00:00:00 --- log: started forth/02.11.26 00:01:39 --- join: onetom (~tom@novtan.bio.u-szeged.hu) joined #forth 00:08:26 re Fractal 00:08:35 Yo. 00:08:58 so whadda we talked abt ? 00:09:25 Dunno. 00:09:50 Ya, I don't know... Whatever. 00:10:19 i had a sleepless night so my brain jams... 00:10:46 i now came up w/ effect 4r my noter in forth 00:11:03 ?? 00:11:37 'noter'= cool text demo w/ editor, can save self w/ editetd text 00:12:31 Oh. For what forth? 00:17:03 --- join: lament (~lament@h24-78-145-92.vc.shawcable.net) joined #forth 00:17:15 Hey lament. 00:18:15 How's the weather in Van now? 00:19:53 it's reasonable :) 00:20:07 What? Not raining? I don't believe it! :) 00:20:19 well, it rained for about a week 00:20:27 Ah, there's the vancouver I know. 00:20:29 but stopped finally 00:20:58 I want to move to Van for a few months. Great city. 00:23:11 You should've came this summer. There was no rain at all. 00:24:25 Ya? Cool. 00:25:24 I don't think it'll ever happen again, though :) 00:25:26 DOS, EGA mess w/ planes - see if GF2Pro can hold it 00:25:36 gpforth, www.forth.ru 00:25:59 lament : Well, you never know... If Alberta gets it's way, we might never have rain! 00:49:04 damn, gforth 'see' is fricking useless 00:49:41 I thought it'd be easy to write tail recursion for gforth; i was wrong 00:49:47 Yes, I know. I believe it keeps a text buffer of the actual string used to create the word. 00:50:30 ppforth see is fairly cool - it prints the memory address and the actual value at address for each instruction 00:51:01 Yes, same with frugal's see. 00:51:40 Yes, I find gforth to be too complex. 00:52:29 Yes, I have to stop prefixing everything with "Yes, ". 00:53:57 No, why should you? 00:54:08 Heh. 00:57:44 hm, and isforth doesn't seem to have 'see' at all 00:58:01 Oh really? Complain to i440r. :) 01:09:51 it has but really wrecked one 01:10:39 it does? Where? 01:14:14 in a pre 1.10 version he gave me, in separate src 01:14:24 hmm 01:35:21 z 01:37:06 isforth also fucks up the terminal :( 01:42:28 Yes, I'm not a huge fan of isforth, really. 01:42:46 Particularly because I can't run it on most of my computers. 01:45:43 you're probably a huge fan of frugal? :) 01:45:52 it fucks term down not up 01:46:06 Well, sort of. If I had to start over, I'd do a few things differently though. 01:46:25 if try to say damn politely it needz scratch rewrite 01:46:44 Actually, my favorite forth is PocketForth for old, old MacOS's 01:47:06 heh 01:47:17 Back then it sure beat the hell out of programming in pascal and C for MacOS... 01:47:50 after isForth i type ( in RH ) butt russian letters go hell after it.... 01:47:51 It was especially great because you could run ROM routines, debug most of MacOS, whole bunches of things you couldn't do with C. 01:48:11 Serg_Penguin : Heh. Why don't you like isforth? 01:48:47 Serg_Penguin: so do english ones, so it's OK 01:49:05 Serg_Penguin : Just because of that? 01:50:40 too much fat and bugz living in it 01:50:55 Yes, lots of fat. 01:50:56 too much fat? in isforth? 01:51:18 lament : Take a look at the .f source. :) 01:51:23 And the .asm 01:51:39 i think creating forth is great creative fun, but releasing buggy stuff is no good 01:51:54 Fractal: well, at least it builds itself in less than a second 01:52:13 builds itself ? it's in NASM ! 01:52:41 well, builds most if itself 01:52:45 s/if/of 01:52:54 lament : Well, it doesn't exactly build itself any more than any other forth system does, but that's not the issue. There's too much complexity and no documentation. 01:53:30 vocabularies are fat begging for axe surgery 01:53:48 Yes, I don't believe in vocabularies either. 01:54:03 Even though they are trivial to implement. 