gigantic amount of operating system textbooks 20:58:38 ehird, so don't complain about me then 20:58:41 but most famous popculture wise, definitely flaming Linux way back then 20:58:47 AnMaster: at least he had a vague idea 20:59:00 ehird, the name did sound familiar 20:59:05 ah, Minix, that was it 20:59:08 why did he flame Linux? 20:59:15 ais523: monolithic kernel 20:59:17 for doing things differently? 20:59:20 Cos he considered it a step backwards in operating system design 20:59:23 yep, I remember now 20:59:40 ais523: http://oreilly.com/catalog/opensources/book/appa.html 20:59:41 and the funny thing is, with modules, Linux managed to work around many of the disadvantages of being monolithic 20:59:49 the full debate 21:00:01 although it was restarted in 2006, it seems 21:00:03 ehird, why did he flame linux btw 21:00:06 oh wait 21:00:10 I'm late asking 21:00:12 AnMaster: Dude! What is UP with your SCROLLBACK? 21:00:22 You never seem to use it... 21:00:23 ais523: Linux is an example of good monolithic design. 21:00:29 ehird, I was searching on wikipedia for it 21:00:33 ehird, didn't find it 21:00:36 ehird, asked here 21:00:37 I suspect a very sophisticated microkernel could be better, but *shrug*. 21:00:41 oh, apparently, according to lingbot, "Tanenbaum" in German translates to "apertium" in English, Swedish, /and/ French 21:00:44 saw that ais asked above 21:00:51 pikhq, QNX style? 21:00:55 pikhq: Microkernels are evil. 21:00:58 I'm just happy using a well-designed kernel. 21:01:01 So's monolithic kernels. 21:01:02 ais523, that is broken 21:01:03 So's kernels. 21:01:13 AnMaster: Minix style, bitch. ;) 21:01:30 now I'm curious as to what "apertium" means, it looks vaguely Latin 21:01:32 http://tunes.org/cliki/microkernel_20debate.html 21:01:36 ais523, um I suspect a bug: http://en.wikipedia.org/wiki/Apertium 21:01:43 it is a translation software it seems 21:01:45 & http://tunes.org/cliki/kernel.html 21:01:48 ais523, :D 21:01:50 & http://tunes.org/cliki/no-kernel.html 21:01:54 [21:01] lingbot: en de "test"? 21:01:55 [21:01] ais523: "Test" (en to de, apertium) 21:02:00 I think it's a bug too 21:02:09 ais523, it is using that software, but it is buggy? 21:02:52 Translation: German » English 21:02:52 Tanenbaum 21:02:52 Tanenbaum 21:02:55 could be 21:02:59 that is Google translate 21:04:05 ehird: according to that page you linked, NT's a microkernel 21:04:06 that surprises me 21:04:09 no 21:04:15 according to it, NT was originally designed as a microkernel 21:04:21 but grew into a monolithic one 21:04:30 and then it says that was disputed by an NT architect 21:04:37 ah, ok 21:05:09 heh, Tanenbaum seems to hate x86 just as much as everyone else 21:06:02 But in all honesty, I would 21:06:03 suggest that people who want a **MODERN** "free" OS look around for a 21:06:03 microkernel-based, portable OS, like maybe GNU or something like that. 21:06:09 I love how GNU was a reasonable OS proposition for the future in 1992 21:06:13 with Hurd! 21:06:44 maybe it would have been, if more people worked on it 21:06:55 it'd have been terribly bloated 21:07:00 I wonder if there's anything intrinsically wrong with Hurd, or just if it never caught on 21:07:03 but then... GNU design 21:07:07 ah yes, it would never have caught on 21:07:42 ais523, why not? 21:07:44 GNU design? 21:07:52 ais523: have you read the HURD design? 21:07:55 it's truly crazy 21:08:04 ehird: bits of it, but I think my brain shut off 21:08:15 AnMaster: GNU programs tend to get bloated over time 21:08:16 it involves kernels being kernels among other kernels while a centralized kernel manages them in a kernel way 21:08:19 sort-of like Firefox 21:08:26 ais523, true 21:08:26 and then you get injected with LSD 21:08:31 which is strange, as other open-source software doesn't 21:08:36 (part of the bootup phrase, I thikn) 21:08:36 well, not all of it 21:08:37 *think 21:09:00 ehird, exokernels? 21:09:02 or what 21:09:07 AnMaster: I don't even know 21:09:13 ehird, or Xen style? 21:09:19 I don't know 21:09:22 kay 21:09:27 It's not anything I've seen before 21:09:29 It's crazy 21:09:56 heh, it has a server with the express purpose of handling crashes 21:10:00 the crash server 21:10:35 $ ./esotope-bfc 21:10:35 /usr/bin/env: python -O: No such file or directory 21:10:36 fail 21:10:43 #!/usr/bin/env python -O 21:10:46 that isn't valid 21:10:47 AnMaster: that's not fail 21:10:52 ehird, it is 21:11:02 AnMaster: it's your system having an arbitrary limitation 21:11:04 ehird, iirc you can only give one parameter on that line 21:11:05 ... 21:11:06 nope 21:11:09 yep 21:11:12 only one is guaranteed, I think 21:11:13 what does POSIX say? 21:11:17 ehird, there you are then 21:11:19 but the only reason more wouldn't work is due to artificial limitation 21:11:24 AnMaster: _guaranteed_ 21:11:26 more isn't invalid 21:11:28 lifthrasiir, it is a bug. 21:11:34 ... no it's not 21:11:44 AnMaster: fail, /usr/bin/env is more than 8 letters long 21:11:50 ooh, burn 21:11:52 ais523, haha 21:11:55 I'm not sure if that's an artificial POSIX limitation 21:11:58 but it ought to be 21:12:03 ais523, oh 21:12:07 POSIX has no limitation on it 21:12:12 it just states that you're only guaranteed one argument 21:12:17 if you can't handle more, your kernel sucks 21:12:30 ehird, going to stay on OS X then 21:12:31 I see 21:12:44 moving to a PC != the linux kernel is good 21:12:47 ah, it seems that the limit's 32 21:13:03 as in, some old shells will cut off #! lines after 32 characters 21:13:10 man perlrun warns about that 21:13:17 in case your command-line options get cut in half 21:13:22 (Perl parses #! lines itself) 21:13:22 Another way that some historical implementations handle shell scripts is by recognizing the first 21:13:22 two bytes of the file as the character string "#!" and using the remainder of the first line of the 21:13:22 file as the name of the command interpreter to execute. 21:13:27 quoting POSIX 2008 21:13:37 about exec() 21:13:38 AnMaster: if that's true, then /usr/bin/env is invalid 21:13:43 as it'd mean you can't specify any arguments 21:13:45 so sir, the fail is on you 21:14:00 incidentally, would just #!python work? 21:14:20 ais523: nope 21:14:24 as in, does it do a path search for the executable you ask, or does it just go to the place you ask/ 21:14:27 ah, pity 21:14:29 [ehird:~/Junk] % cat>a 21:14:29 #!python 21:14:30 print "hi" 21:14:33 [ehird:~/Junk] % chmod +x a 21:14:34 [ehird:~/Junk] % ./a 21:14:36 zsh: ./a: bad interpreter: python: no such file or directory 21:14:55 anyway 21:15:07 lifthrasiir, the wrapper script is broken on Linux. 21:15:11 lifthrasiir, needs to be fixed 21:15:26 AnMaster: can I point out some hypocrisy here? 21:15:40 you refuse to support systems that don't support POSIX properly. why should lifthrasiir make up for a deficiency in your kernel? 21:16:01 ehird, I support systems that support what POSIX + XSI requires 21:16:10 and this is more than what POSIX + XSI requires 21:16:19 he supports systems that don't have artificial limitations on #! arguments 21:16:43 whereas with C-INTERCAL, I aim to eventually support anything which has a program vaguely resembling a C compiler 21:16:47 although I'm not there yet 21:16:57 From XSH: 21:16:59 1. The shell reads its input from a file (see sh), from the −c option or from the system( ) and 21:16:59 popen( ) functions defined in the System Interfaces volume of POSIX.1-2008. If the first 21:16:59 line of a file of shell commands starts with the characters "#!", the results are 21:16:59 unspecified. 21:17:03 xsh? 21:17:12 ais523, the section on the shell 21:17:14 and tools 21:17:24 AnMaster: ok, so depending on #!