00:00:34 Sgeo: What is a norn exactly? 00:01:06 The main focus of the Creatures series, an artificial life form with a brain and biochemistry 00:01:15 https://creatures.wiki/Norn 00:01:21 Sgeo: That must have been an interesting introduction for cryptsting. 00:01:57 cryptsting, we don't torture people here. Unless computer programs are people. 00:02:11 (And even then, if people turn into computer programs someday, I won't torture them) 00:02:31 Sgeo: What makes the Norns go at the programmatical level? 00:02:45 Sgeo: But what if computer programs turn into people? 00:02:45 Actually, someone I know kind of gets on my case about being careful as AI advances 00:03:30 There're probably being smart. But who cares? We have things to torture. 00:03:48 hppavilion[1], neural network made of several lobes. Input lobes which receive data from the engine (including drives and what's nearby), a combination lobe which combines them, and output lobes, one for an action and one for a noun (attn) 00:03:52 (But not living things. For the record. Unless you count a Neural Network as living) 00:03:57 AFK 00:04:20 Sgeo: Ah, that makes sense. I might use that model in my hell. 00:12:22 -!- relrod_ has changed nick to relrod. 00:12:51 -!- cryptsting has quit (Read error: Connection timed out). 00:13:25 so much for a `relcomed newcomer :/ 00:13:33 fungot: were you ever `relcomed? 00:15:56 How does one go about creating an esoteric language? Is it easier to compile it, or translate to something else (c, for example)? 00:17:39 Do you mean define, or do you mean implement? 00:24:04 <\oren\> hi 00:24:32 I mean I'm new, and have no clue what I just asked. 00:25:56 Thanks Google. I suppose I mean implementing. 00:26:21 <\oren\> you can either make a compiler, translator, or interpreter 00:26:50 <\oren\> compiler: takes a program in language X and produces a native EXE or ELF executable 00:27:14 <\oren\> translator: takes a program in language X and produces a program in language Y 00:28:00 <\oren\> interpreter: takes a program and language X and produces the effect of running that program 00:28:32 <\oren\> any language can be used to write these 00:28:51 Hm.. It sounds an interpreter would be easiest? 00:29:01 <\oren\> usually, yes 00:29:24 So as someone who has never done something along these lines, would it be suggested that I go that route? 00:30:15 <\oren\> yeah 00:32:37 Oh hey, people talking about Creatures? 00:32:40 Neat. 00:39:20 MDude: Do you have any idea what I should do with neural networks? I want to do something, but I don't know what xD 00:39:53 hppavellon[1]. MNIST! 00:40:16 boily: MNIST? 00:41:17 B1ood6od: I recommend using an interpreter, definitely. But that's for implementing, not for creating. 00:42:22 -!- jaboja has quit (Ping timeout: 260 seconds). 00:42:49 hppavilion[1]: MNIST. http://yann.lecun.com/exdb/mnist/ 00:43:54 boily: Oh, that's a database. I'm wondering what I should /do/ with neural networking in the /first/ place. 00:45:53 hppavilion[1]: so then does 'creating' imply designing? 00:47:11 it's a very good dataset to play with. with a neural network you can (hopefully) classify stuff, in this case handwritten digits. 00:51:34 B1ood6od: Yes. 00:51:42 Most certainly. 00:53:50 Sgeo: An idea for a hell is, instead of pain, using fear. 01:00:25 OK, Imma make a brain called Hive. Not sure how to go about it, but who cares. 01:00:40 brains are overrated. 01:00:52 I feel bad if I scared cryptsting 01:01:08 I don't think there's really a difference between pain and fear in norn brains 01:01:36 Just what can cause them. There IS a difference between pain/fear and the rest of the drives.. the presence of the former temporarily lowers other drives 01:02:06 (With the exception of one of the hunger drives, which when it returns turns into a different drive... pretty sure that's a bug in the genome) 01:16:50 -!