2021-09-16 06:31:10 <dave0       > i coded today!
2021-09-16 06:31:54 <dave0       > i wrote "find" yesterday and today i optimized it
2021-09-16 06:32:19 <dave0       > i may have prematurely optimized it
2021-09-16 06:32:55 <f-a         > can you share it?
2021-09-16 06:33:04 <dave0       > but it's noticably faster on dosbox
2021-09-16 06:33:41 <dave0       > i suppose
2021-09-16 06:41:47 <dave0       > i gotta clean it up!
2021-09-16 07:00:16 <dave0       > https://termbin.com/pc6u
2021-09-16 07:00:53 <dave0       > it's not meant to be run, i assembled it by hand
2021-09-16 07:01:06 <dave0       > i jut added the comments then, they are pretty bad
2021-09-16 07:03:40 <dave0       > i have 16 pointers in context, where   context 1 cells +   is all words of length 1,  context 2 cells +   is all words of length 2,   context len cells +    for length in general
2021-09-16 07:07:28 <dave0       > i read some advice from a story on hacker news to do a little bit of coding every day
2021-09-16 07:07:36 <dave0       > so i tried it
2021-09-16 07:07:42 <dave0       > it works!
2021-09-16 07:10:54 <dave0       > i hand assembled it because the colon compiler doesn't exist yet, and in order to make the compiler, i need to be able to look up words :-)
2021-09-16 07:15:32 <f-a         > thanks
2021-09-16 07:16:34 <dave0       > it's pretty rough
2021-09-16 07:16:41 <dave0       > not good for much
2021-09-16 07:19:53 <dave0       > i thought i read a story that chuck moore said use dup and over, but don't use swap and rot ... the idea i got was if you have to explain what's on the stack, you're doing something wrong
2021-09-16 07:20:49 <dave0       > and only have a maximum of 2 or 3 values on the stack.. any deeper and you should refactor
2021-09-16 07:20:59 <dave0       > so i started using variables
2021-09-16 07:22:38 <dave0       > the compare word begs to get coded in assembly
2021-09-16 07:23:31 <dave0       > it has 3 values on the stack
2021-09-16 07:23:43 <dave0       > but on x86 there is hardware acceleration
2021-09-16 07:26:03 <dave0       > before i read what chuck said about dup and over, i would have put "name" on the return stack and juggled it with lots of >r and r> ... but a variable is so much nicer
2021-09-16 07:27:16 <dave0       > oh the trick of having an array of linked lists for the words, only takes 32 cells of memory
2021-09-16 07:27:32 <dave0       > but it's a lot faster
2021-09-16 07:28:29 <dave0       > because each linked list only consists of the same length words, it saves a comparison in the inner loop of find
2021-09-16 07:28:46 <dave0       > and i can cache the offset to the <name> field of a word
2021-09-16 07:31:42 <dave0       > the reason the dictionary entry is laid out in a strange way (the name is at a negative offset) is so i can go from an xt to a word by subtracting a cell
2021-09-16 07:41:44 <dave0       > f-a: it's pretty awful code, right? :-)
2021-09-16 07:41:48 <dave0       > i'm a newbie
2021-09-16 07:42:10 <dave0       > i've been writing c for so long that it's worfed my brain
2021-09-16 07:43:06 <f-a         > not really awful
2021-09-16 07:43:22 <f-a         > but I need to read in vim, I am frankly unused to no syntax highlight
2021-09-16 07:44:09 <dave0       > does vim have a mode for forth??
2021-09-16 07:45:13 <f-a         > not a mode, just syntax highlight
2021-09-16 07:45:29 <f-a         > which isn’t the fanciest syntax highlight but works for me
2021-09-16 07:46:40 <dave0       > f-a: have you made your own forth?
2021-09-16 07:47:32 <f-a         > I tried to make a forth for the CHIP-8, but failed
2021-09-16 07:47:38 <f-a         > well not a forth, a cross compiler
2021-09-16 07:50:55 <dave0       > it's hard
2021-09-16 07:53:09 <veltas      > Failed in the "give up" sense, or the "bored" sense?