01:54:55 they may be option, but if kernel needz them to be implemented - such kernel needs axe surgery 01:57:00 Isforth source breakdown: 188K asm source, 152K forth source, 11K of ranting, 0K of docs. :) 01:57:25 heh. 01:57:59 anyway we have no beter one 01:58:02 However, I do have moderate respect for i440r: I've had some good conversations with him. 01:58:29 no better one 4r linux 01:58:35 Serg_Penguin : Well, on the contrary, there are dozens of forths for unix/linux. 01:58:52 pForth is a decent one. 01:58:58 gforth is OK. 01:59:09 bigforth is OK 01:59:18 bah, i'll just get lpforth 01:59:30 4th is strange, but OK. 01:59:40 lament : Heh. 01:59:42 lpforth is ppforth for linux 01:59:52 Oh. 02:00:00 I thought you were making a joke. 02:01:02 Serg_Penguin : In fact, do to it's complete lack of portability, I'd have to recommend not using isforth for your projects. 02:01:17 so at least my tail-recursion routine works without any adjustments :) 02:01:58 Not only will isforth itself not run on any other architecuture or operating systems, it's also highly unlikely your source isforth programs will work on other forth systems without a lot of work, do to isforths large array of "extensions". 02:02:27 lament :) 02:02:37 oh, actually it doesn't. Nevermind ;) 02:02:58 lament : For a decently documented forth, it should be relatively easy to write a tail recursive forth... 02:03:06 lament : lol 02:03:33 Fractal: A decently documented forth? You must be smoking something :) 02:05:27 lament : Well, an adequatley documented forth. :) 02:08:12 Such as? 02:09:08 None pop to mind. 02:09:39 hm, bigforth seems really fucked up 02:10:13 Ya? 02:10:53 well, perhaps MINOS rather than bigforth itself 02:11:18 Oh yeah... I've only played with bigforth a bit... I think it was on windows, actually. 02:11:29 With a "rapid GUI developement editor" that generates object-oriented forth code 02:11:55 OK. 02:12:17 CM can be heard spinning in his grave, and he's not even dead yet 02:12:47 http://www.jwdt.com/~paysan/theseus.png 02:12:59 No picture viewer. 02:14:40 well, it looks pretty much like Microsoft Visual Forth would look :) 02:15:09 Heh. 02:18:19 so what tiny usef00l forth 4r linux will u advice ? 02:24:11 Apparently, someone still has to write one. 02:35:10 Fucking gforth... 02:35:25 I'm trying to figure out how to do tail recursion in it, but it's needlessly complex. 02:36:14 yup, it is 02:36:38 i also tried 2 understand it but i gave it up 02:38:20 mostly because of lack of proper see :) 02:39:47 Well, I don't really mind the see that much... It's just every single gforth word is defined in terms of 300 other gforth words that you have no idea about. 02:40:12 Serge: eforth perhaps 02:40:47 lpforth is eforth, pretty much 02:41:40 http://206.171.116.227/forth/lpforth.html 02:42:05 : immediate immediate-mask lastflags cset ; : lastflags Last @ dup 0= ABORT" last word was headerless" cell+ ; 02:42:15 ^ gotta love that - for setting 1 bit too. 02:42:18 heh 02:43:08 see headerless" <- undefined word 02:43:20 The problem with forth is that you can define anything in terms of something else, and never get to an actual definition :) 02:43:22 So they've unlinked system words. Lovely. 02:43:32 does efort have even fsave ? 02:43:39 lpforth does 02:44:08 -> lament lpforth does have ? 02:44:51 yes 02:46:00 hm, the executable lpforth is really big for some reason, though 02:46:06 110k 02:46:17 the palm version is 30k 02:52:04 lament : Did you strip it? 02:52:18 Either way, that's still too large. 02:52:48 well, you can just really use the kernel if you really want it 02:52:59 yay 02:53:02 i stripped it 02:53:07 it's 5.8k now 02:53:10 :) 02:53:21 oh, and doesn't work 02:53:24 hehe 02:53:30 Heh. 02:54:51 OK, I've added >head and head> now in frugal. Tail recursion is now done like so : ;recurse l @ head> branch , ; immediate compile-only 02:55:10 oh, there's also an actual eforth for linux, by Fare 02:55:16 Think those 2 changes are enough for a version release? 