00:00:00 --- log: started forth/02.05.17 00:37:55 --- quit: onetom ("ircII EPIC4pre2.508 -- Accept no limitations") 00:38:03 --- join: onetom (tom@adsl52033.vnet.hu) joined #forth 01:10:58 --- quit: air ("AIRC v0.2.0pre -- http://www.qzx.com/airc") 01:11:09 --- join: air (brand@12-254-199-50.client.attbi.com) joined #forth 01:21:56 --- join: davidw (~davidw@adsl-243-109.38-151.net24.it) joined #forth 01:46:34 --- quit: onetom (Read error: 110 (Connection timed out)) 01:47:05 --- quit: gforth (Read error: 110 (Connection timed out)) 01:53:28 --- join: gforth (guest@adsl52033.vnet.hu) joined #forth 02:38:58 --- quit: gforth (Read error: 104 (Connection reset by peer)) 02:49:17 --- join: onetom (tom@adsl52033.vnet.hu) joined #forth 02:57:16 onetom: what's the magic recipe/paradigm to make a forth word look at the *next* word(s), and not what is on the stack? 02:57:17 like 02:57:19 ' + 03:00:07 sec 03:00:44 "what is on the stack"? :) 03:01:10 anyway, what would u like 2 do? 03:01:19 get the name of the next word? 03:02:23 sure 03:02:29 or something like that... 03:06:45 hold on 03:07:25 : x name type ; x asd asd ok 03:08:56 interesting 03:09:21 no, its not. its simply THE WAY ;) 03:09:24 thanks 03:11:06 but u can use ' and CREATE inside colon defs 03:11:23 coz they r not immediate so this 03:11:43 ' you can't use inside : 03:11:48 create you can use 03:11:49 input scanning feature of theirs 03:11:55 can b inherited 03:13:01 --- join: gforth (guest@adsl52033.vnet.hu) joined #forth 03:13:15 no I guess that's not quite right 03:13:19 ' can be inside : 03:13:19 hrm 03:13:33 g4 : x ' execute ; 4 x . 03:13:35 onetom: 4 03:13:46 g4 : x ' execute ; 4 3 x + . 03:13:49 onetom: 7 03:22:46 so, how would you create an 'add' function that adds the next two numbers 03:22:48 add 2 3 03:22:50 .s 5 03:25:47 How are interrupts usually handled with forth? 03:31:25 heya 03:32:36 rob_ert: I think that's sort of an oxymoron... interrupts are handled, at least at the base level, in ASM 03:36:41 hoinky and rob yours 03:37:04 rob_ert: what do u mean by that? who can it b handled *usually*? 03:37:19 davidw: let me c... 03:39:34 davidw: dont know. but check the i/o tutorial in gforth 0.5.0 manual 04:07:10 if forth is -the- os in a machine, it's at least potential that the forth involved would have interrupt service routines... but this is a special case; most of the time either interrupts are not handled (i.e., any i/o uses polling) or the underlying OS does the interrupts... 04:07:51 jim: ok, but forth still has to deal with the low level stuff somehow 04:08:39 bbl 04:09:11 generally, forth deals with its -own- low-level stuff... which is not to say forth cannot have asm words that are interrupt service routines 04:10:15 for the case where forth is the only thing in the machine, first, why is that the case? example: in sun machines, forth is only used as a boot loader 04:10:31 Hmm 04:10:41 but there are times when forth is the only thing running in that sun 04:10:49 What I mean was simply how a forth routine should be called 04:11:01 (From a non-forth program, or an interrupt) 04:11:16 interrupt handler* 04:11:37 "Hey Joe, bring that 10M hit/day web server down; I need to calculate this formula... bring it into the boot loader and type ..." 