SizeCoding.org: creating very tiny programs for the 80x86 family of CPUs | Hacker News Reader

SizeCoding.org: creating very tiny programs for the 80x86 family of CPUs

sizecoding.org

100 pointsingve15 comments

I can probably claim to be the inventor of 64 and 32 byte competitions back in my demoscene days.

I was organizer of a small demo party called 0a000h. 4k intro competitions were quite common, 256 bytes happened sometimes, I think we were the first to go even lower. We had some impressive results especially in the 64b.

We also did fast-size-coding competitions. We showed a simple effect and then asked people to come up with the smallest way to replicate that within a couple of hours.

Results are all linked on the webpage: https://www.0a000h.de/

Most of them should run in dosbox, unfortunately some make problems.

It brings smile to my face when I see someone wearing an Amiga ball logo on their shirt :)

There is also this (unfortunately discontinued) competition series, with a lot more tiny programs to examine and maybe even challenge yourself against:

http://www.hugi.scene.org/compo/compoold.htm

These are amazing. All kudos to the authors. This is what CS courses should start with.

Any CS course that includes Asm often includes writing tiny programs, but not tiny programs that do interesting things.

Tiny assembler programs tend to abuse arcane x86 instructions that have one or two byte encodings, but aren't practical for real use (and compilers don't emit them). Add some binary format tricks on top.

A little bit of assembly writing (and more importantly, reading) should be a part of a good CS curriculum but these tiny programs aren't a good place to start.

This seems fairly focused on Windows? Would be interested to see some Linux stuff.

It's not Windows but MS-DOS, which allows for pure binary executables (COM format) that are nothing but instructions. Thus even a single-byte file is executable (C3, RET - i.e. exit immediately.)

AFAIK Linux has a higher minimum size, as the ELF header alone has to be at least 45 bytes:

http://www.muppetlabs.com/~breadbox/software/tiny/teensy.htm...

One could write a com loader for Linux pretty easily.

com programs assume they run on DOS, so you might as well use dosbox for that.

Linux supports some other executable formats for embedded systems, but they don't have much less overhead than ELF.

The problem is that a lot of the demo programs are hooking Int 21h, which is provided by DOS / Windows.

Would it run 16 bit code under x64 Linux (w/o any translation / emulation)? As Windows can't do it (only x32 versions can), thus wondering, if this is a CPU limitation.

Indeed https://github.com/avian2/linux-com-loader

My point was around the DOS Int 21h, really. But I see what you mean about the file format.

Use int 80h for UNIX

Reply on news.ycombinator.com