02:56:54 Not sure why it wasn't that easy in gforth, but whatever... 03:20:01 z 03:22:53 q 03:25:19 OK, I've added docs for the locals wordset too now. 03:25:32 What other things need to be done before a release? 03:29:25 knok lokals out of kernel, make them src-compile option 03:30:12 3 ideal Forth should havle only the features needed to extend it 03:30:21 all others - options 03:30:55 Serg_Penguin : Locals aren't in kernel, they're in progs/locals.fs 03:31:20 If you want to use them, include that file. 03:31:33 it's OK 03:31:49 * Serg_Penguin will rather define extra stacks 03:32:06 * Serg_Penguin did it 4r gpforth 03:39:56 Yes, extra stacks can be useful. 03:40:28 like matrix and vector ones if u do 3d 03:40:32 I have to admit, locals rule :) 03:40:58 maybe but not 4me 03:41:20 if i feel i ned them - i _really_ need 2 factorize 03:42:01 In frugal, locals work like so: : my-word { a b c } ... ; 03:42:16 * Serg_Penguin now writes in one-liners 03:42:17 And where ... is, you can use a b and c as if they were "normal" variables. 03:42:25 Fractal: add comments! 03:42:36 anyway, i consider lcals no good 03:42:38 lament : Next version has docs. :) 03:42:45 i mean 03:42:47 : my-word { a b c -- comments } ... ; 03:43:13 lament : Oh, well maybe... 03:43:28 I like how mine is. Just use ( ) if you want comments. :) 03:43:51 { a b c ( -- comments)} ? 03:43:59 OK, I'm just going to finish this tail-recursive factorial code for an example in docs/USAGE and then I'll release. 03:44:02 lament : No! 03:44:08 lament : No, outside the { } 03:44:11 heh 03:44:18 Because that space is parsed by the { word. :) 03:44:28 Not "interpret" 03:51:10 OK, frugal-0.9.9 is out: http://www.hcsw.org/frugal/ 03:52:07 Featuring tail-recursive words (with a comparison of tail and non-tail recursive factorial), improved docs in several spots, and added docs for locals wordset. 03:52:37 stupid tar 03:52:37 time stamp 2002-11-25 21:53:14 is 7500 s in the future 03:52:52 make: warning: Clock skew detected. Your build may be incomplete. 03:52:53 heh 03:53:02 at least it works. 03:53:45 Heh. You're probably 1 hour behindme. 03:54:09 Er... Look at your clock. :) 03:54:44 The tail-recursive stuff is under the "RECURSION" section of docs/USING 03:55:07 hm, i got a different definition for factorial 03:55:11 : facrec dup if dup rot * swap 1 - tail then drop ; 03:55:12 : factorial ( n -- fac_n ) 1 swap facrec ; 03:55:38 Oh, ok... 03:55:43 Did you write that? 03:56:00 Because that's very close to mine. :) 03:56:01 yes 03:56:01 it's not for frugal, though 03:56:09 yours is a bit different 03:56:19 Just swap tail with ;recurse and it should work. 03:57:02 Did you write that in ppforth? 03:57:11 yes 03:57:13 pp/lp 03:57:13 Cool. 03:57:21 You got tail to work in lp? 03:57:31 yes 03:58:19 By guessing the header offset :) 03:58:36 Heh. Gotta love hacks. :) 03:58:56 Ya, I tested yours. Works great - as expected. 04:00:32 hm, factorials behave very funny with numbers of limited size 04:00:48 30 factorial . 1409286144 ok 04:00:48 31 factorial . 738197504 ok 04:00:48 32 factorial . -2147483648 ok 04:00:48 33 factorial . -2147483648 ok 04:00:48 34 factorial . 0 ok 04:00:54 all the following ones are also 0 04:01:03 neat :) 04:01:24 Heh. 04:01:40 Well, that's what happens when you wrap integers. 04:01:58 One problem with yours though: -1 factorial = bottom 04:02:04 Infinite loop. 04:02:16 not infinite, because of wraparound 04:02:21 :P 04:02:25 Yes, true. :) 04:02:42 interesting, so the factorial gets its 32th 2 at 34 04:02:46 weird 04:03:15 Well, lesse. 