04:12:08 :) 04:14:07 the rules do not belong to forth for that case, you must follow the rules of interrupt service routines, i.e., save ALL state (incl flags, pc, all regs and all other processor state), find the requesting device, find out what it needs, service that need (possibly altering external state, like queues or such) and restore saved processor state, and return from interrupt 04:15:27 one reason I can see for the interrupt handler for some device being in a forth dictionary is to allow forth to directly use the device 04:15:41 That's what I mean. 04:15:59 How do I call a forth routine from an assembly interrupt handler? 04:16:46 probably you won't be able to do that... but you might be able to code an asm "wedge" that does what you want 04:19:50 Couldn't I just save all registers, set IP to the word I want to execute and then run NEXT? 04:21:10 unless you can make calling a forth routine reentrant and make it safe to call during an interrupt, it's generally not a smart approach to call a forth routine from an interrupt, partially because it might be overly expensive to initialize a forth interpreter inside an interrupt, partly because it's too slow, partly because "where does it return to?" "how will you know to restore/return"? 04:21:40 interrupts like raw speed 04:22:03 the quicker, the better (or you might lose the next interrupt) 04:22:35 Hmm 04:22:44 having said all that, it might be worth a try 04:23:00 What happends on an x86 if an interrupt is generated while another one is being processed? 04:23:09 Or if the interrupt flag is set... 04:23:21 Does it just skip the new interrupt? 04:23:25 depends upon what interrupts are masked out 04:23:50 maybe... but, of course, you lose out on handling the interrupt 04:24:41 of course, all bets here are off on things like F21s :) 04:25:30 in that case "next" is what the processor does, and "ip" is the pc... 04:25:32 Well... I only have x86 and older stuff... 04:25:40 Yeah 04:25:46 That'd be quite cool. 04:26:08 Do you know any good page/document about multithreading with forth? 04:26:18 I'd like to see how that's done 04:27:25 you're -far- better off creating an interface between the interrupt and forth, in the entity of a data structure which would be written by the interrupt and read by certain forth words 04:28:20 multithreading with forth? never done that. my guess, tho, is to have a forth-in-C and link it against a thread library 04:28:39 then build api binding words 04:28:40 Hehe 04:28:45 I didn't really mean that. 04:28:53 I meant multithreading within the forth 04:29:09 Somewhere I read about a z80 with a forth system that ran about 50 processes. 04:30:03 You could for instance change thread on every NEXT. 04:30:17 well, again, you'd need reentrancy... and for each process/thread, you'd need a block of data 04:30:18 (Maybe a bit slow, but.. that's forth :-) 04:30:31 jim: ok, well if you have an interface, how does forth deal with it? 04:30:40 I guess each thread will need its own: IP, return stack, data stack 04:30:44 it has to be alerted to the fact that it's got something to handle 04:30:54 jim: Yes, but that can be allocated... 04:31:53 davidw: the interrupt acts by altering the data structure... 04:32:12 but I'm beginning to see you also want unix-like signals 04:32:48 ok, you can alter data structures till you're blue in the face, but unless forth acts on it, your device is just sitting there 04:33:32 davidw: typically, a unix i/o routine is a loop that runs until some sort of end-of-data condition exists... 