-lines is unportable 21:17:29 ehird, yes! 21:17:29 AnMaster: the german spelling is "Tannenbaum" 21:17:32 you're digging yourself into a hole here, AnMaster 21:17:38 ehird, no I'm not 21:17:42 this is exactly why strict POSIX compliance is bullshit 21:17:44 ehird: AnMaster: wait, that's about calling a shell with a file starting #! as its argument 21:17:48 ehird, you fail 21:17:52 that's different from running a program starting #! 21:17:57 ais523: are you sure? 21:17:59 unless the #! line specifies a shell 21:17:59 ./a executes the shell on a 21:18:12 ehird: no, because what if it's a binary? 21:18:14 ./a /runs/ a 21:18:26 ais523: hmr 21:18:31 err 21:18:32 not XSH 21:18:34 XCU 21:18:37 obviously 21:18:41 XSH is system interfaces 21:18:45 AnMaster: I fail to see why that's obvious 21:19:04 ais523, because XSH is system interfaces, the C API. 21:19:08 all the headers 21:19:09 and so on 21:19:17 SH means system interfaces 21:19:19 of course 21:19:20 yes, but I don't imagine that that's common knowledge 21:19:21 ais523, POSIX comes in 4 volumes... 21:19:21 XCU means shell 21:19:23 of course 21:19:28 XBD XSH XCU XRAT 21:19:32 so something that depends on it is not "obvious" 21:19:51 ehird, XCU means "Shell and Utilities" 21:19:57 and yeah it is confusing 21:19:59 C stands for Shell 21:19:59 that's like saying that #1 $ #1 equals #3, obviously 21:20:00 obviously 21:20:03 which is why I mixed them up 21:20:07 Chell 21:20:49 XBD Is Base Defintion. Says stuff like "mount point" being an implementation defined concept 21:20:51 and what not 21:21:14 bbl food 21:21:35 ais523: We should make a DS9K implementation of POSIX 21:21:41 just to trample on everyone touting POSIX compliance 21:21:42 ehird: see Windows 21:21:45 I'd do it on my own, but you're eviller 21:21:50 ais523: that's more a DS9King of the tests 21:21:54 not the OS 21:21:56 well, yes 21:22:16 something that'd pass a stock POSIX test suite run by an unbiased tester, but still be completely broken on more or less anything 21:22:17 is what I meant 21:22:22 but you mean, an implementation that's more interesting than ENOTIMPLEMENTED as a return value for everything? 21:22:25 yes 21:25:56 ais523: would it be possible to implement posix in user-mode? 21:25:59 completely compliant 21:26:13 ehird: I don't see why not, so long as you can implement devices somehow 21:26:33 ais523: just map them to the host devices 21:26:37 wait, which devices 21:26:37 /dev? 21:26:42 or physical I/O devices 21:26:49 well, I suppose /dev 21:26:52 you'd have your own in-memory FS, ofc 21:26:54 from the point of view of the testsuite 21:28:03 BA-K-47: America's No. 1 bacon-based assault rifle 21:28:04 and now it's hit Slashdot: http://tech.slashdot.org/article.pl?sid=09/05/19/1846258&from=rss 21:28:05 http://latimesblogs.latimes.com/dailydish/2009/05/bak47-the-gun-that-will-revolt-and-defeat-terrorists.html 21:28:13 I have no idea if I'm responsible for all this or not... 21:28:22 ais523: ask? 21:28:52 can't be bothered, and don't have a good way to ask 21:28:56 comment 21:29:07 -!- Gracenotes has quit ("Leaving"). 21:31:30 ais523: a kernel that can run in user-mode would be interesting 21:31:41 like, it requires a basis of functions that either map to the host OS or the hardware 21:31:46 and the rest of the kernel is built on top of them 21:31:56 so you could, say, do a VPS or a jail really cheaply 21:32:01 ehird: that's basically how gcc-bf's fake operating system works 21:32:14 right, but it could be both usermode and hardware 21:32:36 I love the way I implemented a filesystem 21:32:50 it's just a dictionary of file against file contents, implemented using linear search 21:32:56 possibly the world's worst filesystem 21:33:23 ais523: oh, with user-mode posix, you could run a standard filesystem on a real harddisk 21:33:29 ehird: So, something like UML? 21:33:36 although you'd probably want to do it in a big file, to avoid the whole root thing 21:33:38 -!- RodgerTheGreat has joined. 21:33:39 and for convenience 21:33:43 howdy folks 21:33:47 * pikhq waveth at the Rodger 21:33:49 pikhq: it just emulates the hardware for linux, doesn't it? 21:33:55 not actually run a kernel properly tuned to usermodeity 21:34:14 yeah, I definitely need to start coming here again 21:34:22 ehird: It's not very well-tuned, but it does implement everything in terms of Linux system calls. 21:34:26 but you mean, an implementation that's more interesting than ENOTIMPLEMENTED as a return value for everything? <-- isn't allowed for most stuff in 2008 21:34:28 iirc 21:34:32 pikhq: eh, even so 21:34:39 for example in POSX 2008 mmap() is required 21:34:44 And it's been ported to Cygwin before. I'd imagine it wouldn't be hard to abstract it further. 21:34:45 while in 2001 it is optional 21:35:30 Hmm. A usermode POSIX kernel would be pretty cool to implement, actually. 21:35:34 RodgerTheGreat, have you been away? 21:35:42 I'm very much enjoying the public reaction to Alpha- "This seems to be pretty useless. It cannot tell me how many cigarettes would fit between the earth and the sun, the nutritional value of Lindsay Lohan, or how big Shaq would have to be to dunk the moon." 21:35:52 AnMaster: I used to hang out here pretty regularly 21:35:55 AnMaster: Just not in this channel. 21:35:56 AnMaster: mmap doesn't force you to map anywhere in particular in memory, though, does it? 21:36:09 RodgerTheGreat, I know 21:36:20 RodgerTheGreat: It also can't plot the distance to jupiter over time, although it can give you the distance to jupiter in a given year. Don't make excuses for its shittiness. 21:36:31 And yes, I worded it every which way. 21:36:31 RodgerTheGreat, I just never noticed you left. 21:36:34 so did GregorR 21:36:34 ais523, huh? 21:36:36 also, children can't legally use it, or so the terms of service claim 21:36:49 ais523, I leave that to the OS. by passing a NULL pointer 21:36:51 AnMaster: so couldn't you just allocate a bitmap and copy the file into it, and then copy the bitmap back into the file when it closed? 21:36:52 yeah, I'm mostly on irc.esper.net these days, so I just haven't bothered to connect to freenode as well 21:36:56 to implement mmap? 21:37:14 ais523, err. Sure why not. 21:37:25 it would be inefficient; but no reason why you can't implement mmap in userspace 21:37:31 ais523, if you can handle the updates on writes from other processes correctly 21:37:34 attempts to map into a particular location would presumably always be rejected 21:37:43 AnMaster: you could commit the file at every context switch 21:37:51 that would be /really/ inefficient, but correct I think 21:37:53 ais523, sure. If you want to 21:38:07 it would make mmap rather unusuable, but if this is meant to be a DS9K, why not? 21:38:13 oh hey- I don't think I ever showed this in here- 21:38:21 http://je.zacbrown.org/code/neurogen.ps 21:38:23 ais523, but I fail to see how it would break any app relying on common behaviour. Would just make them slow. 21:38:28 ais523: it seems this has split into two; DS9Ksix and Usersix 21:38:36 a neural-network backpropagation trainer in postscript! 21:38:38 Ah, the postscript neural network. 21:38:40 RodgerTheGreat: *clicks, browser freezes* 21:38:43 Oh, it loaded. 21:38:48 Neural networks are pretty simpl 21:38:49 e 21:38:54 only takes a second or two to compute 21:39:10 brb→ 21:39:47 ehird: about your link a couple of screens back, there was a story on Slashdot a few months ago about how someone had managed to cut through metal with a piece of bacon 21:39:48 ais523, and anyway I rely on reading different member of union having a sensible result. Should anyone actually run into real issues with it for cfunge I could add a memcpy() variant as a compile time option. 21:39:54 by making it really hot, or something 21:39:56 ←ais523: awesome→ 21:39:57 but then I want actual proof for it failing 21:40:11 especially if ehird says it 21:40:17 or you ais523 21:40:29 You know what'd be cool to build? A computer using cordwood construction. 