- jaboja has joined. 02:13:26 helloily 02:14:15 -!- B1ood6od has quit (Read error: Connection reset by peer). 02:16:08 -!- hppavilion[1] has quit (Ping timeout: 246 seconds). 02:18:50 boily? 02:23:05 fungot, have you seen boily? 02:23:06 quintopia: this is where we are big now it was a nonsense statement. actually, the exception only needs to ,config ,load a packages.scm defining the corresponding structure, i think 02:23:19 ^styles 02:23:22 ^style 02:23:22 Available: agora alice c64 ct darwin discworld enron europarl ff7 fisher fungot homestuck ic irc* iwcs jargon lovecraft nethack oots pa qwantz sms speeches ss wp youtube 02:23:38 ^style homestuck 02:23:39 Selected style: homestuck (Homestuck pages 1901-4673) 02:26:23 quinthellopia! sorry, was watching a horrible sequel... 02:26:34 fungot: I am here. 02:26:34 boily: have a look. a trail of this fluid in the hall. you are going in. 02:26:42 fungot: what? no! 02:26:42 boily: it is a hate so pure and... pumpkin seeds? i sure as hell don't care what the consequences. they allude to other things as well? 02:27:04 fungot: I don't hate pumpkin seeds, and ain't no way I'm going in. ew. 02:27:05 boily: is it. perhaps this one. in a kid's yard. not even that sure why 02:27:19 fungot: don't ask me. you're creepy. 02:27:19 boily: in a good way though 02:27:35 fungot: eeeeh... please consider the expression of my doubts. 02:27:56 but now, time to appreciate the softness of my pillow. 02:28:04 -!- boily has quit (Quit: INDUSTRIAL CHICKEN). 02:30:04 -!- Phantom_Hoover has quit (Read error: Connection reset by peer). 02:55:48 -!- bb010g has joined. 03:05:27 I would first look into general language design. 03:05:32 Wait I'm scrolled up. 03:05:52 And also hppavilion talked to me? 03:08:11 And later left. 03:12:03 `olist 1014 03:12:07 olist 1014: shachaf oerjan Sgeo FireFly boily nortti b_jonas 03:14:12 shachaf, you're fast 03:14:42 Faster than IFTTT is at checking RSS and Twitter 03:18:43 -!- jaboja has quit (Ping timeout: 260 seconds). 03:20:25 There are many uses of neural nets, but for something like norns brains I would try to actually implement Babbage. 03:22:00 Since if the simulae brain is taken as being the cpu's emotion, one can use it to impliment the "might do" loop statement. 03:24:47 Or Else is such a nice error handling method. 03:27:01 -!- MDude has changed nick to MDream. 03:32:45 The ordinal numbers are keeping me awake. 03:33:12 I'm trying to decide what I think the smallest non-well-behaved ordinal number is. 03:33:19 And so this has me thinking about ordinal notations. 03:33:27 Wikipedia defines a certain ordinal notation. 03:33:40 And doesn't follow it? :P 03:33:56 0 is the smallest ordinal number. Z(a,b) is the smallest ordinal number which is greater than a and b, and which is greater than Z(c,d) for c < a or for c = a and b < d. 03:34:20 So, taken together, these definitions form an ordinal notation. 03:34:38 Let, uh... let E be the set of all ordinal numbers defined by this notation. 03:35:20 The notation provides a function (1 + E * E) -> E. A bijection, actually. 03:36:12 Now... 03:36:13 How is Z(a,b) > Z(a,d) if b < d? 03:36:39 Whoops. s/b < d/d < b/ 03:36:56 Ah. 03:37:10 And then the question is, how do you compare ordinal numbers defined in this notation? 03:37:35 Well, let's try to compare 0 and Z(a,b). Hey, guess what? Z(a,b) is guaranteed to be greater than something, which means that it's greater than 0. 03:37:50 Now, let's next try to compare Z(a,b) and Z(c,d). 03:38:58 The easy case: a = c, b < d. In this case, Z(c,d) = Z(a,d) is greater than Z(a,e) for e < d, so Z(a,d) is greater than Z(a,b). 03:39:05 And the hard case: a < c. 03:39:25 Supposedly, in this case, Z(a,b) < Z(c,d) if and only if b < Z(c,d). 03:39:26 Why? 03:39:48 Well, because... 03:40:16 Because, well... 03:40:21 Because a < c, so we can forget about it since Z(a,b) is greater than a,b? 03:40:52 Ah, there's my other mistake. 