04:33:43 uhmmm 04:33:51 often it uses select or something 04:33:58 so that it doesn't poll 04:34:04 polling is wasteful 04:34:27 maybe it can block when looking for input 04:34:36 ok, but interrupts can't really block 04:34:47 you can't sit their blocking on keyboard, disk, ethernet, etc.... 04:35:04 what you need, more or less, is this: 04:35:11 low level asm stuff to handle interrupt 04:35:20 (but this question you're asking is phrased too generally: "how does forth handle interrupts"; stated that way, I have to reply: any way you want it to! 04:35:32 then it passes it to a higher level forth routine, that's less system critical (i.e. it doesn't block the whole system) 04:35:36 Hmm 04:35:38 that's pretty common in other OS's 04:36:00 the bitch is the reentrancy bit 04:36:29 IP has to be a reentrant var 04:36:38 (for one thing) 04:43:01 davidw: the "less critical" thing could be an application-level data consumer loop 04:43:44 which could live in its own thread, and itself get blocked if no data (so it doesn't need to run rampant while waiting for data) 04:44:59 that makes more sense:-) 04:45:16 there's also the issue of atomic updates of the data structure by the interrupt service routine 04:45:35 (i.e., the updates have to be atomic :) 04:47:19 it would be cool if forth had some kind of test and set 04:48:20 like I said, you could link a forth-in-C to appropriate libs and get those things 04:48:31 heh, that's what I'm doing 04:48:48 the bitch is, at the moment, figuring out ecos' interrupt stuff 04:49:00 the ones I introduce don't seem to work...:-/ 04:49:28 you're saving -everything-? 04:51:18 at the moment, I'm still dicking around with the C part of it 04:51:32 I haven't even begun to look at the forth part:-( 04:51:38 I'm using pforth 04:51:47 it has some good stuff, but I'm not %100 convinced 04:51:52 ficl looks interesting too 04:52:55 lunch time 04:59:01 --- nick: mur -> murAway 05:00:53 --- join: Speuler (~l@a161161.upc-a.chello.nl) joined #forth 05:00:58 Hi Speuler :) 05:01:06 hi the_rob 05:14:33 wee gates ? 05:16:45 Het gaat goed, dankje... 05:17:03 *staun* 05:35:16 --- nick: murAway -> mur 05:35:40 re 05:45:05 --- join: gforth0 (guest@adsl52044.vnet.hu) joined #forth 05:46:42 gforth0: bootmessage 05:46:49 gforth0: bootmessage 05:46:54 gforth0: bootmessage 05:47:02 4th: bootmessage 05:47:32 gforth0: bootmessage 05:47:50 Speuler: GForth 0.5.0, Copyright (C) 1995-2000 Free Software Foundation, Inc. 05:47:50 Speuler: GForth comes with ABSOLUTELY NO WARRANTY; for details type `license' 05:47:50 Speuler: Type `bye' to exit 05:54:42 gforth0: s" uname -a" system 05:55:49 s"blablaabl" adds string? is it null ending or pointer in stack? 05:56:07 mur: don't forget the space after s" 05:56:19 leaves addr and count on stack 05:56:33 --- join: onetom_ (tom@adsl52044.vnet.hu) joined #forth 05:56:38 addr = pointer to first char 05:56:44 count = n chars 05:56:53 oh yea, sure, okay 05:57:14 Which is most common, one-byte or one-word string counts? 05:57:18 so pointer and lenght? 05:57:22 mur: left$ ( addr1 count1 u -- addr2 count2 ) umin ; 05:57:27 k 05:57:32 rob_ert: byte count 05:57:57 Do these support support longer strings? 05:58:06 er 05:58:08 systems* 05:58:58 rob_ert: string handling words (expecting addr and count) usually work on strings >255 chars too 05:59:18 rob_ert: but words which assume first byte of string to be count, don't 05:59:35 Okay... 05:59:55 rob_ert: hardly any words assume count byte at first address 05:59:58 Is there any disadvantage other than size when using word-sized counts? 