00:00:00 --- log: started forth/03.07.20 00:21:15 --- join: a7r_ (~a7r@206.72.82.135) joined #forth 00:21:16 hey 00:24:13 --- join: skylan_ (sjh@vickesh01-4772.tbaytel.net) joined #forth 00:26:04 --- quit: skylan (Read error: 113 (No route to host)) 00:27:24 --- nick: skylan_ -> skylan 00:41:54 hey hey 00:42:50 sup 00:42:53 --- nick: RO|_ -> rO| 00:47:04 --- quit: a7r (Read error: 110 (Connection timed out)) 01:30:29 --- quit: flyfly ("using sirc version 2.211+KSIRC/1.2.4") 02:38:18 --- quit: skylan (leguin.freenode.net irc.freenode.net) 02:38:18 --- quit: ianni (leguin.freenode.net irc.freenode.net) 02:38:18 --- quit: Robert (leguin.freenode.net irc.freenode.net) 02:39:24 --- join: Robert (~snofs@h126n2fls31o965.telia.com) joined #forth 02:40:49 --- join: skylan (sjh@vickesh01-4772.tbaytel.net) joined #forth 02:40:49 --- join: ianni (ian@inpuj.net) joined #forth 04:21:47 --- quit: Robert (leguin.freenode.net irc.freenode.net) 04:21:48 --- quit: onetom (leguin.freenode.net irc.freenode.net) 04:21:48 --- quit: rO| (leguin.freenode.net irc.freenode.net) 04:22:07 --- join: Robert (~snofs@h126n2fls31o965.telia.com) joined #forth 04:22:07 --- join: rO| (~RO|@pD9E59674.dip.t-dialin.net) joined #forth 04:22:07 --- join: onetom (~root@cab.bio.u-szeged.hu) joined #forth 04:40:11 --- quit: a7r_ (Read error: 113 (No route to host)) 05:03:02 --- join: CrowKilr (CrowKilr@Ottawa-HSE-ppp3653907.sympatico.ca) joined #forth 05:18:32 --- quit: onetom (leguin.freenode.net irc.freenode.net) 05:18:32 --- quit: rO| (leguin.freenode.net irc.freenode.net) 05:18:32 --- quit: Robert (leguin.freenode.net irc.freenode.net) 05:18:45 --- join: Robert (~snofs@h126n2fls31o965.telia.com) joined #forth 05:18:45 --- join: rO| (~RO|@pD9E59674.dip.t-dialin.net) joined #forth 05:18:45 --- join: onetom (~root@cab.bio.u-szeged.hu) joined #forth 05:24:47 --- join: draq (ident@203-219-179-14-nsw.tpgi.com.au) joined #forth 05:28:22 --- join: RO|_ (~RO|@pD9EE1A0A.dip.t-dialin.net) joined #forth 05:34:22 --- join: mur (murr@baana-62-165-189-64.phnet.fi) joined #forth 05:36:08 --- quit: rO| (Read error: 60 (Operation timed out)) 07:11:35 --- quit: RO|_ (leguin.freenode.net irc.freenode.net) 07:11:35 --- quit: draq (leguin.freenode.net irc.freenode.net) 07:11:35 --- quit: onetom (leguin.freenode.net irc.freenode.net) 07:11:35 --- quit: Robert (leguin.freenode.net irc.freenode.net) 07:11:47 --- join: RO|_ (~RO|@pD9EE1A0A.dip.t-dialin.net) joined #forth 07:11:47 --- join: draq (ident@203-219-179-14-nsw.tpgi.com.au) joined #forth 07:11:47 --- join: Robert (~snofs@h126n2fls31o965.telia.com) joined #forth 07:11:47 --- join: onetom (~root@cab.bio.u-szeged.hu) joined #forth 08:31:07 --- quit: CrowKilr ("going on a real life mini-quest to get info about negative green and pyramids") 09:05:43 --- quit: draq (Remote closed the connection) 09:16:16 --- join: wossname (wossname@HSE-QuebecCity-ppp81178.qc.sympatico.ca) joined #forth 10:05:42 --- quit: wossname ("i am gone") 10:25:29 --- join: kc5tja (~kc5tja@ip68-8-206-137.sd.sd.cox.net) joined #forth 10:25:30 --- mode: ChanServ set +o kc5tja 10:26:57 Hello :) 10:27:09 re 10:27:40 * kc5tja managed to separate his shoulder, he thinks. 10:27:55 One of my shoulders sits about an inch lower than the other, and yes, it hurts. :) 10:31:01 Ouch. 10:37:15 Yep. And I don't know how it happened either. It was rather gradual. 10:44:31 Hmm.. now I can rather comfortably copy 8wpm, only a few years to go until I can do 40 ;) 10:44:41 btw, do you know how to convert between wpm and cpm? 10:45:00 cpm = 5 or 6 times wpm? 10:46:13 5 times 10:46:16 (roughly) 10:46:49 One "word" in CW is assumed to be about as long as the word PARIS (for some reason -- I don't know why). 10:47:07 Yeah.. but I didn't know how long one character was assumed to be :) 10:47:28 Thanks. Now I know the Swedish morse code test is at ~5wpm (25cpm). 10:47:38 So I should be able to pass it, I guess. 10:52:26 The 5wpm limitation is/was international requirement (some countries could go harder if they so desired) 10:53:46 * kc5tja got some amount of design work done on FS/Forth while playing D&D yesterday (lots of dead time -- we had nine players!). 10:56:08 Got some additional behavior of the vocabularies defined, and determined that the minimum size of the dictionary for it is 48KB in size (32K for the vocabularies, plus my usual 16KB for actual code compilation space) 10:58:08 I'm still not quite sure how to store the addresses in the Forth environment though. 