03:40:57 Okay, let me restate the definition entirely. 03:41:00 Heh. 03:41:20 0 is the smallest ordinal number. Z(a,b) is the smallest ordinal number which is greater than a and b, and which is *not equal to* Z(c,d) for c < a or for c = a and d < b. 03:43:06 Now, it's clear enough that Z(a,b) is increasing in b. 03:43:32 So the "easy case", comparing Z(a,b) and Z(c,d) for a = c, is still just as easy. 03:44:05 So now the hard case, a < c. In which, supposedly, Z(a,b) < Z(c,d) if and only if b < Z(c,d). 03:44:23 Ah, I think I'm starting to remember. 03:44:55 Z(c,d) is a fixed point of the function Z(a,-). How do we know this? It's because... um... 03:47:23 -!- hppavilion[1] has joined. 03:47:26 Something something normal function? 03:48:04 Let's see. Is the function Z(a,b) normal in b? 03:49:38 "For all ordinals a < b, f(a) < f(b)." Yup, we know that. "For every limit ordinal g, f(g) = sup {f(n) : n < g}." Um... I think so? 03:49:43 Let me chew on that a bit. 03:50:24 You know what, I think Z is *not* normal in b. 03:51:09 Because Z(0,omega) is omega + 1. 03:51:39 Babbage? 03:52:40 But how about for a > 0, then is it normal? 03:53:46 * tswett peeks at the answer. 03:54:34 Looks like the answer is: never. 03:55:29 Let's give the Veblen functions a try. 03:55:48 Well... wait, I don't want to do that yet. 03:57:23 Hello tswett. 03:57:27 Hey hppavilion[1]. 03:57:38 I'm now interested in Neural Networking 03:57:44 Particularly with PyBrain 03:58:35 I guess a crucial fact is that the function Z(0,b) "misses" "Ord-many" ordinal numbers. 03:58:51 After all, it's just b+1, so it misses all the limit ordinals. 03:59:07 Sorry if I seem a bit distracted. It's because I am. 03:59:22 Weird 04:01:03 So, the function Z(1,b) gives the smallest ordinal number which is greater than b and which is missed by Z(0,-). 04:01:16 But this function *also* misses Ord-many ordinal numbers. 04:01:42 How do we know this? Good question... 04:05:32 Well, suppose you take the function Z(1,-) and iterate it on your favorite ordinal number. You'll get an infinite increasing sequence. Consider the limit of this sequence... call it m. Z(1,m) has to be bigger than m, by definition. But for k < m, Z(1,k) is smaller than m... 04:06:13 ...because once you find a number in the sequence bigger than k, the following number (which is smaller than m) is bigger than Z(1,k). 04:06:37 So, for every ordinal number, there's a bigger number which Z(1,-) misses. 04:06:43 Z(1,-) misses Ord-many ordinal numbers. 04:07:12 In fact, I think this is true for every function f with the property that for all ordinal numbers x, f(x) > x. 04:07:26 tswett: What should I purpose my Neural Net to do... 04:08:00 It'll definitely be bluetoothkinetic- that is, able to sense the bluetooth in the surrounding environment 04:08:34 Well, that rules out everything unrelated to Bluetooth, right? 04:08:37 Teach it to hijack cars' speakers, and broadcast scary messages. 04:09:21 tswett: Well I want it to be more of a generic assistant and/or toy with special abilities than a single-purpose robot 04:09:54 Do you want to actually build it into a robot? 04:10:17 No! Of course not 04:10:22 Don't be ridiculous 04:10:23 xD 04:10:28 :( 04:10:41 So, like, this is totally software, no hardware. 04:10:42 I really just want a fun widget. 04:10:45 Yep 04:10:47 Just software 04:11:15 Hmmm. It'll be hard to do something that hasn't been done before. 04:11:28 Pretty much 04:12:05 You know what I've wanted to make a neural net do? 04:12:11 What? 04:12:13 Take an English word and predict the corresponding Spanish word. 04:12:19 Huh 04:12:24 That'd be interesting 04:12:37 It would, of course, come up with predictions for things that aren't actually English words, like "vonk". 