00:00:00 --- log: started forth/04.06.09 01:54:13 --- quit: htp123 (Read error: 60 (Operation timed out)) 02:03:03 --- join: htp123 (~tehsux@S010600055d233ab7.gv.shawcable.net) joined #forth 02:18:09 --- quit: Herkamire ("goodnight") 02:45:57 --- join: Topaz (~top@exten-halls-131.soton.ac.uk) joined #forth 02:52:45 --- join: fabian_ (~fabian@pD95D76C2.dip.t-dialin.net) joined #forth 02:56:07 --- quit: fabian_ (Client Quit) 03:21:36 --- join: qFox (C00K13S@cp12172-a.roose1.nb.home.nl) joined #forth 03:51:58 hmm, putting TOS in registers does add rather a performance boost 03:52:12 my '+' has gone from 14 cycles to 6 :D 04:01:21 --- join: solar_angel (~jenni@Toronto-HSE-ppp3685160.sympatico.ca) joined #forth 04:08:12 --- quit: Topaz ("Leaving") 04:58:27 so you can create 2VARIABLE's for doubles, but there's no word to store a double? 04:58:34 that doesnt make sense... 04:58:46 then define one. 05:08:44 http://members.home.nl/qfox/output.bmp <-- 15mb rarfile byte frequency graph. 05:10:58 heh 05:11:14 not much room for improvement there. 05:11:20 not bytewise, anyway. 05:11:44 its interesting to see that the frequencies go up every 8 bytes or so 05:12:05 and then drop down again 05:12:32 (and this 8byte average increases a little from begin to end as well) 05:13:25 heh, fair enough 05:20:37 * solar_angel is rewriting a FORTH kernel, from DTC to STC with totally different register assignments. 05:21:10 --- join: arke_ (~Chris@wbar8.lax1-4-11-100-108.dsl-verizon.net) joined #forth 05:21:12 --- quit: arke (Read error: 104 (Connection reset by peer)) 05:28:54 how did you convert a counted string to an uncounted string? 05:29:02 ? 05:29:04 the opposite of COUNT 05:29:12 i don't use counted strings :P 05:29:34 it's not even remotely ansi. 05:29:56 thats irrelevant 05:30:25 it even still has a few ideosyncracies from its heritage as an RPN language that i designed before ever hearing of FORTH :P 05:30:46 uh 05:31:06 what? 05:31:18 i just need to save the filename and an uncounted string is easier to use, but i have the string as a counted string 05:31:21 thats all :\ 05:31:31 ah. add 1 to the address? 05:31:48 unless you don't have a terminating nul, in which case you're pretty much borked. 05:31:52 duh. 05:31:57 the point of an uncounted string 05:32:03 is that the len goes in the first byte 05:32:12 *shrugs* when i use counted strings, i put a terminating nul anyway 05:32:18 it's well worth the extra byte. 05:32:30 yes well but thats not how it's supposed to work. 05:32:36 hehehe 05:33:00 plus i'm fairly sure there's a reverse for COUNT but i cant seem to find it... 05:35:34 pack ( adr1 len adr2 -- adr2 ) 05:35:34 Makes a packed string from adr1 len, placing it at adr2 05:35:39 except, pack is undefined :\ 05:35:53 icky. 05:35:57 which forth are you using? 05:36:09 place ( adr1 len adr2 -- ) 05:36:09 Makes a packed string from adr1 len, placing it at adr2 05:36:11 hm 05:36:21 perhaps its place then 05:36:23 win32forth 05:37:33 ah 05:38:51 ick, recoding DODOES... this should be fun. 05:39:10 dodoes? 05:39:27 it's the internal bit that makes DOES> work. 05:39:40 oh 05:40:03 if you're not doing insane stuff deep in the guts of a FORTH kernel (like, say, writing it, or translating it from DTC to STC...) then you probably wouldn't see it. 05:40:28 my "variables" had a does> by default that would @ its adr. if does> was used the additional words were compiled along with it 05:40:52 was pretty straightforward 05:41:08 yes, but this is the actual definition of DOES> 05:41:30 oh you didnt mean an actualy bit 05:41:31 heh 05:41:33 okay 05:41:43 no, i mean word. 05:41:48 DODOES is the internal guts of DOES> 05:41:57 yeah ok 05:42:05 i thought you were talking about a flag or something :) 05:42:11 if only :P 05:42:34 oh wow, i think it's just 3 instructions in STC. 05:42:40 * solar_angel doublechecks to make sure she's not on crack. 