10:58:33 I could store them as relative addresses (e.g., addresses that are relative to the start of the dictionary), or I can store them as absolute addresses. 10:58:51 Absolute would be more convenient, but I'd have to somehow remember to relocate them every time I loaded the dictionary image from disk. 10:59:16 And where it resides in Gforth's address space certainly WILL differ from where it is in FS/Forth's address space. 10:59:55 However, relative addresses require *constant* run-time relocation using either an explicit ADD instruction or using the [r1+r2] addressing mode. 11:01:05 So that consumes a register that must never be changed. Then, when interfacing to other system libraries, I have to remember to adjust addresses because ! and @ (and related words) will all expect relative addresses. 11:02:07 I'll probably use EBX for the dictionary base pointer register, just as GCC does. It ought to minimize the impact of having to thunk Forth to/from C libraries all the time. :) 11:06:22 --- join: a7r_ (~a7r@206.72.82.135) joined #forth 11:06:38 re a7r_ 11:06:50 hey 11:07:29 Doing some more design work on my Forth environment. 11:07:44 Man, Linux just has to make everything so difficult. :-) 11:07:50 hah 11:07:59 yah,.. well, that's cuz it sucks. :) 11:07:59 * kc5tja is dealing with the need to write relocatable code. 11:08:21 --- nick: mur -> mur_bbl 11:08:41 * kc5tja wishes he had his 65816-based PC finished. I'd be coding for it rather than x86. 65816 is so much simpler. :) 11:08:54 Well, yes and no. 11:09:03 * kc5tja still hasn't figured out his input-related problems yet. 11:11:03 I think that's why pforth doesn't process KEY quite right, is because they don't bother to shove the terminal into RAW mode (the method for doing this on a text-mode console is NOT the same for doing this in an x-term). 11:11:03 That saves a fair bit of code. 11:11:22 yah 11:11:43 I hacked Ficl into SDL last night, and wrote a graphical console for it. 11:12:05 * kc5tja nods 11:12:11 unfortunately, my SHA1 code doesn't pass my tests under Ficl (but it works with pforth and gforth).. so I don't know what's up. 11:12:34 Hmmm 11:12:49 Isn't Ficl supposed to be ANSI-compliant? 11:13:13 Does the SHA1 code produce different results, or does it not compile? 11:17:58 different resuls 11:18:01 t 11:18:14 Hmm...that is weird. 11:18:34 I wonder if Ficl has a bug in it that produces different results than ANSI's requirements. 11:19:31 yah, I'm not sure what the deal is.. the only thing I do that's minorly tricky is a rotate left, shifting the MSB to the LSBs as they roll off the top. 11:20:05 * kc5tja nods 11:26:29 --- quit: a7r_ (leguin.freenode.net irc.freenode.net) 11:26:29 --- quit: onetom (leguin.freenode.net irc.freenode.net) 11:26:29 --- quit: RO|_ (leguin.freenode.net irc.freenode.net) 11:26:29 --- quit: Robert (leguin.freenode.net irc.freenode.net) 11:26:29 --- quit: kc5tja (leguin.freenode.net irc.freenode.net) 11:26:30 --- quit: ianni (leguin.freenode.net irc.freenode.net) 11:26:30 --- quit: skylan (leguin.freenode.net irc.freenode.net) 11:26:30 --- quit: mur_bbl (leguin.freenode.net irc.freenode.net) 11:26:41 --- join: mur_bbl (murr@baana-62-165-189-64.phnet.fi) joined #forth 11:27:10 --- join: kc5tja (~kc5tja@ip68-8-206-137.sd.sd.cox.net) joined #forth 11:27:10 --- join: skylan (sjh@vickesh01-4772.tbaytel.net) joined #forth 11:27:10 --- join: ianni (ian@inpuj.net) joined #forth 11:27:10 --- mode: leguin.freenode.net set +o kc5tja 11:27:21 --- quit: kc5tja ("[x]chat") 11:27:28 --- join: a7r_ (~a7r@206.72.82.135) joined #forth 11:27:28 --- join: RO|_ (~RO|@pD9EE1A0A.dip.t-dialin.net) joined #forth 11:27:28 --- join: Robert (~snofs@h126n2fls31o965.telia.com) joined #forth 11:27:28 --- join: onetom (~root@cab.bio.u-szeged.hu) joined #forth 11:27:44 --- join: kc5tja (~kc5tja@ip68-8-206-137.sd.sd.cox.net) joined #forth 11:27:51 --- mode: ChanServ set +o kc5tja 11:27:53 That was naughty.... 