2026-06-01 01:03:39 <xentrac     > tabemann: Sergeant's three-instruction Forth does have the ability to make new definitions!
2026-06-01 01:04:59 <tabemann    > xentrac: in what way? if it doesn't have colon definitions...
2026-06-01 01:06:32 <tabemann    > oh I see
2026-06-01 01:07:00 <tabemann    > you literally write instructions into memory, and then you execute or write instructions to call those instructions
2026-06-01 01:17:03 <xentrac     > yup
2026-06-01 03:25:10 <xentrac     > I think Linux/m68k lasted significantly after that Linux/m68k page was last updated in 02000
2026-06-01 03:33:36 <xentrac     > hmm, actually it still exists.  I have linux-6.12.85/arch/m68k right here
2026-06-01 06:57:52 <Stalevar    > i8080: 2E 18 13 72 23 73 2B 2B 2B 46 23 4E 23 23 23 EB 09 EB D2 03 00 76 00 00
2026-06-01 06:59:03 <Stalevar    > Works on altair simulator it seems, but really should work on any 8 bit 8080/Z80 if zero is RAM. But if it's ROM, then you have to adjust addresses
2026-06-01 08:30:05 <xentrac     > is that a Sergeant-style 3-instruction Forth?
2026-06-01 08:54:12 <Stalevar    > xentrac, no it's 16 bit fibonacci
2026-06-01 09:35:26 <xentrac     > aha
2026-06-01 10:02:51 <Stalevar    > xentrac, I wonder if you can golf it down significantly. Because its output is only up to fib24=46368
2026-06-01 10:03:26 <Stalevar    > My code is 24 bytes, so it only saves 1/2 of size compared to simply filling with result
2026-06-01 10:22:11 <lisbeths    > I have signifcantly changed the memory layout of my spiritual successor to dc int main(){int
2026-06-01 10:22:11 <lisbeths    > x,y,z,t=9999,*g=malloc((sizeof(int))*9999),*h;while('\n'!=(g[--t]=getchar()));t++;while(1){z=(g+t)[0];x=(g+t)[1];y=(g+t)[2];switch(z){c('q',exit(0))c('-',g[x]-=g[y])c('<',g[x]=g[x]<g[y])c('r',g[y]=g[g[x]])c('w',g[g[y]]=g[x])c('a',h=realloc(g,(sizeof(int))*g[x]);h==NULL?free(g):(g=h))c('i',z=g[x];while('\n'!=(*(g+z++)=getchar())))c('o',z=g[x];while(*(g+z))putchar(*(g+z++)))c('x',t=g[x]-3)}t+=3;}}
2026-06-01 10:22:15 <lisbeths    > crc:
2026-06-01 10:25:09 <Stalevar    > lisbeths, you know code golf makes sense in machine code, but in C it just makes it less readable. Though I guess I can ask an AI to put tabs and linebreaks in right places. but it might change something subtly.
2026-06-01 10:25:31 <lisbeths    > the ungolfing of code is solved
2026-06-01 10:25:37 <lisbeths    > you can ask ai to ungolf this
2026-06-01 10:25:52 <Stalevar    > is this complete program?
2026-06-01 10:26:06 <Stalevar    > z.ai eats all ram
2026-06-01 10:26:07 <lisbeths    > yeah but its just a prototype I havent tested it
2026-06-01 10:26:39 <lisbeths    > may have some subtle bugs
2026-06-01 10:34:45 <Stalevar    > lisbeths, it could undeflow if you enter line longer than 9999 bytes
2026-06-01 10:34:58 <lisbeths    > yeah thats true
2026-06-01 10:35:12 <lisbeths    > you can add safeties to it if you want, its gpl2
2026-06-01 10:37:22 <Stalevar    > Why not gpl3? If realloc fails, you free g and keep going, so you probably need exit in this case
2026-06-01 10:39:43 <Stalevar    > lisbeths, anyway, it's still confusing. You have comparison and some arithmetic but then what? You have to use raw numbers to know where to put strings?