06:00:05 rob_ert: COUNT does 06:00:27 rob_ert: yes. 06:00:33 rob_ert: count won't work 06:00:40 What does count do? 06:01:04 rob_ert: if you fix it, (to work with word), it won't work when it is used, assuming byte access 06:01:19 t'is often used for other purposed than string count byte fetching 06:01:29 : count 1+ dup 1- c@ ; 06:01:54 : count dup 1+ swap c@ ; 06:02:05 I see. 06:02:14 first is often quicker 06:02:14 --- quit: onetom (Read error: 110 (Connection timed out)) 06:02:15 --- nick: onetom_ -> onetom 06:02:40 less stack activity 06:02:47 Yeah... 06:03:00 --- quit: gforth (Read error: 110 (Connection timed out)) 06:03:00 --- nick: gforth0 -> gforth 06:03:07 gforth: see count 06:03:10 Speuler: Code count 06:03:11 Speuler: ( $804C5D7 ) add 204 [esp] , # -4 \ $83 $84 $24 $4 $2 $0 $0 $FC 06:03:11 Speuler: ( $804C5DF ) mov esi , 204 [esp] \ $8B $B4 $24 $4 $2 $0 $0 06:03:11 Speuler: ( $804C5E6 ) xor eax , eax \ $31 $C0 06:03:11 Speuler: ( $804C5E8 ) mov al , 0 [ebp] \ $8A $45 $0 06:03:24 incomplete 06:03:26 But, hmm... I don't feel quite comfortable with such small strings... 06:03:39 gforth: see does> 06:03:43 rob_ert: noname : 06:03:43 rob_ert: dodoes, here !does ] defstart :-hook ; 06:03:43 rob_ert: lastxt 06:03:43 rob_ert: noname : 06:06:28 run time words for string literal are usually implemented for byte count string too 06:06:59 i.e. the strings compile with ." foo" , for example 06:07:04 compiled ... 06:08:39 Hrm.. 06:08:48 Guess I'll have to break a few standards :-) 06:09:35 rob_ert: i'm not sure whether ans still requires strings to have byte count 06:10:00 could be that it is an implementation choice 06:10:01 not sure 06:10:16 but COUNT would be a problem case 06:10:39 count is often used as autoincrementing byte fetch 06:10:53 not only in conjunction with strings 06:11:21 Then I'll simply not use that ;-) 06:11:24 have count produce something else would break exisiting code 06:12:08 Hrm.. 06:12:18 Guess I won't make it standard anyway... 06:12:45 what's wrong with nonaam ? 06:12:54 Huh? 06:13:03 nonama? 06:13:16 Everybody love nonaam ;) 06:13:21 Bah 06:13:23 I meant nonama 06:13:30 He's annoying :-) 06:15:00 Speuler: Do you know a good hashing algorithm that generates a minimum of collisions? 06:15:21 probABLY 06:15:26 need to look it up 06:15:31 did some xperiments 06:15:44 empirically 06:16:00 Okay :) 06:16:07 tried a few, and let a long list of words loose on it 06:16:40 Did you find any algorithm that didn't generate any collisions? 06:16:57 :) 06:16:58 a speedy one, but reasonable distribution, was i think 06:17:02 rob_ert then there is bug: ) 06:17:07 oh 06:17:09 whops .) 06:17:09 mur: ? 06:17:22 nm 06:17:24 rob_ert: no 06:17:41 Okay... Was the number 32-bit? 06:17:44 i tested it to force collisions (number of strings versus size of hash array) 06:18:19 hashed strings to 32-bit, modulo'ed those to #slots in hash-table 06:20:35 used hashed vocs for the psion implementation 06:20:44 (written in a basic-liek interpreter) 06:21:01 to speed up compilation 06:21:20 18 mhz machine ... 06:21:43 128 threads 06:22:09 average of 5 or 6 words / thread 06:23:40 rob_ert: collison-free seems impossible to me 06:23:45 Hmm 06:23:53 as it depends on the number, or string, which is hashed 06:24:00 I mean, for the most commonly used forth words. 06:24:27 To avoid collisions in a dictionary of - say - 200 words shouldn't be impossible, or? 06:24:48 possible i think 06:24:58 sure, possible 06:25:09 rob_ert: jeppis dialekt igen: er hede "jakobstads" öli na goo ? 