04:12:40 There's french and german... http://104.131.78.120/ 04:12:42 I can't figure out how to make Pybrain accept variable-width strings 04:13:13 Is that because Pybrain feeds the entire string into the neural net all at once? 04:13:20 I'm not sure 04:13:28 tswett: You know what I want to do? Make a Neural Net that feels pain. 04:13:36 Then construct Neural Net hell. 04:13:58 Either that or make a small race of neural-net-based individuals and train them all to worship me as a god. 04:14:18 You'll have to figure out what it means for a neural net to feel pain or to worship. 04:14:23 In any case, theoretically, I should be sleeping right now. 04:14:31 The only reason I'm not sleeping is that I'm thinking about ordinal numbers. 04:14:42 So I really ought to either go to sleep, or continue thinking about ordinal numbers. 04:14:44 I choose the latter. 04:14:46 tswett: Feeling pain would just be punishment; put the neural net in a state and program it to avoid being in that state. 04:15:12 Excellent choice, monsoir. 04:15:16 Ah, that's easy. 04:17:02 tswett: And make it a big neural network so that the pain feels meaningful to us 04:20:00 Right, now. Z(0,-) misses Ord-many numbers, so Z(1,-) exists. It misses Ord-many numbers, so Z(2,-) exists. And so forth. 04:20:34 I'm going to temporarily forget about the question of whether or not Z(omega,-) exists. 04:20:37 (Spoiler: it does.) 04:24:04 Let's do a relatively concrete comparison, why not? 04:24:32 Which is bigger: Z(6,b) or Z(7,d)? 04:24:48 Supposedly that question reduces to this one: 04:24:53 Which is bigger: b or Z(7,d)? 04:25:48 And I guess that comes down to the fact that Z(7,d) can't lie in between b and Z(6,b). 04:25:52 Well, why not? 04:26:38 Well, suppose that b < Z(7,d). 04:28:58 Let's see. Z(7,d) is one of the numbers that's missed by Z(6,-). 04:29:33 Which means that Z(7,d) is not any of the elements in this sequence: b, Z(6,b), Z(6,Z(6,b)), Z(6,Z(6,Z(6,b))), ... 04:29:55 But could it be stuck in the middle somewhere? 04:31:35 Well... I think the answer is theoretically yes. 04:32:04 Like, if Z(6,x) were x*2, the answer would be yes. 04:32:33 The answer is in fact no, but we don't know this given the mere fact that Z(6,x) makes its argument bigger and misses Ord-many numbers. 04:32:39 And is strictly increasing. 05:01:20 -!- hppavilion[1] has quit (Ping timeout: 246 seconds). 06:00:26 -!- B1ood6od has joined. 06:33:11 -!- propumpkin has joined. 06:34:15 -!- copumpkin has quit (Ping timeout: 260 seconds). 06:39:10 [wiki] [[Seriously]] http://esolangs.org/w/index.php?diff=45746&oldid=45741 * 202.99.101.111 * (+10) oh, no! 07:24:12 -!- Elronnd has joined. 07:39:34 -!- Patashu has joined. 07:48:25 -!- hppavilion[1] has joined. 08:11:53 -!- mroman has joined. 08:12:00 fnird 08:12:09 @massages-low 08:12:09 Unknown command, try @list 08:12:13 @messages-low 08:12:13 oerjan said 11h 47m 36s ago: getVar name = do (_, _, v) <- get <-- depending on circumstances, just matching enough to ensure the resulting triple is not undefined might be too strict. does it help with a ~ before the ( ? 08:12:13 oerjan said 11h 47m 7s ago: mind you, that might sometimes leak memory instead. 08:17:28 Why do all the english documentaries sound like 50min long movie trailers? 08:18:48 Hm... 08:19:03 What cool and/or evil shit can I do with Neural Networking... 08:19:23 Hemroman! 08:19:35 ( mroman ) 08:19:36 (input):1:8: error: expected: "#", 08:19:36 "$", "&", "&&", "&&&", "*!>", 08:19:36 "*", "***", "*>", "*>|", "+", 08:19:36 "++", "+++", "-", "->", ".", 08:19:36 "/", "/=", ":+", ":-", "::",↵… 08:19:45 Crap. idris-bot. 08:34:11 -!- skarn has quit (Ping timeout: 260 seconds). 08:35:05 -!- ais523 has joined. 08:36:31 -!- skarn has joined. 08:37:55 -!- aloril has quit (Ping timeout: 260 seconds). 