21:40:29 ←DS9C→ 21:40:34 http://en.wikipedia.org/wiki/File:Cordwoodcircuit.agr.jpg 21:40:46 AnMaster: using unions for type punning is legal, IIRC, because that's what they were designed for 21:41:04 ais523, iirc reading from different member is implementation defined 21:41:08 in C 21:41:23 pikhq, wth is that 21:41:46 AnMaster: High-density circuit design using wire-ended components. 21:42:00 You stick the components between two circuit boards. ;) 21:42:09 pikhq, hard to solder I imagine 21:42:16 Yeah. 21:42:29 actually not too bad to solder, but a bitch to assemble on any kind of scale 21:42:29 Even harder how it was originally done. 21:42:30 Tin welding. 21:42:48 one C weirdness: you can use different sorts of pointers to point to different things, but all pointers to unions must have the same representation no matter what sort of union it is 21:42:51 the difficulty in aligning components grows exponentially relative to the number of components 21:43:33 AnMaster: what exactly are you unioning? 21:43:34 RodgerTheGreat: At the time, your alternatives for high-density circuit design were non-existent. 21:43:42 well, sure 21:43:47 This is before surface-mounted components and integrated circuits. 21:43:53 either this or freeforming 21:43:55 ais523, int32_t and float 21:44:04 AnMaster: how do you know float's 32 bits long 21:44:08 that's obviously unportable 21:44:11 And now, you'd only do it for similar reasons to doing a wire-wrapped computer. 21:44:14 ais523, IEEE says so 21:44:24 AnMaster: C doesn't say it's IEEE. 21:44:30 pikhq, true 21:44:33 that is why README says so 21:44:44 AnMaster: C doesn't demand IEE representation 21:44:47 well, wire-wrapping has some distinct mechanical advantages that can occasionally make it worthwhile 21:44:49 ah, pikhq beat me 21:44:59 ais523, and I already replied to why 21:45:08 And floats are often-times done using x87 instructions. 21:45:17 I think building a CPU as a freeformed circuit would be beautiful and impressive 21:45:21 Which are fucking crazy, instead of IEEE. 21:45:41 RodgerTheGreat: Mmm, yeah. 21:45:55 pikhq, sure. But as long as float are 32 bits I'm happy. Anyway FPDP and FPSP won't work if the floats/doubles are of other sizes 21:46:00 in any way whatsoever 21:46:25 if I had more EE chops I'd totally do it. If someone designed one that seemed to be on a feasible scale I'd help assemble it 21:46:26 ok, using a union for type-punning is permitted by 6.5.7, but there may be other reasons it's illegal 21:46:40 ais523, C99 Annex F (normative) 21:46:42 RodgerTheGreat: Same. 21:46:58 I could probably build one out of TTL logic chips, but that seems like cheating 21:47:10 ais523, exact behaviour is implementation defined iirc 21:47:25 I dunno. Free-form even with TTL logic would be impressive. 21:47:29 ais523, err what 21:47:30 6.5.7 Bitwise shift operators 21:47:31 .......... 21:47:35 ais523, that made no sense 21:47:43 Actually, if you could build one with TTL logic chips, you could probably hand-build one. 21:47:58 ais523, what section did you really mean 21:48:01 Surely you could just get a good design for logic gates instead? 21:49:42 AnMaster: I meant 6.5.7, unfortunately there are two versions of the C standard which are identical apart from section numberin 21:49:44 *numbering 21:49:51 ais523, I have the ISO one 21:50:04 ais523, ISO/IEC 9899:TC3 21:50:29 ais523, what section name did you mean 21:50:35 never mind 21:50:39 ais523, what? 21:50:43 let me finish reading, then I'll tell you 21:50:50 and I don't know offhand, I'd have to go back and look 21:50:51 http://en.wikipedia.org/wiki/Intel_80487 21:51:00 ... That is one bizarre 'coprocessor'. 21:51:12 pikhq, I know... 21:51:50 Your i487 SX could in theory be a downgrade. XD 21:52:27 pikhq, compared to what 21:52:46 Your previous 486 SX. 21:52:58 There was more than one clock speed on those. :p 21:53:00 err, "coprocessor"? 21:53:02 oh 21:53:06 pikhq, I see 21:53:22 pikhq, surely you needed matching speeds? 21:53:40 Not necessarily. 21:54:04 ais523: 'Coprocessor', since by inserting it, you disable the 80486 on your motherboard. 21:54:18 it's more an instead-processor 21:54:25 Yeah. 21:54:33 Thus the scare quotes. 21:54:49 http://en.wikipedia.org/wiki/File:Intel_Pentium_III_Katmai.jpg 21:54:55 that's one strange CPU mounting 21:55:30 you should see the heatsink 21:56:03 most PIIs were set up like that, and a lot of Sun's Ultra CPUs mounted the same way 21:56:17 RodgerTheGreat, err. I have a P3 that is normal ZIF mounting 21:56:21 Not was weird as the Slot 1 -> Socket something adaptors that you could use for Celerons... 21:56:46 RodgerTheGreat, copper mine 21:56:51 996 MHz 21:57:20 oh PII 21:57:21 right 21:57:33 pikhq, http://en.wikipedia.org/wiki/File:Pentium_III_on_motherboard.jpg ? 21:57:51 AnMaster: I'm confused: informative appendix J says that section 6.2.6.1 says that reading from a union member you didn't last store into is unspecified 21:57:52 http://en.wikipedia.org/wiki/Slot_1 21:57:52 ok 21:58:03 but I can't see a statement to that effect anywhere in section 6.2.6.1 21:58:22 6.2.6 Representations of types 21:58:23 6.2.6.1 General 21:58:28 yes, that one 21:58:28 are we talking about the same one this time? 21:58:30 ah 21:58:51 -!- jix has quit ("Lost terminal"). 21:59:34 7 When a value is stored in a member of an object of union type, the bytes of the object 21:59:34 representation that do not correspond to that member but do correspond to other members 21:59:34 take unspecified values. 21:59:37 what about that one 21:59:40 I'm not sure 21:59:58 ais523, seems to be that 22:00:12 AnMaster: no, not that 22:00:19 that's mentioned elsewhere in appendix J 22:00:24 ais523, then what does the quoted bit imply 22:00:36 it means, that say you have a union of a short and a long 22:00:45 ais523, I quoted 6.2.6.1 (7) 22:00:45 then if you assign to the short, the bits of the long you didn't assign to end up unspecified 22:00:49 ah 22:00:56 which makes sense I suppose 22:01:05 yeah it does 22:01:06 although leaving them the same would be more useful 22:01:15 The slot-mountable Pentiums always reminded me of NES game carts. 22:01:20 ais523, less efficient 22:01:28 ais523, on some architectures 22:01:40 ais523, say you can only write a 16 bits at once 22:01:56 then have a char and a long 22:02:08 (and 8 bit char) 22:02:23 yes 22:02:27 that's why it's done that way 22:02:42 the C standard is designed to allow implementations to be close to the hardware if they want to 22:02:45 ais523, or maybe reading/writing word size is faster, even if you can still do individual bytes 22:02:51 and do to weird high-level stuff too, if they want to 22:03:04 ais523, which end is GCC at 22:03:15 I don't know 22:03:17 probably both 22:03:28 ais523, you wrote a backend for it. Surely you know! 22:04:15 you don't need to understand all of gcc in order to write bits of it 22:04:54 AnMaster: A backend just does RTL->machine code translation. 22:05:00 hm ok 22:05:03 RTL doesn't tell you much about how it does C. 22:05:22 pikhq: actually, the backend generates the RTL too 22:05:35 ais523, why does it 22:05:45 gcc's architecture is interesting 22:05:51 you create templates that it converts standard opcodes into 22:05:56 like addition, for instance 22:05:58 that's RTL templates 22:06:00 into what? 22:06:06 ah 22:06:07 it then pattern-matches the templates you generated to create asm 22:06:21 but the RTL itself is modified in between 22:06:23 ais523, what about peep hole optimising 22:06:30 that's done on the RTL, normally 22:06:35 although you can do it on the asm if you really want to 22:06:40 yes and it is partly arch specific 22:06:41 well, RTL -> asm 22:07:00 but the RTL itself is modified in between <-- by what 22:07:14 AnMaster: gcc, of course 22:07:17 the less backendy bits 22:07:20 ais523, what part of it 22:07:24 and for what purpose 22:07:26 register allocation, mostly 22:07:30 and also various optimisations 22:07:31 ah 22:07:38 -!