2026-06-01 10:43:47 <lisbeths    > yeah I will fix realloc later
2026-06-01 10:44:09 <lisbeths    > the only type is integer
2026-06-01 10:44:53 <dave0       > bignums are something you should consider to write in assembly
2026-06-01 10:45:27 <dave0       > or use something that already has them like lisp or i think python has bigunms
2026-06-01 10:48:27 <Stalevar    > dave0, or bc
2026-06-01 10:49:39 <lisbeths    > well theres this trick that chuck moore talks about that you can learn if you learn a circular slide rule
2026-06-01 10:50:11 <lisbeths    > chuck said in a talk that you dont need big numbers like five million
2026-06-01 10:50:26 <lisbeths    > you can just encode that as a 5 and as a 7
2026-06-01 10:50:51 <lisbeths    > the 5 is t value and the 7 is the magintude
2026-06-01 10:51:17 <lisbeths    > if you combine that with an additional integer to store a remainder, then you have something somewhat similar to a float
2026-06-01 10:52:28 <dave0       > someone did some research and found in programs most numbers fit in 8 bits and i think it was 99% fit in 16 bits
2026-06-01 10:52:55 <dave0       > i think it also said 30% of constants are zero
2026-06-01 10:54:29 <lisbeths    > wow thats cool
2026-06-01 10:55:01 <lisbeths    > i wonder if they counted initializations to zero
2026-06-01 10:55:31 <dave0       > maybe! it's a good point
2026-06-01 11:09:31 <veltas      > 99% in 16 bits doesn't surprise me, yet if 99% of useful programs could fit in a 16-bit address space that would surprise me
2026-06-01 13:52:50 <Stalevar    > xentrac, actually you can get it down to 8 bytes 31 00 01 13 E5 19 EB D2 04 00 76
2026-06-01 13:53:13 <Stalevar    > I mean 13 E5 19 EB D2 04 00 76
2026-06-01 14:04:57 <Stalevar    > I mean 13 E5 19 EB D2 01 00 76
2026-06-01 15:55:30 <Stalevar    > E5 19 EB C7 31 1A 02 13 C7 this is slightly longer and has to start from 04, but it's self-terminating code
2026-06-01 16:45:53 <xentrac     > Stalevar: that seems good
2026-06-01 16:46:40 <xentrac     > code golf is mostly just fun in machine code too
2026-06-01 16:48:52 <veltas      > What is that code meant to do?
2026-06-01 16:51:45 <Stalevar    > veltas, fill memory with fibonacci
2026-06-01 16:52:36 <Stalevar    > Main purpose was to see if I still remember how to work with front panel computers
2026-06-01 16:52:47 <veltas      > I see
2026-06-01 16:53:03 <veltas      > Did you work with front panel computers back in the day?
2026-06-01 16:55:54 <Stalevar    > Nope. My first computer run Windows XP when bought
2026-06-01 16:56:13 <veltas      > Is it using PUSH to write the sequence?
2026-06-01 16:56:15 <Stalevar    > Unless you count NES clone and PSone and BrickGame pocket games
2026-06-01 16:56:43 <Stalevar    > veltas, yes. Because without push it takes 24 bytes while generating 24 numbers
2026-06-01 16:56:50 <Stalevar    > Not very efficient compression
2026-06-01 16:57:48 <Stalevar    > If you use Radio 86rk monitor though, which let you set a breakpoint and initial register values you can get it down to 5 bytes
2026-06-01 17:00:30 <Stalevar    > Or even 4 bytes, E5 19 EB C7 and you set SP to 021A and HL to 0001
2026-06-01 17:01:00 <Stalevar    > Or  E5 19 EB C7 31 1A 02 13 C7 and start it from address 04
2026-06-01 17:01:33 <Stalevar    > Or if you don't overwrite its own code, then 13 E5 19 EB D2 01 00 76
2026-06-01 17:02:29 <Stalevar    > SP is set to 0000 on Altair and it wraps around, so you don't need to init it
2026-06-01 17:03:33 <Stalevar    > Or to be precise, 256 bytes of RAM is wired to 00-FF and high order byte of SP is ignored
2026-06-01 17:04:01 <veltas      > I guess really it helps to choose a scenario
2026-06-01 17:12:12 <veltas      > The Altair 8800?  I can't find info about that
2026-06-01 17:12:30 <veltas      > Looks like there's no truncation of address bus?