11:33:24 --- quit: mur_bbl (leguin.freenode.net irc.freenode.net) 11:33:24 --- quit: onetom (leguin.freenode.net irc.freenode.net) 11:33:24 --- quit: a7r_ (leguin.freenode.net irc.freenode.net) 11:33:24 --- quit: RO|_ (leguin.freenode.net irc.freenode.net) 11:33:24 --- quit: Robert (leguin.freenode.net irc.freenode.net) 11:33:25 --- quit: ianni (leguin.freenode.net irc.freenode.net) 11:33:25 --- quit: skylan (leguin.freenode.net irc.freenode.net) 11:33:38 --- join: mur_bbl (murr@baana-62-165-189-64.phnet.fi) joined #forth 11:34:02 --- quit: mur_bbl (Excess Flood) 11:34:03 --- join: onetom (~root@cab.bio.u-szeged.hu) joined #forth 11:34:03 --- join: Robert (~snofs@h126n2fls31o965.telia.com) joined #forth 11:34:03 --- join: RO|_ (~RO|@pD9EE1A0A.dip.t-dialin.net) joined #forth 11:34:03 --- join: a7r_ (~a7r@206.72.82.135) joined #forth 11:34:03 --- join: ianni (ian@inpuj.net) joined #forth 11:34:03 --- join: skylan (sjh@vickesh01-4772.tbaytel.net) joined #forth 12:11:35 --- join: crc (~crc@ACA93B2E.ipt.aol.com) joined #forth 12:18:38 --- join: mur (murr@baana-62-165-189-64.phnet.fi) joined #forth 12:21:12 --- quit: crc ("I was using TinyIRC! Visit http://www.tinyirc.net/ for more information.") 12:57:28 --- join: flyfly (~marekb@ip164.ktvprerov.cz) joined #forth 13:59:14 --- quit: skylan (No route to host) 14:20:01 --- join: skylan (sjh@vickesh01-4504.tbaytel.net) joined #forth 14:54:45 --- quit: RO|_ ("!") 15:55:24 --- quit: TreyB () 16:06:45 --- quit: mur ("MURR! save the http://rainforest.care2.com/ (click url there)") 16:38:13 --- join: TheBlueWizard (TheBlueWiz@pc6adn1d.ppp.FCC.NET) joined #forth 16:38:14 --- mode: ChanServ set +o TheBlueWizard 16:38:23 hiya all 16:43:56 gotta go...bye all 16:44:00 --- part: TheBlueWizard left #forth 17:17:36 doh 17:27:56 --- join: a7r (~a7r@206.72.82.135) joined #forth 17:29:07 --- part: a7r left #forth 17:32:46 --- join: CrowKilr (CrowKilr@HSE-Windsor-ppp251236.sympatico.ca) joined #forth 17:32:55 hi 17:36:48 re 18:05:07 --- join: Ro0tSiEgE (Ro0tSiEgE@Ro0tSiEgE.sponsor.freenode) joined #forth 18:05:24 --- part: Ro0tSiEgE left #forth 18:28:58 --- join: TheBlueWizard (TheBlueWiz@pc6adn1d.ppp.fcc.net) joined #forth 18:29:00 --- mode: ChanServ set +o TheBlueWizard 18:29:05 hiya all again 18:29:36 hi 18:29:43 hiya CrowKilr 18:40:44 oi 18:41:21 hiya a7r_ 18:41:38 hey 18:43:05 what's new? 18:46:22 im working on my 5 colors byte token forth system 18:46:46 kc5tja is working on fs/forth, herkamire on his ppc forth 18:46:57 is that a variant of ColorForth? 18:47:40 a little 18:48:05 its inspired by the work of chuck, use the same optimizations ?dup and ?lit in the compiler 18:48:40 the spec is on forth/bespin.org 18:49:01 hmm...ok 18:49:02 i cant go on the site, proxy error 18:49:37 ouch...I offer to whack the proxy server for ya :) 18:57:38 gotta go to bed...bye all 18:58:54 --- part: TheBlueWizard left #forth 19:20:49 Unfortunately, there is no proxy involved. 19:21:00 I really do not understand why you are receiving that error message. 19:22:44 I get a proxy error too... And I don't use a proxy 19:24:58 Neither do I. It works perfectly fine for me. 19:25:11 (and no, it is not my server; it's Tril's) 19:35:18 I think I'll go to bed now, night all. 19:35:57 * kc5tja is getting hungry, so I'm going to grab something to eat. 19:43:52 see ya rob 19:59:12 my system has too many ambiguities, im almost lost in the small amount code I produced 20:00:20 like what? 20:01:17 my litteral handling doesnt work 20:01:21 literal 20:01:49 yeah it works after all, but it looks like this 20:02:13 # ?dup opt 1, B8 ; 20:02:38 definition, macro, macro, macro, bincopy macro 20:03:07 so in code i want a litteral i write: # 12345678 20:03:28 macro, bincopy 20:03:53 since i use tokens, i never had to compile calls 20:04:05 maybe thats because im too early in code i dont know how to express it 20:04:28 but its all defs, macros and bincopy type of tokens 20:05:36 opt means optimization btw 20:05:45 * kc5tja nods 20:05:45 mov ecx, ebx mov ebx, edi 20:05:55 2 element in regsiter stack 20:06:14 What produces the code to push the current value of EAX on the data stack before loading it with the new value? 