2021-09-16 11:17:55 <feoh        > Would be kind of interesting to build a 'fantasy console' a-la TIC-80 or PICO-8 but use FORTH as the programming environment rather than Lua.
2021-09-16 11:18:06 <f-a         > that would be indeed lovely
2021-09-16 11:18:32 <eris[m]     > feoh: uxn
2021-09-16 11:18:48 <eris[m]     > https://wiki.xxiivv.com/site/uxn.html
2021-09-16 11:20:18 <feoh        > That looks like a very nifty project
2021-09-16 11:20:23 <feoh        > Thanks for the pointer!
2021-09-16 11:20:50 <feoh        > However have you looked at TIC-80?
2021-09-16 11:20:51 <feoh        > https://tic80.com/
2021-09-16 11:21:07 <eris[m]     > very different aims and goals
2021-09-16 11:21:13 <feoh        > Yep
2021-09-16 11:21:25 <eris[m]     > let me find the creator's post about it, haha
2021-09-16 11:22:03 <feoh        > I'm suggesting something like TIC-80 that runs on modern hardware but presents a 'retro'-esque runtime environment, using FORTH as the programming language
2021-09-16 11:22:52 <eris[m]     > thats interesting but i dont think it neccessarily offers anything unique over tic-80 then?
2021-09-16 11:23:12 <eris[m]     > not that there's anything wrong with that ^^
2021-09-16 11:23:25 <eris[m]     > but, where does forth's advantage set in here?
2021-09-16 11:24:50 <feoh        > I think it would be interesting because FORTH's stack based nature and use of dictionaries/words could be an interesting way to decompose the workings of a game.
2021-09-16 11:25:53 <eris[m]     > sounds perfect!
2021-09-16 11:26:13 <eris[m]     > i think you could take some inspiration from iKe here
2021-09-16 11:26:53 <eris[m]     > it's not a specifically retro styled thing, and it's more livecode-y than fantasy computer-y
2021-09-16 11:27:04 <eris[m]     > but it uses a more esoteric language
2021-09-16 11:27:06 <eris[m]     > that being, K
2021-09-16 11:27:54 <eris[m]     > https://github.com/JohnEarnest/ok/tree/gh-pages/ike#ike
2021-09-16 11:51:19 <feoh        > Neat, thanks!
2021-09-16 11:52:34 <feoh        > When I hear 'k' I think of the APL like language
2021-09-16 11:52:36 <feoh        > Same?
2021-09-16 11:53:20 <eris[m]     > yes, iKe is a K environment
2021-09-16 11:53:47 <eris[m]     > if you look at the gif(in the link!), you'll be able to see it in action ^^
2021-09-16 11:55:31 <feoh        > Looks neat. Always had a hard time wrapping my brain around APL
2021-09-16 11:55:36 <feoh        > and APL like languages.
2021-09-16 11:56:05 <eris[m]     > it's difficult!
2021-09-16 11:56:14 <eris[m]     > im trying to get better at it
2021-09-16 11:56:47 <eris[m]     > my favourite piece of K code has been `(|+\)` as of late
2021-09-16 11:56:57 <eris[m]     > it's a nifty little thing
2021-09-16 14:42:12 <Zarutian_HTC> h'lo folks
2021-09-16 14:43:21 <Zarutian_HTC> what was the equiv of C's switch statement in many forths loosly following ansi forth "standard"?