- kar8nga has quit (Remote closed the connection). 22:07:47 ais523, why was tha backend involved before that 22:08:13 AnMaster: because it generates the sort of RTL that works well on the target platform 22:08:26 ah 22:08:38 the idea is that say, if you're faster at bitshifts than addition 22:08:45 you'll generate a bitshift for x+X 22:08:46 ais523, so that is where you turn mov 0,%eax into xor %eax,%eax ? 22:08:47 *x+x 22:08:56 AnMaster: no! the mov 0,%eax is never generated 22:09:02 ais523, yes it is... 22:09:05 ais523, at -O0 22:09:27 well, the conversion works differently at different optimisation levels, it's a mess 22:09:35 ais523, ok.... 22:09:38 and the x86 code is /especially/ a mess 22:09:43 ais523, how comes 22:09:49 most interest in it? 22:09:53 presumably they couldn't be bothered to add a new constraint letter for "compile-time constant zero" 22:09:54 So most people working on it? 22:10:01 AnMaster: could be 22:10:02 Since sourcefurge's been talked about here occasionally, and not everyone has a project there I guess, here's a bit of information that was emailed today: "Discussion Forums will be replaced by phpBB." 22:10:11 yes, that was mentioned earlier 22:10:14 presumably they couldn't be bothered to add a new constraint letter for "compile-time constant zero" <-- what has this got to do with the inline asm 22:10:24 or do they have that elsewhere too 22:10:34 constraint letters I mean 22:10:37 AnMaster: gcc is very monolithic, OK, different parts affect other different parts 22:10:45 the constraint letters are used by the pattern-matcher 22:11:00 ais523, are they the same as the inline asm constraint letters? 22:11:01 and also by inline asm 22:11:02 yes 22:11:05 ah the same then 22:11:11 yeah then I know what you are talking about 22:11:22 the inline asm isn't their main use 22:11:27 they're /all over the place/ in the actual backend 22:11:49 and many of the letters, such as "compile-time constant zero", are not particularly useful in inline asm 22:11:56 ais523, they are a pain figuring out if you meant m, =m =&m or whatever 22:11:59 because if you know that something's a constant zero, why did you not just write a zero in your code 22:12:02 I think =&m is valid. Right? 22:12:14 I think so, but I don't deal with the inline asm syntax all that much 22:12:23 the punctuation marks are different there, even though the letters are the same 22:12:32 ais523, interesting 22:14:26 ais523, btw why is it unspecified if char is signed or unsigned without qualifier 22:14:34 since apart from that all are signed 22:14:39 short, int, long and so on 22:14:44 AnMaster: inconsistency between past implementations 22:14:48 C89 modeled existing practice 22:15:02 ais523, but all did signed for int? 22:15:03 the idea in writing C89 was that as many existing programs as possible should be C89-compatible without changes 22:15:07 AnMaster: apparently so 22:15:32 how much shorter would C99 become if you removed all such old compat... 22:15:45 I bet a few pages at least 22:16:44 ais523, btw, it is interesting that gcc shuts up about type punning by pointer if you cast the pointer to void in between 22:16:46 like 22:17:07 AnMaster: doing something extra-verbosely is normally taken by compilers as a hint to shut up 22:17:13 like putting double-parens around an assignment 22:17:24 (const v4sf*)&my_const_v4si; 22:17:27 it dislikes 22:17:35 ((const v4sf*)(const void*)&my_const_v4si); 22:17:36 it accepts 22:17:43 because it shows that you meant it deliberately 22:17:47 well I missed one pair of parantheses there 22:17:54 ais523, yeah I did in that case 22:18:03 ais523, and I deference it just outside 22:18:39 __builtin_ia32_movntps(((float*)(void*)&cfun_static_space) + i*4, 22:18:39 *((const v4sf*)(const void*)&fspace_vector_init)); 22:18:41 ah yes 22:18:45 ais523, like that ^ 22:19:41 ais523, takes quite a bit to make it shut up 22:19:47 bit of work* 22:22:05 -!- oerjan has quit ("Good night"). 22:24:06 ais523, pattern matching seems like the best way to implement most compilers to me 22:24:20 makes it easy to do in a lot of the cases 22:24:36 pentium 3 mounting is fun 22:24:40 because it's rather silly 22:24:45 i mean, why a slot? 22:24:46 ehird, pentium2 you mean 22:24:49 no, 3 22:24:52 ehird, because not all p3 had that 22:24:57 I know that 22:25:01 the katmais did 22:25:10 ehird, P2 did have the same thing as the katmais iirc 22:25:12 http://www.phoronix.net/image.php?id=amd_hd4770&image=amd_rv740_slide3_lrg World's most advanced graphics manufacturing process, revolving around terrible kerning. 22:25:15 ehird, and it is easier? 22:25:20 Surely AMD could do better than that. 22:26:12 ehird, yeah, even I notice the issues with the I in the second entry 22:26:20 or anywhere else 22:26:22 It would be nice if they could spell "semiconductor". 22:26:34 That too. 22:26:42 sure AMD are behind it? 22:26:45 Yes. 22:26:47 Such terrible kerning. 22:26:50 It's the announcement of the 4770. 22:26:54 the spelling is worse 22:26:56 Well, one page of it. 22:26:59 -!- tombom has quit ("Peace and Protection 4.22.2"). 22:27:00 but yeah 22:27:13 ehird, only upper case "I" seems affected 22:27:16 It's appealing to the mainstream market it's aimed at, clearly. 22:27:28 AnMaster: the whole kerning is pretty terrible but the uppercase I is the most noticable 22:27:35 that isn't even a kerning fail 22:27:44 ais523, what 22:27:44 that's just having the letter I far too wide 22:27:44 ais523: it's kerning 22:27:47 er 22:27:48 that's kerning 22:27:50 silly 22:27:58 ehird: kerning's to do with combinations of two letters 22:28:20 hrmw 22:28:21 well fine 22:28:25 err 22:28:30 I thought this was kerning 22:28:49 I think it is 22:28:52 but maybe pedantically it's not 22:29:16 kerning = the changing of letter spacing based on the text instead of a constant tracking, IMO 22:29:23 http://en.wikipedia.org/wiki/File:Kerning.png 22:29:34 If there was good kerning, it would be able to get rid of that extra space. 22:29:40 So, bad kerning *and* font design. 22:29:47 I think the font is verdana 22:29:52 which doesn't have a fucked up uppercase I 22:29:58 ehird, http://en.wikipedia.org/wiki/Slot_1 22:30:05 even so - using verdana in a product announcement? 22:30:09 who _designed_ this shit? 22:30:22 AnMaster: yes, I'm of the opinion it's rather silly 22:30:26 sockets look nicer :P 22:30:31 ehird, what about this http://en.wikipedia.org/wiki/File:Adapter_slot1-socket.jpg 22:30:35 THAT is silly 22:30:37 no 22:30:40 what about it 22:30:46 is it a socket→slot adapter? 22:30:47 THAT is ridiculous 22:30:49 it doesn't seem to mess up lowercase i... 22:30:49 seems so 22:30:50 ehird, yes 22:31:02 I like it, again because http://en.wikipedia.org/wiki/File:Pentium_II_front.jpg looks like a game cartridge. 22:31:13 fizzie: oh, it looks nice, and I agree 22:31:20 I just don't think it's a particularly good way to do a processor 22:31:21 It's like it's calling me to blow on the contacts before I stick it in. 22:31:23 for instance, what about heatsinks and fans? 22:31:35 you can't really do much when it's all encased in a slot 22:31:38 Slot_1 made me think of the DS 22:31:38 fizzie, what 22:31:52 AnMaster: when game catridges lied around 22:31:52 fizzie, blow contacts? HUH 22:31:54 they got dusty 22:31:58 ah 22:32:01 so you blew at the contacts before putting them in 22:32:01 blow in that way 22:32:03 otherwise you'd get glitches 22:32:13 not blow away in a destructive way 22:32:16 AnMaster: I don't think he meant he performed oral sex on game cartridges 22:32:20 dammit, you ruined my joke with timing 22:32:22 type slower 22:32:28 ehird, I meant as in "blow up" 22:32:36 party pooper 22:32:54 at your service :P 22:33:16 anyway why are the contacts hidden on http://en.wikipedia.org/wiki/File:Pentium_II_front.jpg 22:33:55 err, they are? 22:33:57 I can see them 22:34:20 ehird, the contacts doesn't stick out 22:34:24 very odd 22:34:41 Probably to protect against damage/dust/stuff? 