08:42:01 [wiki] [[Three Star Programmer]] http://esolangs.org/w/index.php?diff=45747&oldid=45719 * Ais523 * (+105) /* Implementation */ add a link to my C impl 08:43:04 hppavilion[1]: you still on? Thanks for the befunge-98 suggestion. Been working on it for a few hours, I'm cloning my Rubik's encryption in it. 08:43:07 [wiki] [[The Amnesiac From Minsk]] http://esolangs.org/w/index.php?diff=45748&oldid=45745 * Ais523 * (+1396) /* Implementing level 2 in level 1 */ looks like this part of the document got corrupted somehow (missing paragraphs and the wrong title for the wrong section); fixing that now 08:43:26 B1ood6od: I'm ALWAYS on. 08:44:05 [wiki] [[Language list]] http://esolangs.org/w/index.php?diff=45749&oldid=45742 * Ais523 * (+30) /* T */ +[[The Amnesiac From Minsk]] 08:44:32 -!- aloril has joined. 08:44:37 [wiki] [[Language list]] http://esolangs.org/w/index.php?diff=45750&oldid=45749 * Ais523 * (+18) /* L */ +[[Last ReSort]] 08:44:38 The way I'm doing this code though, just doesn't feel right. I feel like befunge needs to have shapes and look pretty, but so far its massive 200 line column, just to accept any ASCII character and recognize it. 08:45:41 [wiki] [[User:Ais523]] http://esolangs.org/w/index.php?diff=45751&oldid=45310 * Ais523 * (+46) +2; it's unclear what to do about sort order when you create two languages concurrently, each taking influence from the other... 08:45:54 anyone seen my recent languages, incidentally? 08:47:02 [wiki] [[OISC]] http://esolangs.org/w/index.php?diff=45752&oldid=45611 * Ais523 * (+50) /* See Also */ let's create a page the old fashioned way (i.e. creating a redlink to it first), for nostalgia's sake 08:47:19 [wiki] [[ZISC]] N http://esolangs.org/w/index.php?oldid=45753 * Ais523 * (+43) create redir 09:00:35 [wiki] [[OISC]] http://esolangs.org/w/index.php?diff=45754&oldid=45752 * Ais523 * (-6) change categories, this one seems to fit better 09:02:41 oh great 09:03:56 hi b_jonas 09:04:51 [wiki] [[Zero Instruction Set Computer]] N http://esolangs.org/w/index.php?oldid=45755 * Ais523 * (+3009) a new esoconcept 09:05:42 hello, ais523 09:05:53 have you (or anyone else here) looked at the languages I created recently? 09:06:01 ais523: yes, I've looked at them 09:06:21 any opinions about them? 09:07:11 ais523: there's that limited minsky machine thing, 09:07:12 [wiki] [[MATL]] http://esolangs.org/w/index.php?diff=45756&oldid=45743 * Luis Mendo * (+0) /* Compiler */ 09:07:18 b_jonas: right, that's one of them 09:07:31 I admit I'm still stuck on level 3 09:07:54 I don't really know how the lower levels work, but these kinds of limited Minsky machines can be useful for things like proving that M:tG is Turing-complete, 09:07:56 Three Star Programmer, and later The Amnesiac From Minsk, really reignited my interest in esolangs-as-puzzles 09:08:03 and yes, I noticed that too 09:08:09 although I really don't like the exponential slowdown (sometimes double exponential). 09:08:28 (or I probably wouldn't have thought of implementing The Amnesiac From Minsk in StackFlow) 09:08:46 I don't like the slowdown either 09:08:56 although you can partially work around it by writing really fast interpreters 09:08:57 As for three star programmer, the reference interpreter "http://nethack4.org/esolangs/threestar.c" was down recently and I wanted to look at it, but it's up now so I can take a look. 09:09:13 b_jonas: I actually just overwrote it with a new, much better interpreter 09:09:37 I decided to see how fast I could make it go, so I removed all memory allocation and just use the entire data segment as the memory array 09:09:43 oh, that one is scary 09:10:01 so it's not a reentrant or embeddable interpreter 09:10:48 indeed, I actually decided to avoid stdio because I wasn't quite sure it'd work 09:11:07 it beats having to use realloc, though 09:11:22 it's the most literal possible interpretation of "extending an array" 09:12:33 Couldn't you just use a separate mmapped segment for that? Then it wouldn't mess up anything else 09:13:15 you could try to place an mmap at a location in memory at which it could extend a long way 09:13:17 Also, is it possible that it fails to extend memory when pptr goes over the limit but no cell points after that? 09:13:29 but that's a) nonportable, and b) not necessarily going to work 09:13:39 <^^v> mmap <3 09:13:39 and pptr's confined to a small region at the start of the data segment 09:13:46 we store program first, then memory 09:13:55 and pptr just loops around and around the program 09:14:00 oh right 09:14:22 so the program is stored before the data memory, right? 