05:43:08 drugs are bad mkay 05:43:12 hehehe 05:44:04 i'm trying to change the frequency graph of a random file (rar being "random") 05:44:12 i see. 05:44:17 so far only minimal successes though 05:44:35 tried to rotate the bytes a looping increasing amount of times 05:45:05 so byte0 1shl, byte1 2shl, byte2 3shl ... byte8 0shl again... 05:45:57 hehe, i once coded an utility 'frqpic' 05:46:04 in one theory, this should randomize the contents, but in another theory i believe that a true random file will result in the same frequencies 05:46:39 it took byte pair from file as XY pair and incremented the pixel 05:46:41 only the values might shift from byte to byte (but i havent seen this behaviour yet, just minor drops/raises) 05:47:04 like a "fire" algo? 05:47:12 RAR is quite random, but some strips on long archives are visible 05:47:41 yeah but i'm focussing on random data, rar is sufficiently random for that i think 05:48:38 although i guess any correct huffman encoded file should be "random" 05:51:23 no-o-o! 05:51:43 zip is WAY more squareful than RAR on my frq graph 05:52:04 hm i actually havent analyzed a zipfile with the graph yet 05:52:27 what about bzip2? 05:52:39 i first need to fix that stringthingie 05:52:40 i got no one handy to try 05:53:02 btw, i coded the util to see what the unknown file is 05:53:03 because then i can save the output files to be named like the input files, instead of having to copy "output.bmp" manually every time 05:53:11 unknown file? 05:53:46 and i also need to do some bubblesorting output :p 05:53:56 qFox: picture ? wav ? tracker module ? - they all have didtinct fingerprint on the graph 05:54:13 oh 05:54:48 one even can tell apart types of RAW wav or pic data - bits, RGB vs. palette... 05:55:30 on bigramms frequency graph 05:55:52 hm 05:55:54 thats funny 05:56:07 zip appears to have about twice as many zero's as any other char 05:56:15 ???? 05:56:32 here i'll show the graph 05:56:33 what is your analysis tool ? 05:56:42 the one i coded myself 05:57:14 gfx, textual output ? 05:57:18 http://members.home.nl/qfox/output.bmp 05:57:24 that. 05:57:25 :) 05:57:33 no-o-o-o ! PNG it ! 05:57:37 screw that 05:57:39 i wanted to 05:57:44 convert output.bmp output.png 05:57:49 that's all you have to do :P 05:57:50 but the file format was too complicated to simply do it 05:58:00 use libpng 05:58:04 in forth? 05:58:05 :\ 05:58:09 i coded this thing in forth 05:58:21 hooray :) 05:58:27 gimme src ! 05:58:30 well whatever. look at the graph 05:58:34 at the bmp 05:58:44 how big is it ? 05:58:45 its 0compression 24bit (500k) 05:58:57 it has colors! 05:58:58 :p 05:59:00 he-he-he 05:59:06 too bad nothing supports pcx anymore... pcx is good for graph output. 05:59:09 to indicate min/max/avg 05:59:27 well at first i wanted to do png 05:59:35 but looking at the specs i figured that was beyond me for now 05:59:56 i found out what gzipped pnm is as good as TIFF 06:00:02 plus i just wanted a simple straightforward output without having to spend days developing a lib for the image :) 06:00:17 pnm = image in textual numbers 06:00:26 bmp can be too 06:00:31 8bit 06:00:43 then its 1byte per 1px 06:00:51 (now its 3) 06:01:01 why not PCX ? almost anything opens it ! 06:01:11 but anyways. when you look at that image you can clearly see that there are twice as many zero's as any other number 06:01:13 Serg - that's not even remotely true anymore, unfortunately. 06:01:19 Serg> because i never heard of it. 06:01:19 but PCX is a nice format for the amateur. 06:01:30 or no, because i didnt know anything about it 06:01:32 it's rather dos-ey though. 06:01:38 hasn't really been used in years. 06:01:47 i knew jpg was sucky (loss of detail), and gif is suck 06:02:06 and i didnt know many other format details so i went with bmp. 06:02:42 TIFF / LZW ? 06:03:21 well, don't knock bmp... it's simple, if nothing else. 06:03:34 i like palettized bmp more than 24-bit. 06:03:47 it's a nice file format for palettized images. 