2026-06-01 17:12:30 <Stalevar    > https://wixette.github.io/8800-simulator/ this one
2026-06-01 17:12:59 <Stalevar    > Memory only goes up to 0xff
2026-06-01 17:14:31 <xentrac     > well, you could put 64K of RAM into it, or more with bank switching
2026-06-01 17:14:38 <xentrac     > it just didn't come with it originally
2026-06-01 17:22:49 <Stalevar    > xentrac, ofc, but this sim can't run the self-overwriting version because it needs 538 bytes
2026-06-01 17:24:30 <Stalevar    > Though if you don't save two bytes by using RSTn instead of JNC 0xADD3, it will take less stack space
2026-06-01 17:24:40 <xentrac     > you can certainly configure SIMH to have more memory than 256 bytes
2026-06-01 17:24:46 <xentrac     > dunno about wixette's sim
2026-06-01 17:25:06 <Stalevar    > Does simh support front panel?
2026-06-01 17:25:23 <Stalevar    > Last time I only found z80pack which does
2026-06-01 17:25:34 <Stalevar    > It seems that online simulators are more numerous
2026-06-01 17:25:47 <veltas      > I guess my question is whether the 256 byte memory module is ignoring the high address bits, because the bus seems to support a full 64K address space
2026-06-01 17:25:51 <veltas      > Yet you said it wraps round
2026-06-01 17:26:22 <xentrac     > I don't remember but I'm sure it at least had the option to not ignore them
2026-06-01 17:26:25 <Stalevar    > veltas, if  high address lines are not connected anywhere, then of course it will wrap around
2026-06-01 17:26:26 <xentrac     > well
2026-06-01 17:26:46 <xentrac     > when we're talking about the 8800, "hardwired" decisions still weren't that hard to change
2026-06-01 17:27:06 <Stalevar    > also it was sold as a kit
2026-06-01 17:27:17 <xentrac     > exactly
2026-06-01 17:28:01 <Stalevar    > imsai's better though
2026-06-01 17:28:05 <veltas      > So I am asking then whether the high address bits are left unconnected on that 256 byte memory module
2026-06-01 17:28:36 <Stalevar    > Where can you connect it anyway, if you only have 256 bytes?
2026-06-01 17:28:49 <Stalevar    > I guess chip select via logic gates?
2026-06-01 17:28:54 <veltas      > Yes
2026-06-01 17:29:04 <veltas      > That's how backplanes tend to work
2026-06-01 17:29:17 <veltas      > So you can have multiple modules with one address space shared
2026-06-01 17:29:17 <Stalevar    > It could also be this particular sim bug / feature
2026-06-01 17:29:32 <veltas      > I'd have to check schematic
2026-06-01 17:30:46 <Stalevar    > Anyway, it seems that 86rk is much easier to type program is. You just do M or M addr (0 by default) and write the hex codes and then run it with G which can set a breakpoint
2026-06-01 17:31:10 <Stalevar    > And you can set registers with X
2026-06-01 17:31:33 <Stalevar    > https://rk86.ru/
2026-06-01 17:31:45 <Stalevar    > https://86rk.ru/
2026-06-01 17:31:51 <Stalevar    > Both host online simulators
2026-06-01 17:32:08 <Stalevar    > Or you could take emu80
2026-06-01 18:54:42 <Stalevar    > I'm pondering buying a cheap ($18) graphing calculator or not. I already have HP 39gs and don't use it, but the other one is interesting in that it's strange noname and has PROG button
2026-06-01 22:08:33 <veltas      > Stalevar: What are some good projects for the Altair 8800?