2^32 = 04:03:24 4294967296 04:03:53 If you don't include negs that is. 04:03:59 So your max val is: 04:04:13 2147483647 04:04:59 So, what do you think of docs/LOCALS ? 04:05:05 Explain it well enough? 04:07:07 yes. 04:07:10 Also, you should check out docs/DEBUG - I'm particularly happy with debug.fs 04:07:29 Cool. 04:09:41 i tryed factorials in lisp - pages of digits and stack overflow 04:09:59 You probably (mis)implemented it using general recursion then. 04:10:30 i just typed in tutor sample 04:10:36 i dunno write in lisp 04:10:40 Ah yes, cool. 04:10:50 Python doesn't have tail recursion :( 04:10:54 Lisp is a decent language really. 04:11:24 lament : Really? That's odd... Oh well, you can just hack tail recursion the same way as C, then. 04:11:35 what do you mean? 04:11:54 lament : Well, essentially put a goto to the top of the procedure. :) 04:12:10 And just modify your passed variables. 04:12:26 Usually you would use a loop instead of a goto, but you get the idea. 04:13:27 Factorial in python: fac=lambda n,f=lambda f,n:n and n*f(f,n-1)or 1:f(f,n) 04:13:34 (just kidding) 04:13:52 That or tail recursion is implicitly optimized for in python as it is in haskell. 04:14:25 You can write some rather neat functional obfuscated code in Python 04:14:41 Cool... 04:15:30 What do you think of frugal's see? 04:15:45 see see 04:15:45 UNSEEABLE 04:15:45 ok 04:15:45 :) 04:16:25 Well, that's the one drawback of my see: You can't see words with embedded character strings. :) 04:16:51 why not? 04:17:17 hm, frugal's weird 04:17:20 OK, do you understand how the forth s" works? 04:17:43 probably 04:17:47 Yes, it's odd. :) 04:17:59 : foo if 1 else 2 then ; 04:18:11 see foo 04:18:11 134563274 : 0= 04:18:11 134563278 : VM_IFBRANCH -> 134563298 04:18:14 (and so on) 04:18:25 it's as if frugal's desperately trying to be high-level 04:18:55 lament : OK, well rather than risk a dereference of potential character data, and not knowing how long an embedded string is, we just make the word unseeable. 04:19:10 What do you mean? 04:19:31 hm, but you don't need to dereference character data! 04:19:46 OK, consider how s" works: 04:20:33 It compiles a jump to here + length_of_string. 04:20:42 Then it compiles the string right into the dict. 04:21:20 I could make a VM_STRING token, I suppose. 04:21:29 hm, true. 04:21:42 However, how am I going to know how long the string data is? 04:22:08 Because somebody could compile there own string-creation word with ," 04:22:15 Their. 04:22:32 Well, I just figured it's easier, for now, to make such words unseeable. 04:23:00 ppforth simply prints out everything 04:23:36 Well, I like the way of ordering the see with the addresses next to it. 04:23:49 yes 04:24:03 And seeing BRANCH -> some_addr 04:24:55 debug.fs uses similar output, BTW. 04:26:30 I also, if I may, think that my terminal handling routines are much, much, much better than isforth's. 04:26:36 All of them start with term- 04:26:44 And are relatively self explanitory. 04:27:25 term-cls clears the screen, for instance. 04:28:27 by the way, have you done any forth/haskell/C speed comparison? 04:28:39 10 10 term-xy locates the cursor to (10,10, etc. 04:29:12 lament : No, not particularly, but I'd say it goes: C, forth, haskell 04:29:31 That's assuming the standard compilers. 04:29:41 gcc is simply a good optimizer. 04:30:02 If you're interested, check out "The Great Computer Language Shootout" 04:30:08 You should find it with a quick google. 04:30:39 He compared a bunch of languages with various test programs doing certain things, like computing factorials, quicksorting files, etc, etc. 04:30:52 thanks 04:30:59 It's really handy if you want to get quick tips for a language you haven't learnt. :) 04:31:09 He's got probably 40 compilers up there. 