06:04:06 it's for game maps only ;)))) 06:04:27 well in this process i did discover something sucky about bmp 06:04:36 it pads 0's 06:04:45 to make the width an even number 06:04:51 oops 06:04:59 yeah 06:05:08 make which width an even number? 06:05:14 the width of the picture itself? 06:05:15 appearantly so 06:05:17 no 06:05:19 internally 06:05:25 ? 06:05:31 say you have a 9x9 image 06:05:31 you mean, it pads 24-bits up to 32? 06:05:45 ah, you'll end up with a 10x9 image? 06:05:47 the inside width for each line would be padded to even bytes 06:05:51 no not even that 06:05:55 ohhh. 06:05:56 weird. 06:06:16 * solar_angel is writing the FILL word. 06:06:20 9x9, 3bytes/px right, 9x3=27 bytes per "line", one 0 would be padded to make it 28 06:06:36 --- join: I440r (~mark4@168-215-246-243.gen.twtelecom.net) joined #forth 06:06:52 or maybe it was to pad to 10x9 internally, i'm not sure actually 06:07:06 my code works anyhow :p 06:07:55 hm 06:08:00 peachy. 06:08:12 hiya I440r :) 06:08:34 hi :) 06:08:44 converting a FORTH kernel from DTC to STC is not for the faint of heart :P 06:08:50 hang on, gotta go get coffee 06:08:52 sure 06:11:15 solar_angel> in theory, huffman would be optimal when the frequency table, sorted, would be evenly devided right? left side of the middle be equally deviating from the average as the right side 06:11:48 um, something like that. 06:12:18 i'm not entirely sure i caught what you mean though 06:13:25 well huffman algo replaces most frequent bytes for a few bits, and the least frequent bytes for more bits, right... and the few bits and more bits put together will be equally to the two bytes again... 06:13:51 fixed a bug in isforth last night, it woudl segfault if you ran it in an unknown terminil (export TERM=foo) 06:15:00 um, something like that. 06:15:03 I440r - cool. 06:15:10 still not as much evil hackery as i'm doing today though 06:15:31 i'm rewriting the underlying core of my entire FORTH system :P 06:15:52 in pascal! 06:15:54 Heh 06:15:55 hehe 06:16:05 You like Pascal, don't you= 06:16:18 err pascal, c, what is difference 06:16:35 Pascal is a little bit tricker to segfault with 06:17:14 C is counter-intuitive to an extreme 06:17:15 pascal? 06:17:25 and way 2 much typing and parens 06:17:28 actually, it's written entirely in assembly. 06:17:39 well, except for the bits written in FORTH anyway. 06:17:40 solar_angel, yea. real coders do that 06:18:10 moving from DTC to STC has pretty radically altered the way this code works though 06:18:17 Hehe. 06:18:31 So are you writing an optimizing STC compiler in assembly language now? 06:19:46 actually, i'm writing an unoptimizing-but-with-stubs-for-an-optimizer compiler in assembly language 06:19:51 the optimizer will be written in FORTH. 06:20:02 Ah, OK. 06:20:57 heh 06:21:09 Hi qFox 06:21:13 hey 06:21:16 How's the compression going? 06:21:31 finished the graph creating code 06:21:49 http://members.home.nl/qfox/output.bmp for an example 06:21:50 What graph? 06:21:55 *shrug* bmp 06:21:57 frequency graph 06:22:02 yeayea look past it. 06:22:26 Hm. 06:22:27 png appeared to be fuck harder then it was worth 06:22:53 That file looks like a bitch to compress, imho. 06:23:01 this is a zipfile actually 06:23:12 and it compresses to about 30% from this file, with winrar 06:23:13 :p 06:23:31 Heh 06:23:50 note that the frequency of 0 (the first, and blue, line) is more then twice as high as any other except one 06:24:59 Ah, didn't see that one. 06:25:05 hehe 06:25:18 the height is proportional to the max value 06:25:44 not to the total filesize (that would result in pretty unreadable graphs) 06:25:44 Are you writing your compression code in Forth? 06:25:48 ye 06:25:53 Neat. 06:26:40 yay for tailcalls. 06:26:45 * solar_angel optimizes her assembly for tailcalling. 06:27:21 Yes.. what I like about those is the simplicity of implementation, you don't have to worry about a dozen of loop constructions. 