22:34:45 maybe 22:34:46 AnMaster: most game slots are like that 22:34:47 That's exactly what makes it look like a game cartridge. 22:34:51 ah 22:34:55 Indeed 22:35:02 what about the holes on CF cards 22:35:12 you don't need to blow stuff clean there usually? 22:35:19 maybe because they are so small 22:35:19 hm 22:36:12 what is the holgram thingy there for 22:36:19 For kewlness. 22:36:26 huh 22:36:33 AnMaster: not a hologam 22:36:36 it's a diagram of the chip 22:36:44 but yes, for kewlness 22:36:44 no? 22:36:53 read http://en.wikipedia.org/wiki/Slot_1 22:36:58 says it is a hologram 22:36:59 AnMaster: like this one for the nehalem: http://images.apple.com/macpro/images/overview_features_processor20090303.png 22:37:03 well, it may be a hologram 22:37:08 but it has the design of the chip on it 22:37:12 Ooh, there's yet another SD standard, after SDHC? 22:37:15 Didn't know.. 22:37:23 FireFly: exciting! 22:37:26 FireFly, SD? 22:37:31 Secure digital 22:37:33 Mem cards.. 22:37:34 ah 22:37:34 The Secure Digital Extended Capacity (SDXC) format was unveiled at CES 2009. The maximum capacity defined for SDXC cards is 2 TB (2048 GB). SDHC cards also have a maximum capacity of 2 TB based on the card data structures, but this is artificially limited to 32 GB by the SD 2.0 document. 22:37:38 2TB memory cards? 22:37:38 Yeah 22:37:42 FireFly, isn't it closed spec? 22:37:42 That's hot. 22:37:43 Just what I'm reading :P 22:37:43 iirc 22:37:47 or is that MMC 22:37:56 * AnMaster sticks to good old CF. 22:38:02 it is what my camera can take 22:38:11 * FireFly prefers SD/SD micro 22:38:13 Ah, Compact Flash. 22:38:16 It's what my stuff uses 22:38:23 There might be some sort of official reason along the lines of assuring you that you're getting the a real processor and not just some old dusty pentium some counterfeiter's done a plastic slot case for. Maybe. 22:38:24 I prefer USB drives :-P 22:38:25 most high end cameras seem to prefer CF 22:38:27 even these days 22:38:27 FireFly: Compact Flash is really the better format. ;) 22:38:28 stuff = Camera, DS, other stuff 22:38:29 :) 22:38:35 pikhq, indeed it is 22:38:36 Since they come up to 128GB-that-you-can-actually-use 22:38:51 CF is the physically-larger format, though. 22:38:55 Maybe, but not the stuff I use :P 22:38:55 ehird, what do you mean "can actually use"? 22:39:03 AnMaster: " SDHC cards also have a maximum capacity of 2 TB based on the card data structures, but this is artificially limited to 32 GB by the SD 2.0 document." 22:39:10 fizzie: You can get freaking hard drives in CF format, though. 22:39:10 You can only use 32GB of SDcard. 22:39:11 ;) 22:39:11 fizzie, it is still way smaller than the camera battery 22:39:21 FireFly, or the lense system 22:39:23 err 22:39:24 fizzie, ^ 22:39:39 AnMaster: It's uncomfortably large for something like a mobile phone, though. (Which -- surprise, surprise -- might be why no mobile phones use it.) 22:39:47 ehird, SD sure. But who cares about SD 22:39:50 ehird, I use CF 22:39:57 I want a really huge SD micro card :( 22:40:03 AnMaster: People with a camera that isn't high-end. 22:40:10 And mobiles. 22:40:15 fizzie, yeah, I have no idea why I would want it in a phone 22:40:27 People like FireFly do 22:40:28 ehird, um my mobile uses some other card 22:40:32 XD? 22:40:34 Your mobile, perhaps. 22:40:35 iirc 22:40:38 Keyword yours. 22:40:46 ehird, Nokia 22:40:47 wait 22:40:53 SD micro it seems 22:40:57 :D 22:41:01 but I don't have a card for it 22:41:04 nor do I want one 22:41:05 ehird: Currently, CF is only up to about 100GB. The limit, though is something like petabytes... 22:41:19 the built in 30 MB is enough for the address book 22:41:27 what else would I need 22:41:46 MicroSDs are quite extremely small, thought 22:41:48 MicroSD might be the most common mobile phone thing nowadays; Nokia had a lot of (or at least some) MMC using models at some point, though. 22:41:49 also CF is faster iirc 22:42:10 I mean, I'm surprised how you can fit 8 gig in something as thin as a nail 22:42:11 CF is PATA speed. 22:42:24 -!- upyr[emacs] has quit (Remote closed the connection). 22:42:25 pikhq, pretty sure it is higher than most though 22:42:43 pikhq, UDMA5? Or higher? 22:42:56 UDMA 133. 22:43:05 What's amusing is that they still show CF card speeds in "Nx", where N is the 150kbps audio-CD data rate. 22:43:14 pikhq, wow what 22:43:14 Future revisions will implement SATA. 22:43:30 pikhq, so new cards won't work on my current camera? 22:44:01 It will be a different, similar format. 22:44:09 "CFast". 22:44:21 ... Fuck. 22:44:37 pikhq, huh 22:44:38 CFast will use a standard SATA data connector and a slightly different power connector. 22:44:53 pikhq, that will be quite a large connector 22:45:09 for the card 22:45:29 I'd expect a much smaller power connector. 22:45:40 That's the interface speed, though. I don't think I've seen advertised speed more than something like "300x", which would mean 45Mbps. 22:46:01 fizzie: That's because it's Flash. 22:46:11 s/Flash/flash/ 22:46:20 A microdrive can probably hit faster speeds. 22:46:35 um 22:46:45 I thought flash was faster than harddrives? 22:46:47 They have that smaller "microsata" connector (not sure about standardization status there), wonder why not use that in some SATA-based CF-y thing. 22:46:48 Or maybe I'm being dumb. 22:46:57 AnMaster: Er, SSDs are. 22:47:12 AnMaster: An SSD is, flash memory is not generally. 22:47:17 ehird, aren't those basically the same but done as harddrives instead of flash cards? 22:47:28 AnMaster: Not really. 22:47:37 ehird, different technology? 22:47:44 No. 22:47:48 then what 22:47:52 I'm not too certain; ask pikhq. 22:48:24 More flash chips, for one thing. 22:48:28 ah 22:48:35 You can write to/read from multiple in parallel. 22:48:43 Yeah 22:48:53 They use smaller, faster chips. 22:50:15 Flash memory with large capacity stores more than one bit in each flash cell. 22:50:27 This gets a lot of storage, but is slower. 22:50:43 ah interesting 22:50:44 pikhq: Nope, actually. 22:50:49 the X25-M is about as performant as the X25-E. 22:50:51 Sometimes even moreso. 22:51:02 iirc, if you do a fuckton of sequential reads/writes it can be faster, but not by much 22:53:53 ehird, hm it should be safe to open the case of a SSD right (while it is definitely unsafe to do that in case of a harddrive!) 22:53:54 A hard drive is still somewhat faster for sequential operations, and a hard drive is much, much faster for small writes. 22:54:04 AnMaster: Sure. 22:54:10 Anand did that his SSD reviews. 22:54:18 's just flash memory + a controller. 22:54:20 who 22:54:29 Anand Lal Shimpi of anandtech.com 22:54:32 (hard drive doesn't have to seek for the first, and for the second, the hard drive doesn't have to do 1MB-granularity writes) 22:54:33 ah 22:54:35 The author of the ever-helpful SSD Anthology 22:54:52 22:53 pikhq: A hard drive is still somewhat faster for sequential operations, and a hard drive is much, much faster for small writes. 22:54:54 Sorry, NO. 22:55:02 the X25-M is faster than harddrives in every case. 22:55:09 For sequential reads and writes, for random reads and writes. 22:55:10 ehird, [citation needed] 22:55:16 It's OPTIMIZED for tiny writes, chrissake! 