09:14:26 right 09:14:44 -!- hppavilion[1] has quit (Ping timeout: 246 seconds). 09:14:56 where do you make sure that the memory is initialized to zero when allocating the initial part? 09:15:53 in the memory extender 09:16:02 it's used to allocate memory initially by extending it from 0 09:16:11 and it does a memset on the newly allocated area 09:16:36 um 09:16:46 why would you need memset of the newly allocated area? that's zero anyway 09:16:51 I'm asking about the original area 09:16:59 there is no original area 09:17:20 or, well, there's the program's original data segment but it isn't used except for the normal purposes of a data segment 09:17:39 oh! 09:17:40 to create memory we write to, we always use sbrk, and then zero the space between the old and new break 09:17:55 so you sbrk anyawy when you extend, even if the data segment already had some space after? 09:17:58 ok 09:18:03 makes sense 09:18:07 err, I don't think you understand what sbrk does 09:18:23 sbrk moves the end of the data segment, causing everything between the old and new break to become/cease to be data segment 09:18:39 "space after the data segment" doesn't make sense 09:18:52 as far as I can see, sbrk is a special case of mmap/munmap/mremap, which modifes the end of the data segment and returns the old end of it 09:19:06 it exists only for historical reasons 09:20:24 from back when there was no memory mapping, paging, or os-configurable segments, and all the cpu mmu did was to enforce that user mode can't access addresses higher than some limite 09:20:50 the difference is that memory maps created by mmap try to avoid the area that the data segment's expected to expand into, but don't try to avoid each other 09:21:03 and mremapping the data segment is a good way to cause bad things to happen 09:21:12 so it set the notional break between the data segment and the stack segment, where the stack segment grew down from the top of the user area 09:22:16 ais523: even if you mremap the data segment without changing the address? mremap can do that, it isn't only for changing the base address, it's also for changing the size. 09:22:50 ooh, I was assuming that MREMAP_MAYMOVE couldn't be turned off but it can 09:23:17 even then, isn't it much easier to call sbrk, than attempt to find the data segment (or an appropriate place to put a data-segment-alike) and mremap that? 09:23:31 ais523: yes, certainly 09:23:33 also mremap's Linux-specific 09:23:44 sbrk is easier 09:23:48 but you don't actually need mremap 09:23:57 whereas that code runs in the union of BSD and POSIX (and the POSIX dependency is only used for error messages) 09:23:58 you can use munmap + mmap 09:24:08 err, wouldn't that deallocate the data? 09:24:15 unless it was backed by a file 09:24:22 which has its own issues 09:24:34 yes, you'd need to back it by a file or by a "posix shared memory segment" (whcih is just a fancy libc wrapper for a file on linux) 09:24:40 so yes, sbrk is probably easier 09:24:45 for the data segment 09:25:19 ais523: ok wait, but 09:25:56 instead of taking over all the data segment, parts of which could still somehow be used by the compiler or compiler libraries or perhaps even libc, could you instead allocate the initial array with sbrk, and 09:26:07 then later at each sbrk, assert that nobody called sbrk between your calls? 