20:06:31 ok ill explain this small code fragment 20:06:45 # ?dup opt 1, B8 ; 20:06:48 # is a definition 20:07:12 Meaning, # is a "red word." 20:07:26 its tokenized in 2 bytes as 00001000 "#" 20:07:27 yes 20:07:31 OK 20:07:42 spaces are just to make output lisible 20:07:53 they doesnt exist and one dont need them with colors 20:08:12 * kc5tja nods -- I'm familiar with how ColorForths work. :-) 20:08:23 ?dup is a macro, so its encoded as a one byte token 20:09:12 the detoken word, upon decoding this token, jumpt to ?dup code, wich was defined this way: 20:09:38 its a two part definition, since i use jnz in ?dup, dup must be placed before 20:10:10 dup 4, 8D76fc89 1, 06 ; 20:11:15 ?dup 8D47FF39C375EB81388B068D7675E389C7 ; 20:11:33 thats the first optimization 20:11:51 drop definition goes along with the second, wich is ?# 20:12:05 anyway when i use the ?dup macro 20:12:24 the long string of bytes get executed, wich could be deassembled on the fly in the editor 20:13:07 # ?dup opt 1, B8 ; 20:13:58 So the code that does the stack pushing is ?dup then. If the previous stack item is known to be DROPed, there's no need to manipulate ESI; just overwrite EAX. Otherwise, emit the code to DUP first, then overwrite EAX. 20:14:04 opt is a macro too, so its tokenized as a single byte, pointing to: opt 4, 89D989FB ; 20:15:43 ?dup check in the previous code if there was a drop compiled, by first checking in EBX (last optimizable word address), and then checking for the correct dword at the specified memory location for 100% bug free loopback optimization 20:16:15 if theres a drop, just do lea EDI, edi-5, since drop is 5 bytes long and it must be overwritten 20:16:24 if not then a dup is compiled 20:16:50 by jumping to previous dup definition 20:17:40 so my ?dup word is bugged 20:17:49 since its the version when i still used lodsd 20:17:53 ill correct it 20:17:57 ;p 20:18:06 :) 20:18:41 Well, I was asking, because I was pondering how to do the whole dup/drop optimization myself. I've seen it done in ColorForth, but I haven't studied the code for ideas yet, because I haven't gotten that far. 20:19:12 Your solution to just "back up and roll over the DROP code" with new code is pretty neat. I'll have to record that for future reference. 20:19:32 its chuck idea's lololol 20:19:41 im just copying from the master 20:19:45 Well, like I said, I haven't studied his code yet. 20:19:58 So I didn't know. 20:19:59 i spent a month looking at color code ;o) 20:20:12 his two optimizations are ?dup and ?lit 20:20:17 ?lit use the zeroflag 20:20:25 its more complicated 20:20:31 it looks back the ebx pointer 20:20:37 to check for dup's prescence 20:20:42 to overwrite it also 20:20:46 what it does 20:20:54 if theres a lit 20:21:18 it get the pointer at the beginning of it and put the lit value in eax for the primitive macro to handle 20:22:06 in the primitive you do a call ?lit jnz nolitcode litcodewithvalueinEAX 20:22:11 Hehehe :) That's evil. :) 20:22:54 i knew my system would become the best, since it enable to experiment with these concepts pretty easly 20:23:00 but tahts still to be done 20:23:34 since i have not written an editor yet and I yet have to make a cdrom eltorrito bootloader read hd sectors 20:23:39 in raw mode 20:23:46 anyway 20:23:55 ill continue my explanantion ;p 20:24:05 # ?dup opt 1, B8 ; 20:24:27 the correction btw is simple, change FF in previous definition by FB (-1 by -5) 20:25:05 so opt is opt 4, 89D989FB ; 20:25:45 4, is a macro for compiling code inside the definition to compile the following 4 bytes 20:26:31 4, is defined as 4, 8d7f06c747fa8d7f04c766c747fe47fcc3 20:27:01 notice the c3 at the end 20:27:10 i didnt used ; because its not defined yet 20:27:28 in fact, for speed, every primitive can use c3 instead of a macro call to ; 20:27:33 * kc5tja nods 20:27:35 ill explain ; 20:28:14 btw 4, code compiles 6 bytes: lea EDI, [EDI+4] mov [EDI-4], ... 20:29:02 ; is like ?dup, it's a "multiredword" definition 20:29:19 i mean theres a dependency issue 20:29:24 it starts wih 20:29:28 with* 20:30:21 no wait i could do something slower but that would show you more code 20:30:41 after i defined 4, the long way i defined 1, 20:31:04 1, 4, 47c647ffc3 20:31:25 again the c3 is manually put because ; and ret definitions are just after 20:31:42 ret 1, doret C3 20:32:01 no 20:32:05 i cant do that 20:32:13 anyway i wanted to explain 1, 20:32:22 the code that ive written is this: 20:32:22 ok 20:32:38 ret c607c347 20:32:43 doret c3 20:34:13 ; "8D47FB39C375F48038E875EFFE00" 20:35:10 brb -- phone. Keep writing if you want; I'll read it when I get back (shouldn't be too long) 20:35:33 pushopt "8D40FD39C275EA8038E875E5C6006840FF30FF70058F008F4005 ret 20:35:44 