06:27:46 or a limited stack depth 06:27:59 exactly. 06:28:01 tailcalls are fun 06:28:26 i havent implemented it myself so i'm not sure how easy/hard it is to do :) 06:28:40 it's just about trivial. 06:28:59 the only case you have to really watch out for is when you're terminating a conditional or loop on the last instruction of a function. 06:29:14 but since i'm hand-assembling the body of my FORTH compiler anyway... that's not much of a worry 06:29:56 hmmm but isnt the only thing you check for a call before ret? 06:30:17 so isnt it impossible to break a loop or branch that way? 06:30:33 unless the user did this 06:30:45 but i dont see how this can go wrong really 06:31:06 (provided the user cant mess on asm level) 06:31:14 hrm... 06:31:49 well, a call before a return can still land you in a bit of trouble... 06:31:59 because THEN doesn't necessarily generate any code 06:32:14 "then" doesnt need to generate any 06:32:18 it's easy enough to put a RET after the function to guard for that though 06:32:21 I440r - of course not. 06:32:21 hmmmmm but is then called? 06:32:32 only at compile time, isnt it? 06:32:33 tho mine does 06:32:41 it compiles a dothen (a noop) 06:32:47 Why? 06:32:50 so the decompiler knows how to decompile it 06:32:53 heh 06:32:53 Ah. 06:33:11 Do you have a decompiler? 06:33:16 yes 06:33:22 see blah 06:33:23 see foo 06:33:26 Hrm, you never told me that. 06:33:30 lol 06:33:33 you moron 06:33:43 * Robert points at himself and nods 06:33:54 heh, i don't have a decompiler in mine... 06:34:10 Decompiling optimized STC. Yay. 06:34:14 i haven't found any really compelling reason though 06:34:18 hehehe 06:34:29 you think that's bad... i've got 3 registers free for dynamic register juggling... 06:34:38 whiii 06:34:38 i'm going to STC now so that i can later add an optimizer. 06:34:38 :p 06:34:46 What do you use the others for? 06:34:49 I440r - what register assignments are you using? 06:35:14 esi = IP, ebx = top of stack 06:35:19 ebp = rp 06:35:34 esp = parameter stack 06:36:12 wait... how are you doing STC like that? 06:36:13 eax, ecx, edi are scratch 06:36:16 that sounds more like DTC 06:36:30 mine IS direct threaded 06:36:35 ahhh, alrightie 06:36:37 im opposed to STC 06:36:39 :) 06:36:50 solar_angel: How about your register allocation (in the STC) ? 06:36:52 my old one is direct threaded, my new one is STC. 06:37:02 my register allocation used to be pretty much identical to I440r's. 06:37:04 it's now... 06:37:11 IP -> EIP, obviously... 06:37:19 W -> nonexistant/unnecessary 06:37:25 RSP -> ESP 06:37:28 PSP -> EDI 06:37:35 TOS -> EAX 06:37:48 EBX, ECX, EDX -> reserved for the optimizer, unused. 06:37:57 ESI -> scratch, clobbered regularly. 06:38:10 EBP -> FSP 06:38:18 (sorry for the flip of meanings and sides) 06:38:24 FSP ? 06:39:04 Function Stack Pointer. 06:39:08 my FORTH has a few goodies. 06:39:17 Sounds exciting. What is it? 06:39:20 whats a function ? 06:39:48 it's essentially a callback stack. 06:40:18 ...which is? 06:40:19 err callback ? 06:40:31 it lets you call a word in the context of other words. 06:40:47 if you feel like doing metaprogramming in forth :P 06:41:17 I still don't get it. Explain what calling a word in the context of other words means. 06:41:33 USING PIXEL LINE 06:41:44 or USING MEMPLOT LINE 06:43:53 does that help? 06:44:50 Sorry - no. I don't know what USING does. 06:45:10 USING pushes the following function onto the function stack :) 06:45:55 a parsing word ? 06:46:08 yes 06:46:18 Did you mention what a function is (as opposed to a normal word) ? 06:46:25 do you also have [using] ? 06:46:28 oh sorry... a normal word. 06:46:48 I440r - honestly, the whole function stack system isn't implemented yet... 06:46:51 does using take the next two items from the execution stream or from the input stream 06:46:54 Ah. So what good does a word on that stack do? 06:46:57 it was slated to go in sometime after i convert to color-based syntax. 