09:26:50 that would be much more portable, because then libc (including the runtime, signal handlers, etc) or libgcc could still use areas of the data segment that it's already allocated, 09:26:55 it just wouldn't be allowed to allocate more. 09:27:29 b_jonas: no, that's exactly what I do 09:27:41 including the assert 09:27:45 great! 09:27:46 although I guess you could remove the assert for performance 09:27:52 meh 09:27:55 I don't clear the existing bit of the data segment 09:28:01 the assert doesn't really slow down anything 09:28:04 given that that's a good way to cause your constants to start not existing 09:28:14 and indeed, not with the current code structure 09:28:32 the assert got inlined into the middle of a loop with some previous code and broke code locality as a result 09:28:34 ok, good 09:28:57 but the nested-loop style of writing the program persuaded gcc to not inline the allocation, which makes sense given how rarely it runs 09:29:12 -!- ^v has quit (Read error: Connection reset by peer). 09:29:25 you could tell that to the compiler explicitly 09:29:37 that the condition is rarely true that is 09:29:40 oh well 09:30:35 it's funny that the program doesn't actually have three consecutive stars by the way 09:30:53 you could try to store pointers instead of integers maybe 09:31:17 I know, storing pointers would be more efficient on many architectures 09:31:36 that depends 09:31:38 but x86 (-32 and -64 but not -16) has this really nifty addressing mode that can do the pointer deref in one instruction 09:31:55 it depends on whether you want to restrict the integers to 32 bit 09:32:01 if you want 32 bit, you can't use pointers 09:32:07 (in x86_64 that is) 09:32:16 if you want 64 bit integers, then you can. 09:32:35 but 32 bit integers would be much more efficient if you're sure (and assert) the memory never grows above that 09:33:36 [wiki] [[Last ReSort]] http://esolangs.org/w/index.php?diff=45757&oldid=45744 * Ais523 * (+138) mention the influence of Malbolge 09:35:36 I'm sort-of hoping Last ReSort is TC, because its main loop is one byte shorter than MiniMAX's and thus would set a new record 09:35:56 although it's not as elegant because MiniMAX is capable of zeroing its own memory as memory expands, and Last ReSort isn't 09:38:15 um, one byte shorter in what what? 09:38:24 like, some cpu implementation or something? 09:38:34 and which cpu among that? 09:40:03 oh well, at that point I probably don't much care about how short it is, as long as it's reasonably short. 09:40:34 -!- ^^v has quit (Ping timeout: 245 seconds). 09:40:52 in x86 family 09:41:12 "x86 family"? but x86_32 and x86_64 are different 09:41:16 for these purposes 09:41:39 b_jonas: yes, I said that intentionally, because MiniMAX used the 8086 (which is actually relevant in this case) 09:42:00 whereas Last ReSort's ZISC I was using x86_64, although it easily ports to other 86alikes 09:42:01 the 8086 as opposed to what? 09:42:21 "easily ports" sure, it's a freaking C program afterall, but the number of bytes changes a lot 09:42:43 b_jonas: as opposed to the 80386 09:42:54 ... 09:42:57 and no, I'm talking about MiniMAX/Last ReSort 09:42:58 how does that matter for this? 09:43:01 they were originally written in machine code 09:43:11 8 bytes versus 7 for the main loop 09:43:30 hmm... let me think what the differences are that make 8086 better (as opposed to worse) 09:43:45 I think push sp or pop sp work differently, 09:43:48 the #1 biggest advantage is a default 16 bit data size 09:43:51 which may help 09:44:05 meaning your constants use up less space 09:44:10 um, you can still do 16 bit data size on a 386 09:44:27 all x86 cpus can do that, except in long mode 09:44:35 push sp or pop sp might help actually. 