those 4 lines are my ; definition 20:37:03 ; is broken in two parts, the first, following the def token is to compile the same optimization as chuck: if the previous compiled thing is a call, replace it with a jump (opcode E8 becomes E9) 20:37:26 there's a " typo btw 20:38:16 you could remove pushopt entirely, the word would be ok if terminated by ret or C3 20:39:06 back 20:39:08 push opt implement another opt: it detects "call1 jump2" code through the same 100% bug free algo and then replace it with "push2 jmp1" 20:39:19 pushopt* 20:39:54 so testing the speed of this opt and its validity would be easy 20:40:09 Oh, now that's an interesting optimization. 20:40:38 * kc5tja hadn't thought of doing that -- push2 would push the address of word2 on the return stack I assume, so that word1's RET causes a forced branch to it. 20:41:06 wonder if one could push subsequent call adresses in stack, but that would require first of all a deeper opt stack 20:41:17 subsequent... i meant previous 20:41:40 I'm not sure it'd be faster though. 20:41:51 chuck didnt impplemented it, even with his 2 deep opt stack, he could have, but seems its not worth it 20:42:07 me neither, but least the code is there 20:42:27 so now you can understand my # word.... 20:42:40 but while explaining lot of ideas rush through 20:43:08 new ways of doing things, like string table things that could solve some problems im seeing in my detoken design 20:43:22 what jumps in front of me is 20:43:28 1 - i dont use the call token at all 20:43:34 at least now now 20:43:57 2- i can comment out pushopt def with a single bit, but its long bincopy thing would still be there 20:44:15 so it would be nice to tie a string to each def, using an external table 20:44:39 so when i comment the def, i comment the code, and tokens can be use to compile that string, jump to it, etc 20:44:50 no overhead involved like in my system and less ambiguities 20:44:57 but the picture isnt clear still 20:45:33 at least im having fun doing it right now, thinking about this, so i guess im on the way to get something better than waht i thought i had 20:46:45 * kc5tja nods 20:46:56 and its not everyday i can give back something to kc5tja, who makes me learn so much ;p 20:47:10 * kc5tja has little to no artistic talent in and of himself, so he rarely endeavors to explore such things. 20:48:46 I guess I'm more "imaginative" in other areas. But really, nothing I've done to date is truely "new" or "innovative." 20:49:00 I haven't blazed a new trail in my entire life. 20:49:11 me neither, but im on it right now and its pretty exciting 20:49:53 Actually, that isn't strictly true: I *did* exploit Forth's dictionary and colon-definitions to implement key-bindings for VIBE. :) 20:50:15 I had never seen any Forth program do that before I did. Nor have I seen any do it since. :) 20:50:35 what do you mean by using the dictionary to bind keys? 20:52:42 FINDing the word associated with a key press, and EXECUTE-ing it. 20:53:03 (executing a default handler if no explicit word is found) 20:53:22 you man you serach for the scancode value in the dictionary? 20:54:03 i thought about using a short-lived tmeout to get a counted string of input from keybaord to do exactly that, universal kb handling, onthefly keyb configuring with least amount of code etc 20:54:26 In other words, if I receive an ASCII key value of $41 (ASCII "A"), I look for a word $$c41 or $$i41, depending on whether VIBE is in command- or insert-mode at the time. 20:55:27 if it works for you it should work for me ;p 20:55:44 * CrowKilr didn't tried VIBE yet 20:56:13 It's currently written for GForth, since that's the Forth environment I'm using to write FS/Forth's target compiler. 20:57:43 It's usually pretty easy to port to other Forth environments; however, it does require a truely interactive version of KEY -- a7r_ has run into problems using VIBE with pforth, for example, since pforth treats its input as a series of lines, rather than a series of characters. 20:59:01 bahhhh, go native =) 20:59:36 I've considered going straight native right away, actually. 20:59:40 But I lose a lot if I do. 20:59:56 For instance, I lose the ability to browse the web while running in FS/Forth, because I don't have a Forth web browser. 21:00:11 And that is a huge, huge, huge, huge, huge negative for me and my business. 