22:55:24 AnMaster: http://www.anandtech.com/printarticle.aspx?i=3531 22:55:33 read the 31 pages or just trust me 22:55:35 ehird, independent source 22:55:39 that is independent 22:55:44 no ties to Intel or anything 22:55:48 independent from anand 22:55:49 duh 22:55:51 and intel 22:55:53 try the google 22:56:00 or buy an ssd and try it yourself 22:56:03 Is the X25-M Flash memory or battery-backed RAM? 22:56:06 pikhq: flash memory 22:56:12 there's a reason it's recommended as a disk drive: 22:56:16 it's really really fast on <4KB random write 22:56:17 s 22:56:26 (and everything else, but especially so) 22:57:13 -!- olsner has quit ("Leaving"). 22:57:50 ehird: ... That X25-M is slower for sequential writes. 22:58:00 in some cases 22:58:06 And by "small writes", I should've said "for a single small write at a time". 22:58:17 err, no 22:58:25 for random reads/writes, an SSD is definitely way faster 22:58:39 Obviously, if you're doing more than one at a time, the SSD is guaranteed to be faster. 22:58:43 No seeking. 22:58:46 also... 22:58:49 pikhq: http://images.anandtech.com/graphs/thessdanthology_031809001858/18640.png 22:58:56 The x25-E is faster there 22:59:00 Which is still an SSD 22:59:24 so it's MLC that does that 22:59:30 still, you don't do sequential writes much on an OS drive 22:59:39 and the sequential read performance is still great: http://images.anandtech.com/graphs/thessdanthology_031809001858/18639.png 22:59:47 and the random read/write performance blows everything else out of the water 22:59:49 Okay, the X25-E is just stunning. 23:00:07 so you get about 30MB/sec slower than a 5400rpm HD with an X25-M when doing the rare operation of big sequential writes 23:00:16 that's not really a big deal at all 23:00:32 ehird, rare seq read? 23:00:37 since it's a rare operation on an OS drive, and not actually all that much slower than HDs 23:00:40 AnMaster: sequential WRIT 23:00:40 E 23:00:43 ah 23:00:44 * pikhq does sequential writes somewhat often. :p 23:00:48 -!- BeholdMyGlory has quit (Remote closed the connection). 23:01:01 pikhq: You do a bajillion random writes every time you load a page in your browser. 23:01:14 And a bajillion random reads when you start a program. 23:01:25 ehird: Uh, no. 23:01:30 Large sequential writes are stunningly rare in comparison to everything else an OS drive does. 23:01:34 *OS drive*. 23:01:39 Not media drive. 23:01:42 I would think the whole cache thing would mitigate the write issues somewhat, especially given that you're likely to have 6 gigabytes of memory free for caching data. 23:01:50 It's not like firefox's going to sync() after each write. 23:01:52 Why the fucking hell would a browser do random writes to your hard drive? 23:01:56 pikhq: Dude. Cache. 23:02:04 pikhq: Look at firefox sometime. 23:02:10 fizzie: actually, a bug in firefox 3 rc meant it /did/, on Linux 23:02:11 It does a ridiculous amounts of writes per page loaded. 23:02:13 Browsers tend to keep cache in RAM and then later commit it to disk. 23:02:13 but that was fixed 23:02:14 ais523: he was joking 23:02:19 ehird: ah, ok 23:02:30 The random reads, though, is a good point. 23:02:38 pikhq: anyway, if you're seriously arguing that an HD is faster than an SSD for OS drives 23:02:44 go buy an SSD, and pop it in your machine, and copy your OS over. 23:02:51 ehird: I'm not. 23:02:52 and then weep at the performance ;-) 23:03:00 pikhq: you seemed to be implying it 23:03:00 I'm just saying that an SSD isn't better in all cases. 23:03:13 It's not better when you're doing large sequential writes and totally need max speed on them. 23:03:14 And I stand corrected with that X25-E. 23:03:20 Which is terribly rare, to be honest. 23:03:27 That's just stunning. 23:03:35 pikhq: Yes; but the X25-E is expensive. 64GB max space, for the cost of the 160GB M one. 23:03:46 And the M beats it slightly on the more conventional operations quite a bit. 23:03:50 Eh, not worth it yet, but it will be. 23:04:01 Nah; MLC is the way forward. 23:04:05 For now, I'll content myself with cheap magnetic disks. 23:04:08 Except in enterprise server environments. 23:04:11 Which is the target market of the _E. 23:04:12 *-E 23:04:33 pikhq: Have fun with that; I'll be attempting to snap my fingers fast enough to measure the opening speed of applications. 23:05:00 I don't really see how the opening speed of applications is relevant at all, but I guess that might be just my peculiar use case. 23:05:09 ehird: I'll be holding onto my money. 23:05:21 fizzie: Well, it's not; let's say operating speed of applications. 23:05:23 But that's so vague. 23:05:41 "And this stack of bills went towards my education instead..." 23:05:48 pikhq: I'll be spending thousands on every other part of my machine and thus making the reasonable decision for a high-end assembly :-P 23:05:53 Mwaha. 23:05:57 <_< 23:06:01 >_> 23:06:03 ehird: Well, sure, if you're spending thousands. 23:06:21 Get a single SSD and a bunch of cheap disks for bulk storage. 23:06:22 'snot my fault "hardware is cheap" only applies relatively. 23:06:32 (or if you're spending tens of thousands, bunch of pricy SSDs) 23:06:38 pikhq: Single SSD on / + single 2TB disk on /home/ehird/media. 23:06:55 ehird: That's a high-end setup there. 23:06:58 Spiffy, though. 23:07:14 pikhq: a 2TB 7.5k RPM drive actually only costs around ~$170 23:07:23 which is just a little more than a 1TB drive 23:07:35 "Little more"? 23:07:41 1TB drives are about $130 23:07:42 1TB drives are under a hundred. 23:07:43 I don't actually think the operating speed of applications is relevant for me either, since I don't do much waiting; I guess I could be mildly pleased with faster compilation times, though. 23:07:48 Under a hundred? Where? 23:08:03 fizzie: Well, if you're totally satisfied with performance, obviously increasing it isn't going to satisfy you. 23:08:22 fizzie: Compilation is not very disk-bound, though. 23:08:28 Over here 1TB is around 100 €, 1.5TB is around 130-140 €, 2TB is around 300 € 23:08:32 fizzie: An SSD is about general snappiness. 23:08:54 ehird, matters for boot mostly 23:08:57 pikhq: OK, cheapest 1TB drive on newegg is $79.99 23:09:00 so I stand corrected 23:09:03 AnMaster: no, it matters for everything 23:09:13 ehird, once system is up I tend to use the same running set of apps all the time 23:09:14 mostly 23:09:16 Deewiant: 81.90 eur ($112 approx.) seems to be the cheapest 1TB drive in verkkokauppa.com. But Finland's not a cheap place for hardware. :/ 23:09:25 it's the difference between 0.5-1 second of waiting and 0.1 second of waiting 23:09:28 which is huge, psychologically 23:09:34 verkkokauppa.com is what I was looking at, too 23:09:39 So, that $170 2TB drive is more than twice the cost of a 1TB drive. ;) 23:09:47 Deewiant: It's the de-facto standard. 23:09:51 pikhq: a cheap one. 23:09:57 And that 81.9 € one is 5400 RPM 23:09:58 $130 seems to be more common 23:09:59 Fair enough. 23:10:17 And yeah, getting hardware in Finland sucks. 23:10:17 Deewiant: The 82.90 is 7200 RPM. 23:10:26 Right you are. 23:10:56 Getting hardware in continental Europe sucks in general. 23:11:06 Deewiant: It seems like buying overseas and paying the shipping could be cheaper for finns. 23:11:13 They do advertise a lower power usage for the 5400 RPM model. 23:11:31 ehird: Most US stores seem to have a shipping model of "California and Wyoming" or whatever 23:11:51 I'm annoyed that I can't use the nice Nexus Value 430W psu for my machine; maybe I'll get one of those "10 watts less!" super-green harddrives to ease my soul! 23:11:56 They don't even bother to state on their pages that they ship to the US only, it's so obvious :-P 23:12:03 There's also the "whole import tax and other stuff" thing for outside-EU ordering. 23:12:11 The UK seems to have okay prices on tech. 23:12:13 s/"whole /whole "/ 23:12:21 And shipping probably wouldn't be a huge problem. 