V-FORTH - Forth-83 for the VIC

[Forbidden64]

You realize of course, that this has historically never been done on any system...it is for all intents and purposes, a standalone FORTH OS because of these features no one else apparently thought of. Likely because they were too busy trying to adhere to a standard from 1976 made for tele-terminals(FIG). Incidentally, since V-FORTH doesn't have an editor, and also reads sequential files as though they are being typed...doesn't also then mean that any VIC word processor can be an editor, as well as you mentioned, any modern machine's text editors?
Or in essence, as you mentioned, any sequential file...? I'll do some testing and find out...like if quick brown fox can make V-FORTH compatible code structures, that would be interesting. Cool, would be an editor like take for instance a text editor made with minigrafik. :O.
Especially since they don't even need to be loaded at the same time. Although I am just hearing about minigrafik, and don't know much about it yet.

[srowe]

Things were different when the VIC came out. No one had a disk drive (at least in the UK) and the concept of using a standarized editor etc would have been quite alien. The integrated line editor that BASIC had was the model of the day.

In theory, yes, any word processor or similar program should do. From memory some of them do non-standard things, I think the editor that came with my assembler used characters 0x01 to 0x08 to indicate tabs.

I think a text editor that uses tokra's Maxi-Edit

http://sleepingelephant.com/ipw-web/bul ... f=2&t=8375

would make developing on a VIC much easier.

[Mike]

[...] I am just hearing about minigrafik, and don't know much about it yet.

The thread to the MG batch suite I pointed you to (in the post you're most probably referring to) contains links to a disk image and manual for MG, in the top post.

Besides this you're taking a good chance, that a program/tool/game in the Announcement section (especially those released by me, but there are also other users) sports MINIGRAFIK under the hood. MG has been around for over ten years now.

Cool, would be an editor like take for instance a text editor made with minigrafik. :O

This idea, a tool set of symbolic assember and text editor, hosted on the VIC-20, both working in a software 40 column mode, and capable of constructing themselves pops up here in Denial every now and then, see:

o First Step in Vic-20 Software Development,
o 40 column word processor?, and
o Programming language implementation ...

... including the reasons I didn't start out with that project.

[Bobbi]

For what it is worth, the little VICScript language I made last year has a primitive line editor which can be used to edit PETSCII SEQ files. Programs are loaded from plain text SEQ files. These files can be created in any text editor or word processor, so you are not stuck with my dreadful line editor! I chose to use PETSCII encoding but I could also have decided to use ASCII files so I could edit them easily on another platform. I am thinking to add an option to load and save in ASCII to allow files to be easily migrated from PC to VIC-20 and back.

[Forbidden64]

Grr, I have been so busy I haven't been able to toy around with this since my last post. I have been studying my book on the road when I get chances though. Hopefully, in a month or two things won't be as crazy...and I can set aside some proper time for this very cool new language. I guess having a VIC based editor is superfluous in the sense that, as the author pointed out, there are better modern alternatives. I just wanted to point out that this was an option theoretically. You could develop the entire thing directly on the VIC even though there is no formal editor.

All you would need is a 40 column word processor to make it comfy. That is, of course, independent from the language itself, which is already stand-alone...

I also found yesterday when I was searching around some archives, some very cool stuff. A 6502 & 65c02 cross assembler for dos, and unix. An 6502 object code disassembler for amiga(it's a ".1" file?), a 6502 assembler/ linker for the atari, a 6502 disassembler for UNIX, As well as a smorgasbord of other similar programs. I was having a quick look around to see if there was anything in there for my arrested LSI-11 project for later related to the LSX operating system, which is essentially a sysgen for UNIX 4.0, that fits onto an 8" floppy an runs with only 20kw of memory. Sadly, that project has been shelved for a while.

Also, apparently there is already a VFORTH...Vax FORTH, which is also in that archive. Apparently, this archive was ripped off of usenet circa '88/'89, along with even some message board content.
http://cd.textfiles.com/sourcecode/

[J.E.E.K.]

[..]
I meant to mention, the source and build can be downloaded from http://eden.mose.org.uk/vforth.git

Great stuff, nice work!

FIG Forth which this Forth is based on has historically a lot of bugs. Couple of weeks ago I went through the U/ or UM/MOD field and looked for implementations in various Forth sources for the C64, VIC 20 or even 65816.
Fine to see that your version looks pretty good and efficient (nearly the same as my version from the conclusion on the different sources I took into consideration). Nevertheless it has a bug ...

The original code (from the Git repo):

Code: Select all

 259 CODE  UM/MOD     ( 31 BIT DIVIDEND-2, -3,  16 BIT DIVISOR-1  *)
 260                  ( 16 BIT REMAINDER-2,  16 BIT QUOTIENT-1    *)
 261   SEC 2 + LDA,  SEC     LDY,  SEC 2 + STY,  .A ASL,  SEC    STA,
 262   SEC 3 + LDA,  SEC 1+  LDY,  SEC 3 + STY,  .A ROL,  SEC 1+ STA,
 263   10 # LDA,  N STA,
 264   BEGIN,  SEC 2 + ROL,  SEC 3 + ROL,  SEC 2+ LDA,
 265           CS  NOT  IF,
 266              SEC,  BOT SBC,  TAY,
 267              SEC 3 + LDA,  BOT 1+ SBC,
 268              CS IF,  SEC 2+ STY,  LABEL L442  SEC 3 + STA,  THEN,
 269           SEC ROL,  SEC 1+ ROL,
 270           N DEC,  SWAP  0=
 271       UNTIL,    POP  JMP,
 272       THEN,  BOT SBC,  SEC 2+ STA,  SEC 3 + LDA,  BOT 1+ SBC,
 273       L442 JMP,
 274 END-CODE

Line 272 has to be fixed to

Code: Select all

 272       THEN,  BOT SBC,  SEC 2+ STA,  SEC 3 + LDA,  BOT 1+ SBC, SEC,

The problem becomes visible, if the MSB from the dividend is set, this is handled as special case (line 272). So far so good. In this case only the remaining 16 bits taken for calculation (we know the dividend bit 17 is 1 and the one from the divisor is 0, the outcome is known). But the value of the 16 bit from the dividend may or may not greater than the divisor. So, the carry flag is not predictably set or unset after the subtraction. However, the quotient part needs the carry set which is rotated into the quotient later (line 269).