21:00:18 simulator or cross-development 21:00:35 umbilical dev as they call it in holon 21:00:47 For that matter, I lose all support for networking too, since my network card's driver source is indecipherable. 21:01:03 * kc5tja was going to write my environment for SDL under Linux. 21:01:20 once its done, you beat future software and your hardware into submission 21:01:36 This would give me a native-like run-time environment, but with the convenience of being able to run under Linux at the same time. 21:02:51 Well, the problem is that no matter what I do, the step from going from Linux to native hardware is going to be so huge that it'll be as destructive as starting over from scratch. 21:03:10 linux has a magnetic effect on me when i was in 14th grade, then when i got my system (tv ati video card running under 2.2 kernel and otehr kewl hacks, no real programming though) my bubble popped, litterallly, i lost the fire 21:03:45 2 weeks later i f*ck** the hd and it was all gone, i didnt bother to rcuperate it 21:03:50 i lost so much time 21:04:02 own personal point of view here 21:04:32 my bubble was fueled by the fact that i would get total control 21:04:44 The issue is that Linux has a lot of software already written for it which enables me to run my business right now, which is sorely needed right now. 21:04:49 i guess seing the kernel source from my own eyes, and using the system discouraged me 21:05:14 yeah ok business 21:05:19 If I go native Forth right away, I am basically 30 man-years in the whole as far as software I can use to run my business. 21:05:22 im still in school =) 21:05:28 s/whole/hole/ 21:05:44 not if I help you and if others help you 21:05:50 if we form a true community 21:06:00 with everyone understanding everything 21:06:07 liek we could do a small test 21:06:18 Can you write a JavaScript-enabled web browser that supports XHTML 1.0 and HTML 4.2 by the time I get my next web development client? 21:06:28 ok lectures wahetevr so tahte veryone knows about formalities as ?dup and ?lit in compiler design 21:07:35 hmmmm 21:07:44 i dont know enough about html 4.2 21:07:48 xml i could 21:07:56 if i had drivers for framebuffer 21:07:58 kb 21:08:01 network card 21:08:15 The key words: by the time I get my next client. 21:08:18 once the drivers are in fun begin 21:08:21 It isn't going to happen. 21:08:30 imagien you have the drivers, would it take much time? 21:09:02 No human being on this Earth, even Chuck, would be able to write a browser capable of competing with Opera or Mozilla, within two weeks. 21:09:34 For that matter, no team of humans could do it either. 21:09:55 2 weeks full time, with flawless simple drivers and all the tcp/ip/html/etc specs, would be easy to brainstorm something in 1 or two day and start from there 21:10:06 Maybe in two years, I can see that. Maybe, if you manage the project just right and the workers are dedicated, maybe a single year. But not two weeks. 21:10:16 No, it wouldn't. 21:10:17 okok you got more experience 21:10:25 i cant argue ;p 21:10:39 So you see where my dilemma is. :) 21:10:53 but the idea of counted string from kb + dictionary search could be used 21:11:07 I wanted to make an evolutionary path, moving each major piece of software I needed from Linux to Forth, as I needed them, and as my time permitted. 21:11:44 Eventually, FS/Forth would be self-sufficient enough to run native-hardware, and then I'd have all my important projects available to me. 21:14:35 fs/forth and my project and everyforth for that matter needs what im trying to do: establish a true elegant standard that would utterly render every other way of thinking about forth obsolete 21:14:43 colorforth is close to that 21:15:14 Well, I don't subscribe to so militant a view with respect to Forth environments. 21:15:39 Despite adopting many of ColorFOrth's characteristics, it's still a "traditional" ASCII-source-based Forth environment. 21:16:21 everyone here knows it exist, the right way to do things, yet they cant put their finger on it 21:16:42 there's still fat left 21:16:51 even in my minimalistic design 21:16:56 theres still fat lol 21:17:29 5 tokens, they dont match the intent, the code meant to run on them 21:17:44 i was seduced the other day by the string table idea 21:17:58 im sure it can bring simplicity 21:18:06 like i dont like my big pushopt 21:18:09 i comment it 21:18:22 the def could be a ref into the string table 21:18:30 * kc5tja is pretty happy with the structure of his vocabulary layout, actually. 