09:44:42 b_jonas: you can, but I said "default" 09:44:54 they all start up in 16 bit real mode, don't they? 09:44:57 you need to drop extra 66s and 67s into the program 09:45:03 also I consider that to be an 8086 emulator 09:45:12 rather than the "proper" 80386 instruction set 09:45:16 what? 09:45:26 the 80386 adds a lot of new features to 16 bit real mode 09:45:43 lots of new instructions, including 32 bit arithmetic, in a legal and documented way 09:45:52 it's not just a "8086 emulator" 09:46:01 [wiki] [[MATL]] http://esolangs.org/w/index.php?diff=45758&oldid=45756 * Luis Mendo * (+261) /* Specification */ 09:46:13 plus the 286 adds 16-bit protected mode 09:46:39 well, let's put it this way: MiniMAX runs in real mode, Last ReSort runs in protected mode 09:46:44 ok 09:47:06 but really, does the different meaning of POP SP perhaps help in one of those variants, for one-byte dereferencing? 09:47:34 let me look it up how it works and when it's changed... hmm, where's that documented 09:47:50 it must be somewhere in the intel manual probably 09:48:12 [wiki] [[MATL]] http://esolangs.org/w/index.php?diff=45759&oldid=45758 * Luis Mendo * (-350) /* Specification */ 09:48:17 unless they decided you should just check that bit in the flags to detect the cpu 09:49:09 POP SP sounds like the sort of instruction that's generally basically useless, but might be really helpful for an esolang 09:49:15 what does it actually do? 09:49:29 ais523: I don't know if it's push sp or pop sp 09:49:47 one of them differs somewhere between the 8086 and the 80386, but I'm not sure at which stepping and which direction, 09:49:58 in whether they decrement/increment first and load/store next or backwards 09:50:03 oh, it's not quite the Last ReSort ZISC all by itself, I was hoping it would be 09:50:10 because it's sp, not *sp, that gets incremented 09:50:20 and iirc it was the recommended way to detect the difference between two of those cpus, 09:50:32 although these days you can just shortcut it by detecting a later cpu directly 09:51:34 two of the steppings use a bit in the flags register for detection 09:52:11 (one is somewhere between 8086 and 80386, the other is between pentium with cpuinfo and pentium without cpuinfo, almost everything after pentium with cpuinfo just uses bits in cpuinfo for detection) 09:52:17 (almost) 09:52:40 (there's more steppings between these and I don't recall what they use 09:52:41 ) 09:53:26 But these days the intel manual doesn't even bother to document those old cpus, so I'm not sure that part is documented. 09:53:51 Just learned not to program at 4am. I had a 200+ line if-else statement doing something I just did in 5 lines, oops. 09:54:43 It only documents everything starting from the 386. 09:55:01 hmm no, it actually documents a bit earlier too, partly. 09:55:56 POP SP might be useful for removing a stack frame. I don't the ENTER and LEAVE were available before the 286. (I might be wrong) 09:55:58 wait, there are more bits in EFLAGS for detection than I thought 09:56:03 (some of them are also used for other things) 09:56:10 three or four bits 09:56:12 interesting 09:57:09 aha, here it is 09:57:12 it does document that 09:58:45 it says that PUSH SP differs, namely 286 and later push the old value, the 8086 pushes the new value to the stack 09:59:04 (it pretends the 80186 doesn't exist, which is mostly true) 09:59:59 bit it also gives another way to distinguish between the 8086 and the 286, so I think one of those is above and one below the 80186. 10:00:10 It doesn't say anything about POP SP, but you can probably find that info somewhere else. 10:01:22 This is chapter 22.17 and 22.16 in the intel manual, and it also tells how to distinguish between 386 and 486 etc 10:02:20 -!- ais523 has quit (Ping timeout: 246 seconds). 10:02:44 But very few people use real 8086 these days, because the 80386 has obsolated it. 10:09:57 8086 didn't have leave 10:11:05 also push bp wastes valuable stack space 10:11:20