06:25:10 but depends on number of threads 06:25:58 mur: Heh? 06:26:13 haven't got it at hand 06:26:23 Number of threads? 06:26:30 What does that have to do with it? 06:26:38 thread = hash slot 06:27:03 in case of collision, the slot becomes start of list for entries, using the same slot 06:27:13 could double hash too 06:27:30 but often, making a list of colliding words is cheaper 06:28:04 Hrm... 06:28:11 I think I'll skip hashing instead :) 06:30:55 PROC hash:(name$) 06:30:56 local hashvalue& 06:30:56 hashvalue& = 0 06:30:56 i& = len(name$) 06:30:56 while i& 06:31:05 hashvalue& = hashvalue& + asc(mid$(name$,i&,1)) * (i& * 2 + 1) 06:31:05 i& = i&-1 06:31:05 endwh 06:31:05 return (hashvalue& and KThreadMask%) + 1 06:31:05 ENDP 06:31:17 that's the simple psion version 06:31:21 hashing for speed 06:31:33 therefore very simple hash algorithm 06:32:03 Dankje. 06:32:17 np 06:32:17 Seems pretty fast :) 06:45:04 --- join: gforth0 (guest@adsl52089.vnet.hu) joined #forth 06:53:03 Speuler: Hm... I need a way to load and intepret/compile a source file in my forth... 06:53:15 Speuler: But I also want to be able to get keyboard inout. 06:53:17 input* 06:53:24 Any ideas on how to design that? 06:53:41 rob_ert: what's the problem ? 06:54:02 Well... 06:54:10 I can't really figure out where to start. 06:54:13 different solutions 06:54:16 consider: 06:54:32 Easlier I've just had a "read-line" function that reads from keyboard. 06:54:34 key reads filehandle from a var 06:54:42 usually, standard in 06:55:23 when you open a file, write that handle to the variable key refers to 06:55:37 then key read next chars from file 06:55:48 save the previous handle on a stack 06:56:05 when file has been read, close it, restore previous handle 06:56:19 (allows you nestable includes) 06:56:30 Yeah... thanks :-) 06:57:00 So I'll simply call "compile ( file-handle -- )" ? 06:57:13 don't use compile 06:57:17 --- join: onetom_ (tom@adsl52089.vnet.hu) joined #forth 06:57:37 consider "from" or "include" or "fload" 06:57:53 Well, the name doesn't matter. 06:57:57 But that kind of structure? 07:03:30 --- quit: gforth (Read error: 110 (Connection timed out)) 07:03:30 --- nick: gforth0 -> gforth 07:03:35 --- quit: onetom (Read error: 110 (Connection timed out)) 07:03:35 --- nick: onetom_ -> onetom 07:06:30 --- quit: davidw (Read error: 113 (No route to host)) 07:45:30 --- join: tathi (~josh@wsip68-15-54-54.ri.ri.cox.net) joined #forth 07:46:29 Hi :) 07:48:25 hey 07:49:01 Gaah 07:49:09 ? :) 07:49:14 I can't find any instruction timings 07:49:20 :-( 07:49:22 for x86/16-bit. 07:49:30 Hrm 07:49:34 Guess I should get HelpPC 07:49:44 But then I got to get networking to work on my other computer... 07:49:48 grr.. 07:50:02 rob_ert: for what 4th? 07:50:33 ehhh 07:50:44 Hehe 07:50:46 Not forht.. 07:50:52 forth*' 07:51:05 what then?!?!goaway!;) 07:51:09 Bah ;) 07:53:45 :)) na, jo.. kegyes leszek, visszabbveszem 07:53:50 aaarrgghhhhhhh 07:53:56 Hehe 07:54:47 * onetom kicks the monitor down 2 the ground 07:56:01 onetom you need cursor which follows eye pupils and actives then the right window 07:56:06 oe sumting 07:56:54 mur: Why can't you write a spelling corrector for your IRC client? =) 07:57:24 rob_ert sorry, no 07:58:13 :((( 08:00:13 mur: yeah, iknow :/ 08:07:13 is there anybody interested here in a lil nice programming excercise? 08:07:35 forth ? 