http://wikiti.brandonw.net/index.php?title=83Plus:Hooks:Catalog

Well, when opening the program editor from ICE, it leaks memory, about 13K, way too much
Can anyone help me please? I used the routine from https://www.cemetech.net/forum/viewtopic.php?p=171310#171310 and this is my code:
Code:

So, to double check, are the steps something like this?
1) ICE decides to edit a program.
2) ICE invokes the routine above, jumping _Mon.
Do you try to catch the change-of-context out of the editor, or do you assume that the OS now has control forever? The obvious answer is that if you're handing control over to the OS, need to _DelMem size_of_ICE - 2 bytes at UserMem. Another option would be to exit ICE, but leave a key hook installed that will delete itself and launch the editor as soon as the cursor starts flashing at the homescreen again.

Edit: It's worth mentioning that the _DelMem at UserMem approach only works if you're positive that the OS was the one running your program, not a shell. The OS copies your program to UserMem, while shells swap the program to UserMem, allowing you to run programs with very little free RAM. The hook approach might be preferred in case the OS/shell/whatever has other cleanup tasks it wishes to run.

I'm curious about linking at the TI-84+CE. Let's say I want to get some bytes from the linked calc, how would I do that? And how would I even check if a calc is linked? Are there some calls for it? Thanks in advance

EDIT: Mateo gave me one call that checks if the USB is powered, _usb_BusPowered, and he said there is another call somewhere in the OS that can write/read bytes, but he didn't explore it yet. I don't really mind, because I can't even test things, since I don't have 2 CE's..

I'm struggling with Z80/ADL mode and prefixes etc. When I'm reading the documentation, I see (all in both Z80 and ADL mode):
call
call.sis
call.lil
call.is
call.il
but where are call.lis/call.sis and call.lil/call.sil? What happens if the CE triggers such opcode (in both modes)?

This is not really ASM-related, but I have a question about spasm. I have a lot of sprites (AoCE) which takes up one giant appvar. All these sprites are (compressed and) compiled with ConvPNG. But I need to display these sprites, and thus I need to know the offset in the appvar. So my idea was to include the sprite data in the main program, such that the names/widths/heights are defined, but not included in the output. I tried .nolist and .list, and that works fine for the labels, but the sprite data is still included. Can someone help me?

You could have an offset table at the beginning of your appvar, pointing to where each sprite starts

That will work, indeed, but the sprites might change a little bit, and updating the offsets each time is a hell

You could write yourself a script to create dynamic labels for each sprite, like _sprite_offset_<num> and then in the offset table thing do like

Code:

Indeed, a table containing offset expressions computed by the assembler is simple and makes perfect sense, especially when coupled with a C enum to define the indices

A more complex approach would be to use some form of archive file with an index and a bit of extra metadata for every chunk, in the same vein as ZIP files / tarballs. There's precedent in the TI community: for instance, TICT eBooks for the TI-68k series are stored as an introductory image, then multiple eBook parts, into a so-called "TTArchive" container supporting optional per-part compression with PuCrunch, created by a computer-side generator driven by a config file.
However, given that the approach listed by Sorunome works beautifully for your use case, I won't elaborate further unless you want me to

Okay, I made a PHP script that does this automatically, and generates this output:


Code:


<serious stuff> Now that I have everywhere offsets, is it better to have a relocation routine, like the C libs, or a table containing the offsets? I prefer to have a relocation routine, because it is slightly faster during the game itself.

"Relocation routine", as in embedding an array of pointers into your program's image, and at startup time, storing pointer values computed from the base address of the variable (which shall hopefully either not change during one run because of archive memory garbage collection, or be automatically recomputed after an address change if you have a way to catch GC events...) and the contents of the offset table ? Yeah, sure.

What are these events and can they happen during the execution of an ASM program?

On the TI-68k/AMS platform, archive memory GC can occur during the execution of an ASM program, when said program wants to archive variables (ROM_CALL EM_moveSymToExtMem), for instance its own savestates, but there's no more suitably usable space within archive memory. Or when calling the EM_GC ROM_CALL directly, of course.
Admittedly, my memory is fading, but I'm not aware of a direct way for the running ASM program to receive notifications from the system when a GC occurred (though I can imagine some dirty and/or heavyweight schemes). Correct programs need to be written in such a way that they're resistant to both RAM heap compaction and archive memory GC, and the main way to do that is to use HANDLEs instead of SYM_ENTRYs.

On the TI-Z80 and TI-eZ80 platforms, I can't see why archive memory GC couldn't occur during the execution of an ASM program, for the same reason. And people even more knowledgeable in TI-(e)Z80 calculators than you are, which is far more than I am, can tell you what happens on these platforms. I was just attempting to point a potential limitation of a scheme based on direct pointers to data in some cases: you might have to recompute after archiving variables, or something like that

So basically such event would only occur if the archive is full?

Other spasm question: like with the C libs, I use \.r , and now I need to know how many times this macro is called. In the C lib, I see something like #define numrelocations numrelocations+1, which works fine for me too, but if I have somewhere in the code ld b, numrelocations, it sets it to 0, causing a crash.

You wouldn't happen to have a multiple of 256 relocations, would you?

If not, then spasm's up to its usual weird define scoping. If you want the value of numrelocations as it would be at that instant, eval(numrelocations) may help. Otherwise, I'm not sure what to suggest without more context.

In notepad++ is it possible to highlight all the label names? Such as highlighting everything before the ":".

Do you mean inside a User Defined Language?
If so, yes it is possible, as long as you have an underscore starting the label (I haven't figured out if you can without an underscore). If you edit your defined language, and go to the tab called "Operators & Delimiters", choose one of the boxes to use it for. Add these things so it looks like this:

Open: _
Escape:
Close: (( : EOL))

That will do it for you. Obviously, change the Styler settings for highlighting.

PT_ asked for help with a 24-bit ez80 sqrt routine on IRC. I soon realized that my 8.8 z80 sqrt routine already computes a 24-bit sqrt, so I just made some changes to that:

Code:

What is the fastest way to check if signed HL is in a signed range?
i.e. HL = 2 in range [-1,8] should return true, but range [3,12] and [-20,-11] should return false.

EDIT: Runer solved it. Basically move the range to [0,X] and then unsigned compare HL with X.