2021-09-16 14:44:10 <Zarutian_HTC> often gets compiled/translated to jump tables
2021-09-16 14:47:01 <Zarutian_HTC> here we go http://lars.nocrew.org/dpans/dpansa6.htm#A.6.2.0873
2021-09-16 14:47:31 <MrMobius    > that type of translation is an optimization the C compiler does. not a lot of forths optimize so i wouldnt expect many to create a jump table
2021-09-16 14:49:35 <Zarutian_HTC> well, some forths I have seen use a case variant of jump table
2021-09-16 14:51:01 <Zarutian_HTC> something like (CASE) followed by a cell containing number of entries followed by the entries
2021-09-16 14:51:41 <Zarutian_HTC> each entry being two cells, one for the selector value and one for the jump target
2021-09-16 14:54:27 <Zarutian_HTC> the whole thing followed by the code for the default case then the code for each case
2021-09-16 14:57:22 <MrMobius    > neat
2021-09-16 14:57:29 <MrMobius    > you could set something like that up yourself
2021-09-16 14:58:06 <MrMobius    > i think C only makes the jump table though if it's a simple pattern it can recognize
2021-09-16 14:58:34 <MrMobius    > so you can calculate the jump address based on the index without having to do any comparisons
2021-09-16 14:59:26 <Zarutian_HTC> yeah, one of the patterns is that the case selector value is less than 11 bits in size
2021-09-16 15:01:03 <MrMobius    > i mean a pattern such as 1,2,3 or 5,7,9. if you have cases with values like 2, 78, 104 then the C compiler wont find a pattern
2021-09-16 15:03:38 <Zarutian_HTC> ... and the switch statement has a case for each possible value or near it then a jump table is made iirc
2021-09-16 15:06:52 <MrMobius    > right but with 2, 78, 104 the addresses need to be arranged in memory in some way that calculating the address to load from the table is faster than doing the comparisons
2021-09-16 15:07:08 <veltas      > Rest in peace, Sir Clive Sinclair
2021-09-16 15:07:09 <MrMobius    > so for that youd need about 100 empty entries in the table which the compiler would not do
2021-09-16 15:07:21 <f-a         > rip Clive
2021-09-16 15:14:06 <veltas      > I think they are probably smart enough to do a very basic hash and mask to allow unconditional jumping, only one way to find out though
2021-09-16 15:14:36 <veltas      > I would expect them to generate a perfect hash, but maybe I expect too much of them :P
2021-09-16 15:39:33 <MrMobius    > ive played around with it on godbolt before. its pretty slick: https://godbolt.org/z/Trf1G488s
2021-09-16 17:39:56 <feoh        > So are all you folks just hard core chips & bits types or do any of you use FORTH on vanilla 'modern' hardware like Factor and the like?
2021-09-16 17:42:23 <veltas      > There's a good mix in here I think
2021-09-16 17:42:58 <veltas      > The hardcore chips and bytes people tend to gravitate a bit more to #mecrisp
2021-09-16 17:43:42 <veltas      > I'm currently writing a Forth for x86-64 UEFI systems, not the first to write a x86-64 Forth in here
2021-09-16 17:44:01 <veltas      > Most people in here learning Forth will start with gforth
2021-09-16 17:44:50 <f-a         > never tried factor
2021-09-16 17:44:54 <f-a         > I use gforth
2021-09-16 17:45:03 <f-a         > but frankly the allure of forth lays in its simplicity I feel
2021-09-16 17:45:10 <f-a         > I need to get a frigging board and experiment
2021-09-16 17:45:11 <MrMobius    > im building a graphing calculator with a 1mhz 6502 which is forth based
2021-09-16 17:45:48 <veltas      > Whereas I have no idea what the allure of Factor is
2021-09-16 17:46:52 <veltas      > MrMobius: Why use a 6502 today?
2021-09-16 17:48:23 <MrMobius    > veltas, for this project there was a competition to build something with a 6507
2021-09-16 17:48:36 <veltas      > Ah
2021-09-16 17:48:57 <MrMobius    > but in general its one of the best choices in DIP if you need external bus
2021-09-16 17:49:40 <MrMobius    > 14mhz with modern 65C02. maybe only 68000 or 65C816 are better
2021-09-16 17:55:48 <feoh        > I tend to be interested in things with a relatively high level of abstraction, but IMO that's one of the really cool things about Forth. The abstraction levels can go as high or as low as you want.