23:12:25 For hardware, US > UK > Japan >> Germany > the rest 23:12:35 The US just rocks because of newegg. 23:12:35 Not sure if China should be there somewhere. 23:12:43 God, I wish they shipped to the UK. 23:12:45 The Atom box and assorted hardware came from UK, since it's EU-internal. 23:12:52 Deewiant: Japan > UK; we just have a few okay sites. 23:12:56 Japan has *everything*. 23:13:14 Well, whatever. There's still that >> there. 23:13:16 Scythe's website is so unprofessional. "6Heat Pieps" 23:13:45 The problem isn't shipping costs; I'd gladly pay shipping costs if there were any store that actually shipped what I wanted to Europe 23:13:56 Deewiant: China's probably cheaper, because of insanely crappy knock-offs. 23:14:10 Price wasn't the only thing I was considering 23:14:23 If the only stuff available is crappy knock-offs then that's not too good either :-P 23:20:23 it's the difference between 0.5-1 second of waiting and 0.1 second of waiting 23:20:25 for what 23:20:46 For operations to complete. 23:20:48 ehird, most of the stuff is in page cache early on for me 23:21:28 ehird, what specific operations. Most stuff I do is either CPU bound, memory bound, or just very harddrive trashing (would wear out a flash drive soon) 23:22:12 i challenge you to wear out an SSD quickly 23:22:20 ehird, use it as swap :P 23:22:25 I said quickly. 23:22:26 I won 23:22:43 ehird, sure. Use it as swap on a a system with 128 MB RAM 23:22:45 Run KDE 4 23:22:51 I said quickly. 23:22:59 ehird, it will be quickly 23:23:16 -!- FireFly has quit ("Later"). 23:23:23 Yes, because SSDs are the least reliable things ever and break with 3 seconds of use. 23:23:53 um. 3 seconds... is that how you define quickly? 23:24:03 I define quickly to "within two years" 23:24:09 Please look up "hyperbole". 23:24:12 while true;dd if=/dev/urandom of=/dev/hda;done 23:24:14 :p 23:24:17 pikhq: hda? 23:24:21 You think they use *IDE*? 23:24:21 sda 23:24:40 Linux no longer uses hda 23:24:43 even for PATA 23:24:49 it uses /dev/sd* for all 23:24:51 ... 23:24:52 Ah, right. Thinko. 23:25:00 AnMaster: er, for IDE... 23:25:01 Well, that's up to you; certainly the old PATA stuff is still in the kernel. 23:25:04 ehird, yes too 23:25:06 I should know that, I've got /dev/sd* for IDE and SATA here. 23:25:09 Heh. 23:25:14 fizzie, well true 23:25:15 (though to be fair, 23:25:19 fizzie, but why use it? 23:25:23 I just use /dev/mapper) 23:25:46 AnMaster: When I last bothered to look, quite a lot of the PATA drivers on the libata combined-SATA-and-PATA side were marked as EXPERIMENTAL. 23:25:58 fizzie, hm maybe. 23:26:04 That was quite a while ago, though. 23:26:15 fizzie: There's the general-purpose PATA driver. :p 23:26:18 fizzie, last I read most worked fine even thought they were marked like that 23:27:12 The Atom box has the CF card adapter in the PATA bus (speed's not an issue) and it's /dev/sda there, because CONFIG_ATA_PIIX wasn't marked experimental any longer. 23:27:44 Quite a lot of the experimental markings have gone away. But the heading still reads "Serial ATA (prod) and Parallel ATA (experimental) drivers". 23:28:44 On this desktop I just twiddled the bios bit to turn the SATA controller to AHCI mode, and forgot about PATA support; I don't have any hardware in the bus anyway. Truly a box of the future, even if it's some years old. 23:28:52 I think it's going to be marked that way until they can remove the old PATA stuff. 23:29:34 * pikhq goes out to mow 23:29:42 fizzie, what about DVD 23:29:50 AnMaster: SATA for that too. 23:29:57 mow? 23:30:12 fizzie, heh why. surely the DVD is slower than SATA 23:30:17 why would it benefit from it 23:30:20 'Cuz pata sucks? 23:30:30 The cable is smaller, blocks airflow even less than rounded IDE cables. :p 23:30:36 i hate pasta. 23:30:41 oh you said pata 23:30:44 not pasta 23:30:46 EVEN SO 23:30:48 i hate pasta. 23:30:52 what 23:30:53 how 23:30:54 can 23:30:54 fizzie, ah true 23:30:56 you 23:30:58 hate 23:31:00 pasta 23:31:02 you 23:31:03 I did mention the drive ("TSSTcorpCD/DVDW SH-S183A") but admittedly the from the name it's not obvious which interface it has. 23:31:04 demon 23:31:06 :| 23:31:08 pasta rocks 23:31:09 i just dont like it 23:31:09 :D 23:31:10 in general 23:31:14 i mean, some pasta is ok 23:31:16 dont get me wrong 23:31:21 ehird, Pasta with Swedish meatballs 23:31:31 but my grandfather makes pasta and its so bad when he does it, so its completely ruined pasta for me. 23:31:33 ehird, very nice 23:31:36 I'd think psygnisfive's tastes would be more pasta with Swedish balls... 23:31:49 mmm swedish food 23:31:55 ehird, especially when my grandmother makes the meatballs 23:32:02 psygnisfive: That's not what I meant. 23:32:03 anmaster: ikea sells swedish food and i love it. 23:32:06 It was a pun on your gayness, you see. 23:32:11 i know it was 23:32:13 Good. 23:32:14 fi:pata equals something like en:stew (as in the food class). 23:32:16 I'm glad we're all aware. 23:32:16 "Your gayness" 23:32:20 Yes. 23:32:22 psygnisfive, IKEA food is horrible 23:32:25 this is the title for drag queens 23:32:25 :D 23:32:25 at least in Sweden 23:32:29 Your Gayness! 23:32:38 psygnisfive: You have a gaycore in your body, which transmits gayity to the world. 23:32:38 anmaster: im sure it is, but i dont mean the food they prepare 23:32:41 i mean the packaged stuff 23:32:47 It also provides alternate power in case your heart stops beating. 23:32:58 psygnisfive, they don't sell that here 23:33:13 like that yummy chocolate-coated hard caramel candy 23:33:13 AnMaster: Of course not; it'd be like selling ice to an eskimo. 23:33:16 i forget what its called 23:33:18 AnMaster: duh; you guys can get it in a food store 23:33:19 fizzie, yeah 23:33:26 it's curios for the foreigners 23:33:34 its in the orange and blue packaging 23:33:37 ehird, you can get stuff from all over the world in the food shop 23:33:38 surely you know this stuff anmaster 23:33:38 daimo! 23:33:39 IKEA does sell some packaged food supplies here in Finland, though they aren't exactly... exotic. 23:33:39 thats it 23:33:41 daimo. 23:33:48 AnMaster: well sure, but not with swedish pacakging and the like 23:33:51 *packaging 23:33:51 It's just "Daim" here. Do they add an o in your place? :p 23:33:56 ehird, so can you 23:33:58 oh maybe its just daim 23:33:59 what about rice 23:34:00 i dont know 23:34:02 anyway 23:34:03 it's from China after all 23:34:04 OMG DAIM 23:34:09 AnMaster: does it have chinese packaging? 23:34:09 no? 23:34:09 and other countries close to it 23:34:22 ehird, what does the packaging look like then at IKEA 23:34:25 also, you swedes have some sort of soft flatbread 23:34:28 Swedish. 23:34:31 It looks very Swedish. 23:34:33 its very white and looks like a cushion 23:34:33 psygnisfive, uh... 23:34:33 Infallopedia says Daim is called "Dime" in the UK/Ireland. 23:34:36 Like Swedish food direct from sweden. 23:34:37 tunnbröd? 23:34:39 fizzie: Oh, Dime? 23:34:40 yes! 23:34:40 is that what you mean 23:34:41 That stuff is ick. 23:34:42 tunnbrod 23:34:46 i love it mm 23:34:47 psygnisfive, NO NO 23:34:52 psygnisfive, bröd 23:34:56 im lazy 23:34:58 psygnisfive, different letter remember 23:35:00 tunnbroed 23:35:04 Seriously. 23:35:06 Dime is awful! 23:35:11 Apparently it's Dajm in swedeland. 23:35:11 ehird, "dime"? 23:35:15 AnMaster: Dajm. 23:35:16 daim is delicious, shut your mouth 23:35:19 ehird, oh that 23:35:31 ehird, yeah it is awesome 23:35:37 NOOOOOOOOOOOOOOOOOOOOOOOO 23:35:40 lets see what else man 23:35:43 where is Dajm from? 23:35:45 YOU ALL HAVE TERRIBLE TASTES IN CHOCOLATE 23:35:48 AnMaster: Sweden. 