Test case:
HEX
0 8000 C000 UM/MOD U. U.
AAAA 8000 OK
should be the correct value.
Currently I got
0 8000 OK
As described above the missing carry (unset after subtraction of the divisor) is rotated into the quotient. The remainder is ok in this case.

[J.E.E.K.]

In addition to the previous one might replace the final JMP by a BCS which is in this case always taken.

Code: Select all

 272       THEN,  BOT SBC,  SEC 2+ STA,  SEC 3 + LDA,  BOT 1+ SBC,
 273       L442 JMP,

could be written as

Code: Select all

 272       THEN,  BOT SBC,  SEC 2+ STA,  SEC 3 + LDA,  BOT 1+ SBC,
 273       L442 HERE 2- CS NOT C, C, 

This saves the byte lost which "SEC," costs because the branch takes only 2 bytes instead of 3 for the "JMP,".
The branch opcodes are missing in Ragsdale's Assembler (we would need such as "BCS,") therefore we have to do the corresponding like the stuff in THEN, (I hope the offset calculation is right - not tested).

[srowe]

Thanks for reporting this, I did pick up a number of bug fixes I found on 6502.org but I don't recall seeing this one mentioned.

I think this is the correct change to handle both the missing carry set and switch to use a relative jump

Code: Select all

diff --git a/vforth.fs b/vforth.fs
index 84d989d..cdbf64c 100644
--- a/vforth.fs
+++ b/vforth.fs
@@ -265,12 +265,12 @@ CODE  UM/MOD     ( 31 BIT DIVIDEND-2, -3,  16 BIT DIVISOR-1  *)
           CS  NOT  IF,
              SEC,  BOT SBC,  TAY,
              SEC 3 + LDA,  BOT 1+ SBC,
-             CS IF,  SEC 2+ STY,  LABEL L442  SEC 3 + STA,  THEN,
+             CS IF,  SEC 2+ STY,  BEGIN,  SEC 3 + STA,  SWAP  THEN,
           SEC ROL,  SEC 1+ ROL,
-          N DEC,  SWAP  0=
+          N DEC,  ROT  0=
       UNTIL,    POP  JMP,
-      THEN,  BOT SBC,  SEC 2+ STA,  SEC 3 + LDA,  BOT 1+ SBC,
-      L442 JMP,
+      SWAP  THEN,  BOT SBC,  SEC 2+ STA,  SEC 3 + LDA,  BOT 1+ SBC,  SEC,
+      CS  NOT  UNTIL,
 END-CODE
 
 CODE 2/

Rather than use explicit branch instructions my code uses the standard ASSEMBLER conditionals. There's no check for pairings so with a bit of stack manipulation you can interleave BEGIN, IF, UNTIL, THEN, etc if you're careful.

[J.E.E.K.]

Thanks for reporting this, I did pick up a number of bug fixes I found on 6502.org but I don't recall seeing this one mentioned.

I think this is the correct change to handle both the missing carry set and switch to use a relative jump

Thanks Simon for the quick response. As I remember Mike Barry's latest version from 6502.org got this fixed too.
Just for reference, see the thread Re: integer division and multiplication

Rather than use explicit branch instructions my code uses the standard ASSEMBLER conditionals. There's no check for pairings so with a bit of stack manipulation you can interleave BEGIN, IF, UNTIL, THEN, etc if you're careful.

Ok, I see, that's python assembler without pairing you're referring to as standard assembler?

Did you pushed the changes into the repo already?

Johann

[srowe]

Thanks Simon for the quick response. As I remember Mike Barry's latest version from 6502.org got this fixed too.
Just for reference, see the thread Re: integer division and multiplication

Thanks for the link, I hadn't come across that thread.

Ok, I see, that's python assembler without pairing you're referring to as standard assembler?

Did you pushed the changes into the repo already?

Yes, the implementation of the ASSEMBLER conditionals in the python compiler mimics those in the Ragsdale code (just without the pair checks).

I've tagged a 3.8 release with this fix and pushed it.

Cheers

[J.E.E.K.]

Thanks a lot, Simon!

BTW, will you provide a link to the ZIP for distribution or upload the image here again?

Johann

[srowe]

Sure, I've uploaded both the D64 image and a zip file containing source, docs and binaries to my web site

https://eden.mose.org.uk/download/

[pixel]

This is mind-boggling. Never cared about FORTH before. Well done!

EDIT: HINT: There's open source code for bitmapped display and 40 columns one could plug in.

[srowe]

Thanks, I had a lot of fun learning the internals of a Forth VM.
I'm using the full 8K for the ROM-based version but I could look at adding 40 column mode to the PRG version. It would reduce the maximum size of the dictionary a little but would be more usable in a visual way.

[funkheld]

Hi good afternoon.

a nice vforth

why are the graphics-screen so small please?

How can you make it big, please?

is there a new version of vforth from 2019?

Thank you.
greetings from Germany