14 comments

[ 2.8 ms ] story [ 38.2 ms ] thread
This is more like "by communicating with the X server", which is not exactly the level of "from scratch" I was expecting, but then again, it's more involved than the equivalent in Win32 that's not much more than filling in some structures and invoking a few functions.
This reminds me of the time I was a young, naïve idiot, and decided I would learn GTK by writing some basic app in GTK. In x86 (not x86-64) assembly. Like all of my other project ideas from that era, never went anywhere, and I horribly underappreciated just how complicated things were.

Kudos for actually getting somewhere in their attempt to do this, a further state than I ever managed.

When you don't know what you don't know...

In my youth, kids learned assembly to crack games.

Hehe, when I was a teen, I wrote a gameboy emulator. Was not fun
I learned X86 ASM by sinking my teeth into the Intel 8085 manual, then lighting up LEDs on a hardware emulator and later on a 8085 simulator that me and my brother built.

What certainly helped was that I had did some digital design and instruction set architecture, etc.

Later on, I did some real-world assembly programming for the PIC microcontrollers and some inlined assembly in C, which I did not find daunting at all because of my previous experience.

I guess the best prerequisite for this material is having done some low-level C, the kind where you know about text/data sections and being comfortable with calling conventions, the run time and the linking process.

Using X11 isn't exactly from scratch.
XQuartz does not support high resolution displays, that is not what macOS users looking for.
I was curious about the rep movsb, so I tried to compare this with how a compiler would copy a known small size chunk of memory. It seems pretty complicated. I didn't manage to make a compiler generate the `rep movsb` idiom but I managed to find out some interesting stuff:

- small, known size moves are stitched from a fixed number of mov instructions, sometimes overlapped. For example 21 bytes is qword (8 bytes) + xmmword (16 bytes), overlapped.

- char-copying loops like "a++ = b++ c times" with c not known at compile time are either realized as simple increase, compare, conditional jump, or, at higher optimization, a monster that has like 10 branches to use anything from xmmword to byte depending on the amount of data

- big, known size moves (> 256 bytes) just generate a call to memcpy

(on Linux/X11, which should have been mentioned in the title)
I have to say, writing a GUI is one of the last places I’d ever find it appropriate to use assembly language. Man, those click handlers are going to be fast.
Does anyone know of a C version of this, communicating with the X server without libraries?