23:35:48 AnMaster: Sweden, says WP. 23:35:51 oh 23:35:52 http://en.wikipedia.org/wiki/Dime_Bar 23:35:53 ehird its not the chocolate thats awesome 23:35:54 I thought it was from US 23:36:01 "originated in Sweden in 1953" 23:36:13 oh, anmaster 23:36:20 your pear soda stuff? <3 23:36:28 ehird, my grandmother makes a very tasty cake with bits of dime in it 23:36:29 AnMaster is actually behind all things swedish 23:36:30 true story 23:36:48 and just all sorts of swedish food 23:36:49 so yummy 23:37:07 http://sv.wikipedia.org/wiki/Daim <-- "Dajm"...? 23:37:17 mmm daim 23:37:31 "Daim är en sötsak tillverkad av Kraft Foods." <-- "Daim is a sweet produced by Kraft Foods." 23:37:32 anmaster, check en.~ 23:37:39 googling shows Kraft Foods is US based 23:37:42 yeah 23:37:42 yeah en. says it's Dajm 23:37:43 riggt 23:37:45 right* 23:37:47 AnMaster: it originates from sweden 23:37:51 but I guess a US company makes it now 23:37:53 kraft foods is english yes 23:37:53 ehird, I have seen both spellings in Sweden 23:37:56 but you can buy it in UK Ikea anyway 23:37:57 er 23:37:59 eamerican 23:38:02 it is funny 23:38:03 eAmerican. 23:38:04 english speaking surely! 23:38:06 They are cyberspace americans. 23:38:09 Kraft is a Swedish word 23:38:20 meaning Power/Energy 23:38:24 lol 23:38:25 err 23:38:29 Force I think 23:38:29 POWER FOODS 23:38:30 rather 23:38:33 POWER THIRST 23:38:37 MADE WITH LIGHTNING 23:38:39 REAL LIGHTNING 23:38:44 no 23:38:45 Force 23:38:47 As far as I can make out, Kraft Foods ate the whole Marabou thing, which was the Dajm-maker. 23:38:49 is the right translation 23:38:54 can i just say, that NEVER gets old 23:39:05 AnMaster: Powerthirst: http://www.youtube.com/watch?v=qRuNxHqwazs That Munctional-Erlang thing parodies it. 23:39:11 as in the think you measure in Newton 23:39:14 munctional? 23:39:24 psygnisfive: http://www.youtube.com/watch?v=1yH_j8-VVLo 23:39:24 is this a programming language that Mr Monk uses? 23:39:25 what? 23:39:26 monktional. 23:39:27 Functional programming for men. 23:39:31 AnMaster: That MUNCTIONAL Erlang video. 23:39:37 HAH 23:39:39 MUNCTIONAL 23:39:39 It's a parody of the Powerthirst thing that psygnisfive was quoting. 23:39:41 http://www.youtube.com/watch?v=qRuNxHqwazs 23:39:45 brilliant 23:39:47 ehird, I'm not familiar with it 23:39:53 got a link to it ehird ~ 23:39:59 AnMaster: http://www.youtube.com/watch?v=qRuNxHqwazs 23:40:04 But you've seen the munctional one. 23:40:06 ENERGY FOR ME 23:40:06 ehird, what? Where? 23:40:07 You said it was old when I linked it. 23:40:08 MENERGYYYYY 23:40:10 ehird, hm 23:40:12 men* 23:40:13 :P 23:40:14 ehird, which one... 23:40:15 AnMaster: http://www.youtube.com/watch?v=1yH_j8-VVLo anyway 23:40:28 http://www.youtube.com/watch?v=qRuNxHqwazs or http://www.youtube.com/watch?v=1yH_j8-VVLo 23:40:48 YOULL HAVE SO MANY BABYES 23:40:51 FOUR HUNDRED BABIES 23:40:57 AND THEYLL RUN AS FAST AS KENYANS 23:41:00 IN A RACE AGAINST KENYANS 23:41:01 AnMaster: First is powerthirst, second is munctional. 23:41:07 AND THEYLL GET IN A TIE WITH KENYANS 23:41:08 First is the original, second is the erlang-based parody. 23:41:12 AND GET DEPORTED BACK TO KENYAAAAAAAAAAAAAAAAAAAAAA 23:41:18 psygnisfive: Yes. I've seen it too. 23:41:29 sorry, im just enjoying it ok. 23:41:30 god. 23:41:36 have you seen the nutrigrain one? 23:41:44 similar play play on the babies thing 23:41:58 I love the p-maps tour maps part. 23:43:18 ehird, which is the original and which is the joke 23:43:23 Powerthirst is original. 23:43:25 But it's a joke too. 23:43:29 The Munctional is the derivation. 23:43:32 ehird, on what 23:43:33 How many times have I said that? 23:43:39 AnMaster: It's just. A joke. 23:43:42 ah 23:43:46 It's parodying energy drink adverts to a degree. 23:43:55 the MUNCTIONAL guy doesnt have the right amount of energy in his voice 23:44:00 HE SHOULD DRINK SOME POWERTHIRST 23:44:16 also, too much PARALLEL 23:44:26 erlang is all about PARALLELISM. 23:44:39 PARALLEL search on your PETABASE sorted IN PARALLEL. 23:44:56 lol 23:44:58 this is ridiculous 23:45:21 this is poorly thought out 23:45:33 so's your faec. 23:45:35 face 23:45:41 oh no mah faec 23:45:42 D: 23:45:47 i liked it tho 23:46:10 ehird, actually sorting in parallel with mapreduce is fun 23:46:18 I have done it in erlang of course 23:46:24 mapreduce isnt just map + reduce innit 23:46:24 "sorting in parallel" by itself is impossible 23:46:29 well, quite impossibl 23:46:30 e 23:46:31 its map + sort-into-buckets + reduce 23:46:32 across multiple nodes 23:46:33 ehird, no 23:46:37 ehird, mergesort 23:46:43 AnMaster: that's not sorting every element in parallel 23:46:46 can be made parallel 23:46:59 psygnisfive: no, it's just map/reduce 23:47:14 ehird, of course not fully, but reduces workload 23:47:16 nuh ehird 23:47:21 its also got a sort component 23:47:23 since combining two sorted lists is easy 23:47:36 psygnisfive: maybe google map reduce 23:47:42 yeah google mapreduce. 23:47:46 dunno of any other 23:47:55 actually. I was talking of google style map reduce 23:47:59 in this case 23:48:04 hadoop specifically, i think, has it. 23:48:08 rather than the two functions map and fold 23:48:25 duh 23:48:52 ehird, ? 23:48:55 mapreduce is kinda cool tho 23:49:00 23:48 AnMaster: rather than the two functions map and fold 23:49:03 i knew that 23:49:08 but there are non-google implementations of MapReduce 23:49:12 ehird, yes that is the alternative meaning I can think of 23:49:19 i watched a SICP class from berkeley and the prof demoed hadoop 23:49:23 which is what I meant by "maybe google map reduce" 23:49:24 or taught about it, anyway 23:49:29 it was quite interesting 23:49:34 ehird, well yes. I'm using an implementation in erlang then 23:49:36 especially with the scheme interface 23:49:51 but for sorting you of course need a sorting step 23:49:57 I think that is what we were talking about 23:50:00 ... 23:50:15 map/reduce isn't really applicable to sorting 23:50:20 other than that there wouldn't be any sort involved 23:50:25 since it's hard to do a sort efficiently with reduce 23:50:29 and map is rather useless for it 23:52:07 ehird, what about merge sort 23:52:21 AnMaster: that's not a pure reduce 23:52:27 ehird, true. 23:52:59 hmm with 85% tdp utilization for the cpu, 90% system load and no capacitor aging, http://extreme.outervision.com/PSUEngine says my system will use 328 watts 23:53:07 but tweaking a bit puts it over 430W 23:53:08 ehird, anyway I would suggest quickmergesort, merge sort to split load across the number of nodes/cpus. Then each node/cpu doing it's own sorting with heap sort 23:53:10 ok im off for a while guys 23:53:10 err 23:53:11 see ya 23:53:11 i wonder 23:53:12 or quicksort 23:53:15 -!- psygnisfive has quit ("Leaving..."). 23:53:16 -!- ais523 has quit (Read error: 104 (Connection reset by peer)). 23:53:17 Deewiant: know anything like that? 23:55:37 ehird, anyway, I much prefer sorting networks 23:55:51 a pitty it needs specialised hardware 23:56:59 fpga? 23:57:21 ehird, that is slow 23:57:22 :P 23:57:42 AnMaster: fpgas can raytrace at 5fps 23:57:45 that's not slow. 23:57:58 ehird, compared to ASIC it is 23:58:09 AnMaster: do you have any idea how hard raytracing is 23:58:10 ? 23:58:10 ehird, where are GPUs that can do way better 23:58:13 5fps is basically a miracle 23:58:14 there* 23:58:14 LOL 23:58:15 no 23:58:20 ehird, the last ones yes. 23:58:20 you cannot do 5fps raytracing with gpus 23:58:39 there's a reason there's a whole company based around their realtime-raytracing FPGAs 23:58:46 and it's that it's way ahead of anything else 23:58:49 ehird, I have seen software ray tracing on a i686. Demo guy made it 23:59:07 at 6 FPS iirc 23:59:08 Yes, there are demos with it. But the scenes they render are simple, and the quality is not high. 23:59:17 ehird, yeah 23:59:20 these do 5-10FPS for high-quality images 23:59:24 as in, same quality as end result 23:59:37 so, yes. FPGAs are fast. 23:59:46 ehird, what about Larrabee 23:59:49 spelling