08:07:41 doesnt matter 08:07:47 Tell us ;)( 08:07:49 tell us then :) 08:07:56 but actually it ... okay :) 08:08:25 imagine a matrix of numbers, a so 2 say M*M table 08:08:53 there r some given elements in it 08:09:27 from the [0,N]EN interval, 08:10:02 where N E (0,M] 08:10:06 right? 08:10:17 * rob_ert doesn't know anything about matrixes. 08:10:22 yeah 08:10:34 rob_ert: they're ... "tables" 08:10:36 rob_ert: for now, u wont have 2 08:10:53 u have 2 fill the not given elements 08:11:24 in all the possible combinations w numbers from the [0,N] interval 08:11:42 but there r 2 constraints 08:11:45 1st: 08:12:03 oink: I know what they _are_, but not how to count with them. 08:12:25 if u walk the table along a spiral starting from the left top corner 08:12:57 rob_ert: not need 4 any number-matrix operatoins 08:13:11 OK :) 08:13:51 u have 2 get a series of acsending numbers periodically 08:14:10 zeros dont count, so 08:14:22 (is that your homework or what ? :D) 08:14:33 1002003120003001000020030102... 08:15:04 oink: sure. it was a homework @ some univ. 08:15:35 oink: actually it should b coded in some declarative prg lang, like prolog or sml 08:15:59 ugh.. 08:16:00 oink: but its not so simple, so any solutions do play ;) 08:16:32 oink: coz a student is very curious about any algorithms 08:16:50 oink: tho, that student is not me 08:17:15 oink: im interested in the problem, because i wanna introduce forth 08:17:34 so what's the other constraint? 08:17:35 oink: on a linux-flame mailinglist via this problem 08:17:42 2nd: 08:18:15 none of the rows or columns can contain the same number more the once 08:18:22 zeros dont count again 08:18:45 they r the joker values ;) 08:18:48 ---------- 08:18:54 is that clear? 08:18:59 rob_ert: hm? 08:19:34 I'm not clear on _exactly_ what the first constraint is... 08:19:40 what about holding a contest, huh? ;) 08:20:16 tathi: ok. whats wrong: the spiral path? 08:20:31 tathi: or the structure of the series? 08:20:35 the series 08:21:09 u have 2 fill the empty elements w numbers from the [0,N] intervall 08:21:10 you mean that, ignoring 0s, you need to get 12..n12...n etc? 08:21:12 -l 08:21:17 yup 08:21:24 ok 08:21:55 if N>M its impossible 2 meet the 2nd condition, ithink 08:22:18 unless N>M*M... 08:22:33 that is every field is uniqe 08:23:03 but there r prefilled elements certainly... 08:24:08 a recomposition of the text of this problem is welcome from a 08:24:44 person havin a better knowledge of english mathematical terms & style 08:25:40 a brute force algo in forth will also do 08:26:18 but 1 optimized 4 cheap (that is fast) execution would b more welcome 08:34:31 nah? 08:35:57 --- quit: Speuler (Read error: 113 (No route to host)) 08:44:35 No thanks :-) 08:44:44 (I'm back, btw) 08:45:46 rob_ert: that is? 08:46:28 Huh? What? 08:52:01 what do u mean by nothanks? 08:52:20 Your problem. 08:52:28 I'm not going to try it .) 08:52:32 why? 08:52:44 Too hard ;-) 08:53:08 oooh, nooo 08:53:20 the bruteforce method isnt that hard 2 08:53:31 not even difficulty ;p 08:53:35 not even difficult ;p 08:53:40 Hehe 08:54:07 its a perfect excerice 2 try out the power/weekness of 4th 08:54:40 rob_ert: imagine the bruteforce method: 08:54:59 u have a series w holes in it 08:55:48 consider the series of the holes as an x digit number 08:56:05 and simply generate all the possible numbers 08:56:26 that is N^x variations 08:57:02 and then simply test all the possible "numbers" against the conditions 08:57:40 2 implement the condition checking routines inst too difficult 08:57:55 probably the spiral-walk is not so obvious 08:58:17