Re. note in DB.4th, "DEFINE UM* and UM/MOD"

[marktmiller]

Hi,

Not opening an issue, just a comment.

I got here through the Antic Atari 8-bit podcast "Are the Grandkids Listening?!". I've been working in APX Extended Fig-Forth for a while. I was looking through your code, and saw in DB.4th that you need implementations for UM* and UM/MOD. It just so happens I have implementations for both, but perhaps you've already been working on your own.

UM* is pretty simple (if I understand it correctly):

: UM* ( n1 n2 — d) M* DABS ;

I wrote up an implementation of SM/REM in Forth assembler. UM/MOD uses it. If you'd like, I can post the code for SM/REM and UM/MOD here in a comment.

[BartGo]

Wow, thank you (and thanks to Kay)!

I will appreciate any tunings for this Forth, as well as your FP words.

[marktmiller]

I'm remembering now, I wrote SM/REM in assembler, but realized after I got it done that APX Forth already has it, but calls it M/. So, you can just write an alias for it:

: SM/REM M/ ;

(n2 = remainder n3 = quotient)
: UM/MOD ( d1 n1 — n2 n3)
SM/REM ABS ;

I've been meaning to put my Forth code on Github. The FP listing is rather lengthy, and I need to do more testing. I ported all of the FP functions from Turbo Basic. Plus, I've rewritten, and added FP comparison operators, mostly in assembly. I'll see what I can do.

[marktmiller]

Hi Bartosz,

I realized last week that I had a significant error in some of the integer arithmetic words I'd written, including the two unsigned operators I put in my comments here: They don't work correctly at the full bit width they're supposed to. I feel rather embarrassed, because I misinterpreted what "Starting Forth" said about the specifications for these words; that the inputs "are unsigned." I assumed that meant you take the absolute value, and use signed operations to do the calculation. To some extent, this was correct, but it's not complete, because the true aim of the unsigned operators, UM*, and UM/MOD, is to allow the programmer to use all 16 bits (for single precision) or 32 bits (for double) to represent positive quantities! IOW, you ignore the sign bit. What I was doing was only getting correct results up to 15 bits for singles, and 31 bits for doubles. If you go above that, you get incorrect results. So, I'm presently working on fixing these unsigned words so that the full bit width can be used.

[BartGo]

Bugs around division in the fig-Forth itself were known - you will find different patches of related primitives around the web, like:

• http://forum.6502.org/viewtopic.php?f=9&t=7025 (Garth's fixes)
• http://6502org.wikidot.com/errata-software-figforth#toc0 (some assembly code suggestions)

I am not experienced in assembly and didn't test these so I can't recommend a specific fix but please be aware that fig-Forths have these known issues.

[marktmiller]

Bugs around division in the fig-Forth itself were known - you will find different patches of related primitives around the web, like:

• http://forum.6502.org/viewtopic.php?f=9&t=7025 (Garth's fixes)
• http://6502org.wikidot.com/errata-software-figforth#toc0 (some assembly code suggestions)

I am not experienced in assembly and didn't test these so I can't recommend a specific fix but please be aware that fig-Forths have these known issues.

Thanks. I had the thought I might need to resort to assembly to fix the issue, and I was kind of dreading that. However, some experiments I've been doing indicate I might not. I always appreciate finding code that does the job, though.

BTW, I posted some code I wrote to make APX Forth more compatible with the "Starting Forth" book. I posted it on my Github at:

https://github.com/marktmiller/atari-8-bit-forth-projects/tree/main/fig-forth1.1/starting-forth

An oddity I found in "Starting Forth" is it doesn't talk about a word to print out unsigned double-length numbers. So, maybe the maintainers of the Forth standard didn't think people would want to use the unsigned operators to produce final results?? Instead, it talks about the numeric formatting words. From that, it's pretty easy to come up with what's needed.

I saw this in Garth Wilson's post:

UM* (called U* in FIG-Forth) which is used in */.

Oh... (my original used M*). Anyway, the point is U* is buggy.

[BartGo]

I will probably need some of the words you have prepared.

I was playing with doubles in APX Forth recently, Alan Winfield's "Complete Forth" (on Archive.org) has this word to print unsigned doubles:

: UD. <# #S #> TYPE ;

[marktmiller]

I will probably need some of the words you have prepared.

I was playing with doubles in APX Forth recently, Alan Winfield's "Complete Forth" (on Archive.org) has this word to print unsigned doubles:

: UD. <# #S #> TYPE ;

That's what I came up with, lol! :) Even named it UD.. Only difference is I added SPACE to make it like .. I was looking through "Starting Forth," because I thought surely there's a word to do this, and all I came up with was the numeric formatting words, <# starts formatting for unsigned double, #S pulls all the digits off, and #> ends the formatting. There are other words used inside the brackets for signed output. These words save the string above the dictionary, and return an address and length, sufficient for TYPE. I gave it a try, and it worked.