04:31:44 hm, his tests are pretty weird 04:31:49 or rather his results 04:33:34 Seems pretty decent to me, although he doesn't program functionally very well - sometimes using IOrefs and other such nonesense, so poor ol' GHC doesn't do too hot. 04:35:25 ocaml is consistently good 04:35:29 interesting 04:35:59 gforth is consistently bad 04:36:29 Yes, well unless you really work at forth, it's difficult to write really decent code. 04:36:45 yay, a hello world test 04:36:57 -since most forth compilers do little to no optimization. 04:37:01 lament : Heh. Ya. 04:37:07 java dies on that one 04:37:41 Heh. 04:40:18 heh 04:40:25 http://www.bagley.org/~doug/shootout/bench/random/ 04:40:57 either his haskell algorithm sucks, or there's something weird there 04:43:43 Hm, well what're you going to do? :) 04:44:12 Well, in any case, I gotta get going. 04:44:14 Night. 04:44:35 night. 04:53:53 --- quit: lament (forward.freenode.net irc.freenode.net) 04:53:53 --- quit: onetom (forward.freenode.net irc.freenode.net) 04:53:53 --- quit: TreyB (forward.freenode.net irc.freenode.net) 04:53:53 --- quit: OrngeTide (forward.freenode.net irc.freenode.net) 04:54:55 --- join: lament (~lament@h24-78-145-92.vc.shawcable.net) joined #forth 04:54:55 --- join: onetom (~tom@novtan.bio.u-szeged.hu) joined #forth 04:54:55 --- join: TreyB (~trey@cpe-66-87-192-27.tx.sprintbbd.net) joined #forth 04:54:55 --- join: OrngeTide (orange@65.19.141.250) joined #forth 05:31:43 --- quit: Soap` ("Sleep!") 05:53:04 --- quit: Serg_Penguin () 06:07:28 --- quit: onetom (Read error: 54 (Connection reset by peer)) 06:07:35 --- join: onetom (~tom@160.114.27.131) joined #forth 06:21:32 --- quit: onetom (Read error: 104 (Connection reset by peer)) 06:23:19 --- join: Herkamire (~jason@wsip68-15-54-54.ri.ri.cox.net) joined #forth 06:23:34 mornin forthers 06:24:16 --- join: onetom (~tom@novtan.bio.u-szeged.hu) joined #forth 06:33:53 Hi. 06:47:53 --- join: Serg_Penguin (~Z@nat-ch1.nat.comex.ru) joined #forth 06:48:06 re 06:48:17 mornin :) 06:48:32 was it u working on enth ? 06:49:14 Herk, when can I try the amazing PPC colorforth? 06:50:34 btw, what assembler are you using for this? 06:54:48 Serg_Penguin: not I :) 06:55:25 XeF4: it uses gcc to compile the core. 06:55:42 Herk: oh, what OS? 06:55:43 XeF4: you can try it now if you're boored. it doesn't do much yet. 06:55:48 XeF4: linux 06:56:20 no ppc linux yet, are there any single-floppy versions around? 06:56:27 s/versions/dists/ 06:56:30 XeF4: it uses linux syscalls for the io and exit 06:56:47 Using a few simple Forth primitives instead of assembly, i.e. compiling "drop" directly to "add esi,byte 4" or whatever, is that a good idea? 06:57:39 is there some other way that I'm missing? :) 06:59:00 fpos will stand alone as an OS eventually 06:59:22 don't hold your breath though. I'd guess a year or two 06:59:54 Robert: depends who you ask. it certainly makes it more complex and larger compiled code. what I440r is doing is pretty fast 06:59:56 Herk: do you know of any single-floppy PPC Linux dists? 07:00:33 XeF4: no. but I haven't looked. I don't have a floppy drive :) 07:01:23 I wouldn't say "certainly" makes larger code. add esi,4 is 3 bytes; a 32-bit XT is 4 07:01:40 and pop edx (if you use esp for the data stack pointer) is 1 byte 07:03:02 XeF4: true. :) 07:04:46 Robert: I think it really comes down to whether you would like your compiled definitions to be native code, or addresses. 