06:47:21 Robert - the words following it can call it without knowing the word's name. 06:47:45 Like ' MYWORD EXECUTE ? 06:47:50 I440r - the next item from the input stream. the second item isn't affected by USING, although it uses the function stack. 06:48:03 Robert - um... not sure. 06:48:32 Except that uses the parameter stack. 06:48:47 solar would be better if USING took from the executoin stream (so you can do this at run time too) and [USING] took from the input stream 06:49:01 so [USING] would be immediate 06:49:08 Any advantages using a special stack for passing "function pointers"? 06:49:21 Except of course taking some load from the parameter stack. 06:50:04 I440r - i haven't decided the exact semantics yet. 06:50:34 Robert - there are a few advantages for some things i have in mind... it factors along a second axis, makes code smaller... 06:50:43 i think im gona rename "and" to "&" and "or" to "|" and "xor" to "^" 06:50:43 but it's a pretty obscure feature to most people's viewpoint. 06:50:48 i threatned that a while back 06:50:54 heh 06:51:48 solar_angel, obscure is the NORM in forth :) 06:52:31 I440r - exactly. 06:52:54 solar_angel: I rarely use those features in Forth. Going for something more Lispy? (lambda functions) 06:54:57 something lambda functionesque, yes. 06:55:19 you think that's bad, i've also got a CLASS system planned :P 06:56:13 Depends, if you make a bloated implementation without actually using those features in a good way it's bad. But I guess you've got something figured out. 06:56:51 nothing about this is bloated... 06:56:54 i'm very careful about that. 06:57:33 can i make parse continuing parsing after a newline? 06:58:05 i'm still a novice, just playing with string functions.. trying to buffer in multiple lines 06:58:12 Just rewrite it. 06:58:19 What Forth are you using? 06:59:01 gforth? 06:59:09 Is that a question? 06:59:38 ~ is not right ? 07:15:35 sup forthers 07:15:41 i love you guys 07:15:52 what about us girl? 07:16:26 oh holy shit 07:16:31 i love the girls too! 07:16:39 lol 07:16:54 in his defense, the odds on any of those being here were slim :p 07:17:17 qFox - that's why i used the singular sense of the word. 07:17:21 hehe 07:17:23 lol 07:28:34 Hi warpzero. 07:28:53 * Robert <-- a male forther (!) 07:29:03 So how's the Forth in your end= 07:29:03 ? 07:29:39 good 07:29:42 though i don't use it 07:29:49 as often as i would like 07:30:02 :) 07:30:05 Nobody is perfect. 07:30:20 (Sorry, Chuck, I meant almost nobody) 07:30:32 do what i do... write an OS in FORTH with your own FORTH engine. 07:30:44 what i, chuck, and about 2 other people in the world do, i should say. 07:30:46 OK. 07:30:53 Heh 07:31:04 So my efforts aren't counted? *sob* 07:31:15 hehe, you're working on an OS in FORTH? 07:31:23 half of the people in here are 07:31:25 http://members.home.nl/qfox/output.bmp <-- frequency deviation from average (sorted). the graph outlining (bottom/top) is the same atm so dont mind that 07:31:34 Your memory is as bad as mine. http://robert.zizi.org/f.html 07:31:39 Yes. 07:31:51 warpzero is right, more people than you might think are doing that. 07:31:52 ah right... 07:31:54 solar_angel: im writing an OS in a different language 07:32:03 sinner. 07:32:06 warpzero: Which? 07:32:08 warpzero - which language? 07:32:14 Robert - yes, your effort is counted... 07:32:27 solar_angel: :) 07:32:30 solar_angel: the one im implementing 07:32:38 and there's one or two others in here with an OS in FORTH... 07:32:41 warpzero - i see. 07:32:50 well, mine is in my own variant of forth... i like it that way. 07:32:55 solar_angel: kc5tja, Herkamire.. uhm.. and probably a few others. 07:33:29 right.... which one of those is designing the kestrel? 07:33:36 kc5tja 07:33:39 coolness. 07:33:43 the big daddy of this chan :p 07:33:51 and i thought i was the only person here who designs CPU's to run FORTH on. 07:33:56 qFox - how so? 07:34:02 just so 07:34:04 :) 07:34:07