21:18:35 editor would be simplified: the library of words would be a string table 21:18:48 i like counted strings 21:19:05 High level assembler is bloated: use counted and zero termination 21:19:34 * kc5tja uses explicitly counted strings all the time. I never use implicitly counted or null-terminated strings anymore. Ever. They're just too much trouble. 21:19:39 IMO, at least. 21:19:41 counted strings enable you to use zero, so they're better imo 21:20:13 yay we agree ;p 21:20:20 They also reduce the need for so many temporary buffers too. 21:20:51 And, it can be extended to types other than strings -- a string is just a byte array. The concept can be extended to any kind of array. 21:21:09 dynamic array, thats the name 21:21:16 (addr u) is a powerful pair of stack cells: addr is the base of the array, and u is the number of elements (not necessarily the number of bytes). 21:21:29 an array is of defined length, a count enable it to vary in size 21:21:59 yep, with the adress on top or u? 21:23:14 u on top 21:23:25 (usually; it can be any convenient order, but I usually store u on top for consistency) 21:23:48 The idea is that a reference to the array consists of both its base address AND the number of items within it. 21:28:54 got an idea 21:29:07 since i use macros everywhere etc i came to think of something 21:29:07 What's that? 21:29:29 every word refernce that was a macro could be a compilecall token 21:29:43 at the end, the end word that signal the end of source file 21:30:04 would convert the last call, if any, into a jump and execute it or something 21:30:31 the idea is that the call everywhere makes the code be a "compilation bomb" 21:30:59 when one calls the top level word 21:31:15 everything compiles 21:31:23 everything beneath it that is 21:31:52 i think thats how chuck do it when i think of it 21:32:10 he says compilation on the fly, takes no perceptible time 21:32:14 Well, he has some top-level words that are defined like this: 21:32:21 : icons 36 load ; 21:32:32 where typing "icons" causes code to be compiled every time. 21:32:58 load is his token interpreter, equivalent to my detoken word 21:33:06 * kc5tja does this with VIBE, since I do not want the editor to overwrite what has already been compiled, but I don't want it to last in memory either. It works like an overlay. 21:33:07 it depends what inside the block 36 21:33:40 what do you mean, you use a scratch space? 21:33:55 The dictionary is the scratch space. 21:34:07 MARKER foo ....code code code... foo 21:34:44 As soon as the code is finished, it drops to 'foo,' which causes the whole chunk of code to be removed from the dictionary. 21:34:57 chuck probably uses empty at start of block 36 21:35:13 Yes -- he uses EMPTY in front of all his applications. 21:35:23 And it's much superior to MARKER I find. 21:35:35 I've made some simple ANSI-compatibility definitions for it: 21:35:48 empty is a default state with primitives in memory 21:35:57 : mark S" marker _empty_" evaluate ; 21:36:04 : empty S" _empty marker _empty" evaluate ; 21:36:15 : empty S" _empty_ marker _empty" evaluate ; 21:36:27 GODDAMMIT!!! What is wrong with my )(*$&@(#* keyboard? 21:36:31 : empty S" _empty_ marker _empty_" evaluate ; 21:36:43 Geez, you'd think the _ key was broken or something. 21:36:55 : unmark S" _empty_" evaluate ; 21:37:23 It's a dirty hack, but it works. 21:38:02 * CrowKilr don't know about ansi forth, so doesnt see much 21:38:53 It's equivalent to this: 21:39:05 : mark " : _empty_ ;" evaluate ; 21:39:19 : empty " forget _empty_ : _empty ;" evaluate ; 21:39:28 : unmark " forget _empty_" evaluate ; 21:43:29 doesnt ring a bell much, anyway i get the idea even if i cant understand it fully 21:43:48 * CrowKilr is still thinking hard about efficiency in his system 21:44:32 GOT IT 21:44:44 WOAH lol if it works it will change lot of things 21:44:52 okokokok 21:44:59 the source is alive, dynamic 21:45:20 tokens get transformed and erased 21:45:39 the def token get erased, pointer -1, but the address is saved in memory 21:45:56 bincopies: 21:46:20 when the token is shifted out, i got a counted string (counted