07:05:37 with addresses it's much simpler, and I was quite suprised how fast isforth is. 07:06:16 I was quite surprised too 07:06:51 -> robert u say good thing 07:07:02 usually it is only slightly slower than bigforth (which is native coded) 07:07:50 except for number crunching, since bigforth optimizes arithmetic fairly well 07:08:13 but it should be different language, maybe macro asm, borrowing main ideas from Forth 07:09:01 1) take asm written in forth 07:09:20 2) define macros like @ ! dup drop !a+ etc... 07:09:57 Serg: bigforth is written like that 07:10:09 3) add forth'ish control structures so one can write asm-fast proggie w/o any grasp of CPU guts 07:10:21 last week somebody wrote something (that dunder primes thing) in isforth, then ported it to asm, and found that the isforth version only took about %40 longer 07:10:43 sounds like an unskilled ASM coder 07:10:57 or braindead algorithm 07:11:56 XeF4: neither version was optomized I think. my point is that there isn't that much overhead in isforth vs native code 07:13:45 hard to say without seeing the code 07:14:26 if isforth doesn't cause eg.. a lot of branch mispredictions and some really horrid asm code does, that can account for some huge performance losses 07:22:41 --- join: tathi (~josh@wsip68-15-54-54.ri.ri.cox.net) joined #forth 07:48:04 --- quit: onetom (Read error: 110 (Connection timed out)) 07:56:11 --- quit: Serg_Penguin () 07:56:34 --- quit: proteusguy ("Client Exiting") 08:00:13 --- join: proteusguy (~username@65.191.88.177) joined #forth 08:14:47 --- quit: lament ("mental mantle") 08:22:29 Does anyone know how IsForth does a jump to the beginning of the current word? 08:23:49 um...it doesn't? 'cause I440r doesn't believe in recursion? 08:23:57 a jump or a call? 08:23:58 Robert: I think it doesn't. I440r is predjudiced against tail recursion. (that or he doesn't get it) 08:24:05 Hrm. 08:24:17 Bleh. 08:24:25 yep. 08:24:43 I've tried to explain the concept to him a few times, but I'm not sure he gets it. 08:25:44 I have to say, I'm highly amused by the comment in one of the isforth sources that says he's making the stacks "very small so as to discourage the use of evil recursion" 08:25:46 How should you use it? I just wanted to play with it as an alternate variant of ordinary loops. 08:25:51 but they're like 1000 cells deep 08:26:02 tathi: :) 08:26:18 Robert: that's it :) 08:26:21 Robert: well, there is goto 08:26:30 then on the other hand we have Chuck saying that the stacks on one of his chips are 18 cells deep, which was a number chosen so as to be practically infinite :) 08:26:45 Robert: tail recursion can be used just like : foo begin .... while ... repeat ; 08:27:06 Don't the lates chips have like 6 cells deep parameter stacks? 08:27:10 later* 08:27:15 : foo ... if ... exit then ... tail-recurse ; 08:27:25 Herkamire: Yeah.. 08:27:32 dunno, think that was an article on ultratechnology.com talking about F21... 08:27:42 colon 'recurse', recurse 08:27:42 dd last, fetch ;foot in self shoot 08:27:42 dd nameto, fetch 08:27:42 dd comma 08:27:42 dd exit 08:27:44 Heh. 08:27:55 Robert: not as far as I know.. 08:28:08 the latest to have been implemented in silicon have 18-deep 08:28:33 or, if you put your definitions in the dictionary right away, and ; optomizes the last instruction from a call to a branch: : foo ... if ... exit then ... foo ; 08:29:15 XeF4: Ah, OK. Maybe it was the varpourware one, er..X18? 08:29:22 : foo ... if ... foo then ... ; 08:29:33 I thought that still had 18-deep, but I'm not certain 08:30:54 : spaces dup 0= if exit then 1 - 32 emit spaces ; 08:30:56 besides, 8 deep is the minimum I can stand to use 08:31:32 Who said you're supposed to stand Forth chips? ;) 08:31:49 : spaces dup if dup 0 do 32 emit loop then drop ; 08:32:07 Robert: hehe :) 08:32:10 I have to agree on that, a 4 cell deep return stack and a 6 cell deep parameter stack is too little for many tasks. 08:32:21 Like...large, factored programs. 08:32:47