https://www.youtube.com/watch?v=Imquk_3oFf4
https://www.pouet.net/prod.php?which=85227

@tomasz your post reminds me of Wolfenstein 3D game



another one 32b:
https://www.youtube.com/watch?v=uTCnMpNHBIw

One thing that always feels clunky to me in such style coding is the non-reliability of starting conditions in DOS, which leads to having variants of code that work only in a very specific environment (or even worse - under a specific configuration of a specific emulator). If only MS/PC DOS had defined something like a standard values for registers at entry point of a program! But - of course - this would add another layer of legacy compatibility that any version of OS would need to provide, and in these days when every byte of memory still mattered, this certainly was not an option.

Even worse version of this problem lingers in the rules for PC Booter category of our 512-byte coding contest. At least at the entry point of a DOS program we know the values of IP and SP registers, while during bootstrap we are not even guaranteed to have IP=7C00h, as it might be 0 (with CS=7C0h). My TetrOS example needs 8 lines of code just to set up an uniform starting point because of that.

This gives me an idea - perhaps something to consider for some future contests - that we could design a well-defined starting conditions for coding small demos, and provide a universal launcher program that would execute the demos as such. There could even be separate but compatible launchers that could either boot or run from DOS command line (this reminds me of something I had done while l was still working solely under DOS - for testing boot sector code I had a small tool that could start a boot sector directly from a freshly assembled file).

Should/Can things be done to enhance disp8 addressing in terms of the initial execution environment?

I do miss PUSHA/POPA (MOVEM on 68k) in long mode.

256 bytes should be enough for anyone.

Funny enough, I have interesting case with my 512B MBR entry (not submitted yet)
It works in Bochs, Qemu, VirtualBox, Thinkpad T420

However it doesn't work correctly when executed with boot option in DOSBox.

Same code works fine in DOSBox if executed as COM file, or when loaded via Turbo Debugger

Relies on some initial register values except DL?

all the archive from this marvelous contest on pouet.net.

i feel admirative for those who coded some of theses incredible demos.

http://f00l.codercat.org/misc/256b_complete_demos_archive.zip

enjoy them

Thanks for hint. I initialized almost everything.... the one thing that I didn't was stack. I wrongly expected it to be initialized (haven't been coding assembler of almost 10 years).

Thank you very much

This is my favorite
https://www.youtube.com/watch?v=BzfsDfHqIM4

awesome one, download link:
https://files.scene.org/view/parties/2002/0a000h/256b/comatose.zip

running it on dosbox with the settings:
core=dynamic
cycles=max

and its still slow, but better than default dosbox configurations.

edit: and graphics are bad, i will run it on 32-bit machine whenever i have time.