at sign change, to get even range) 21:46:47 so i write back this count where its supposed to be and i decal every bytes of the string there 21:46:55 compilation would be called mutation lol 21:47:07 ok imagine i have my ret def 21:47:21 ret c607c347c3 21:47:57 in tokens its 1+ret+1+c607c347c3 = 10 bytes 21:48:25 when compiled or mutated 21:49:14 it would see a def token: save the address in memory like before, the source pointer esi adavnces but edi points at rel address 0 21:49:49 esi:4 EDI:0, because def token skips the definition word 21:50:25 now the detokener right shift the token, detects a bincopy, jumps to code 21:50:45 that bincopy handling code takes that count, put it at the [EDI] 21:51:27 and after compiles: mov AL, [ESI] 21:51:36 inc ESI 21:51:52 mov byte [EDI], al 21:51:55 inc edi 21:52:02 or just movsb for small code size 21:52:10 6x smaller 21:52:38 * kc5tja is a bit too tired to figure out what you're saying. 21:52:42 ooo my code is buggy, better use ah lol 21:52:58 i woudl compile over old source code 21:53:05 But I did realize something: if I use SDL for FS/Forth, on my 1024x768 screen, with an 8x8 text font, I'll have 125 columns across, and 87 rows. 21:53:05 not in a new place in memory 21:53:16 That's an awful lot of text to put up on the screen. :) 21:53:33 second level driver: text on graphic display, having this in forth would be great 21:53:56 What's second level about it? That's precisely what I'm going to write for my Forth under Linux. 21:54:08 But, here's the dilemma: do I want 80x25 text, or 125x87? 21:54:16 I suppose I just have to code it to find out. 21:54:33 go for 80x25, will be able to use it in native mode ;p 21:55:06 That requires I emulate a PC text mode in the graphical mode. I'm not sure I want that kind of overhead. 21:55:15 second level, because you need a frambuffer driver before thinking plotting pixels 21:55:26 Well, SDL covers that. 21:55:27 yeah better to use two tasks like chuck did 21:55:33 No!!! 21:55:42 lol ok i said nothing ;p 21:55:46 Chuck's style of multitasking is horribly inefficient! 21:56:01 The CPU never gets a chance to halt when it has nothing to do, so it drives the CPU temperature right up. 21:56:42 good point for running it on higher end machines than a 486 ;p 21:56:49 For running on ANY machine. 21:57:03 Heat dissipated is a function of how many transistors are being worked at any given time. 21:57:17 If the CPU has nothing better to do with its time, HALT it. Otherwise, it'll suck power needlessly. 21:57:47 Very few processors lack a proper HALT instruction. The 6502 did, but the 65816 makes one available (thankfully). 21:58:24 * CrowKilr tells himself: note to self: since im gonna use interrupts, halt use becomes possible 21:58:24 The only key, is that HALTing a CPU requires interrupts to be enabled, because it's an interrupt that wakes the CPU. 21:58:31 * kc5tja nods 21:58:33 Precisely. 21:58:50 another arg in favor of ints 21:58:59 chuck disabled ints ;p 21:59:02 * kc5tja nods 21:59:11 He re-enabled them for his networking code. 21:59:23 thats exactly what i was going to say ;p 21:59:35 i saw the code posted on the cf mailing list 21:59:45 * kc5tja nods -- that's what I have too. 22:00:03 he got his ir transducer working!!! didnt looked carefully ifts a parport /serial one or built in 22:00:19 i want to use my laptop as a fat ass remote too lol 22:00:43 would enable me to test remotes at work 22:01:01 * kc5tja intends on using his computer to control various aspects of his radio, and for accessing AX.25 networks. 22:01:08 This just requires RS232 capability. 22:01:13 we have a "chirper" for this right now, little black box with speaker that chirps if it receives a signal 22:02:02 parports for umbilical dev is great on old boxes, network would be even better because of boot, capabilities cable range etc 22:02:43 Thing is, I need to write a network driver for my Forth environment if I'm going to support networking. And like I said before, the Linux sources for my card are virtually undecipherable. 22:02:55 what is your card? 22:03:06 I suppose I have no choice but to reverse engineer the code though. 22:03:11 buy a realtek if you can 22:03:22 i got SiS hardware in my desknote laptop 22:03:25 3Com 3C95x chipset. 22:03:38 was hoping i could get docs but they dont give them 22:04:10