I've been working on porting GWMON-80 from Intel 8080 to NMOS 6502 (naturally, called GWMON-65). Part of the requirement for the port is trying to fit GWMON-65 in 512 bytes, just like GWMON-80. There's a practical reason for this: I have 6502 systems that use 1702A 256-byte EPROMs!

I've got the monitor basically ported and started working on shrinking it down to meet the goal. Originally it was 13 bytes too big, and didn't include the vectors since development has been done at 0x1000 in RAM. Going through the source, I noticed that there was a lot of space consumed by strings in general: the actual strings and printing routine of course, but especially in the setup for calling the string printer. I was loading two zero page variables with the pointer to the string, which requires eight bytes by itself! So, I came up with a way to get that down to two bytes:


Do note that the high byte termination is being stripped in the COUT routine.

Now, the string printer is called like this:


The indexes are calculated at assembly time, so that makes this kinda ugly solution a little more bearable. An index needs to be one less than its actual offset from the start of the string block, as this makes the string printer slightly shorter. The ^ byte cast operator had to be used on the first offset since it's a very large negative otherwise.

Anyway, I thought I'd share since I haven't run across this solution in particular in any old documents, and a cursory search on the Internet didn't reveal anything exactly like this. It's obviously limited to string blocks that are not longer than 255 characters, which does make the high bit termination more attractive. I can't see using this other than for very space-constrained applications.

Not a bad little approach, though, if space is at a premium and you don't need too much text

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

I am not sure if it is normal for a ROM print routine to preserve the processor bits on return, but I don't think it should be taken for granted. Normally it is done this way.

PRTSTR: INY ;Point to start of string
LDA STRNGS, Y ;A = next char in string
BMI DONE
JSR COUT
JMP PRTSTR

DONE JMP COUT

One mans normal is another mans "Huh?"

Today I mostly have in-line strings because I want to:

PRINT "Hello, world"

so rather than "load address of string into register ; call print", I:



If I were tight on space I could do the top-bit set thing to terminate the string, but I'm not, but on a system I worked on some 40+ years ago where I wrote a lot of 8080 ASM code was (initially) right on ROM space, so I had to have a plan B, so I made strings containing tokens where each token as a lookup to a table of words, so I could re-use words to make sentences out of the most commonly used words or phrases. It was tricky, but I semi-automated the process by machine generating the data where I could.

So... give 10 programmers a problem, get 11 answers still applies

-Gordon

_________________
--
Gordon Henderson.
See my Ruby 6502 and 65816 SBC projects here: https://projects.drogon.net/ruby/

I'd seen your tutorial before, and encountered the inline strings method in the wild when disassembling some Ohio Scientific code! It ended up being somewhat longer in this particular instance, but was something I considered/tried. Neat trick, too!



I'm with drogon...what? That method is more bytes and doesn't strip the high bit, which will produce garbage on some (many) serial terminals when running in 8N1 mode.

I get what Rob is saying. There is a dependence in your code for COUT to preserve the N flag. His solution addresses that potential issue. I did something similar years ago:ORMASK would be defined as $80 for an Apple ][ style COUT and $00 for most others, so if you are really pressed for space you can eliminate one of the instructions that uses it and save two bytes. The PUTCH label was a sneaky way to print a single lead character before the string, or just the single char if its high bit is set.

_________________
Got a kilobyte lying fallow in your 65xx's memory map? Sprinkle some VTL02C on it and see how it grows on you!

Mike B. (about me) (learning how to github)

Ah, yeah, I didn't show COUT, which does preserve the A register. Last thing it does is pull it back off the stack. Required for other parts of the program, so it's a freebie optimization

Nice! I thought of something similar for a text-heavy program that used BRK instead of JSR to save two bytes, and packed the strings as six-bit chars (or maybe it was five-bit chars and the carry flag to indicate which set to access). I never finished that project, and no one missed it.

_________________
Got a kilobyte lying fallow in your 65xx's memory map? Sprinkle some VTL02C on it and see how it grows on you!

Mike B. (about me) (learning how to github)

if COUT is in your code I'd do this:



You can even have that .Y is the exact offset and do this:

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

FWIW: the Acorn MOS (BBC Micro) uses BRK to indicate an error of sorts and includes the error code and message in-line. So:



Adapting that to a generic string print would be trivial - unless you wanted to use BRK for anything else.

-Gordon

_________________
--
Gordon Henderson.
See my Ruby 6502 and 65816 SBC projects here: https://projects.drogon.net/ruby/

I'd thought about doing that, but I will likely split COUT from the main monitor as I did with the drivers paradigm for GWMON-80. Right now CIN falls into COUT to effect echo, but that will likely get split up too, since CIN stays in the main monitor (a CIN without echo goes in the driver).



Considered this one too, but I figured a monitor shouldn't take control of BRK -- might want to use it elsewhere, or for actually setting breakpoints!

I thought that null terminated strings were awful. Now we have negative terminated strings! To make the string one byte smaller and make the loop two bytes smaller!



It might be useful to split uses of BRK:

• BRK zero: output literal string.
• BRK positive: error message.
• BRK negative: breakpoint.

_________________
Modules | Processors | Boards | Boxes | Beep, Beep! I'm a sheep!