Notices by Adrian Cochrane (alcinnz@floss.social), page 40

To output a digital image to a screen you essentially need to keep an X & a Y counter. Most games added a bit of extra circuitry allowing the CPU to configure a handful of coordinates which when reached by those counters triggers a configured image to start being sent out. These are called "sprites"!

Some early machines like the Atari 2600 were too cheap to afford a full sprite renderer, not handling the Y axis at all. Amongst other things...

3/4!

Designing your own hardware specifically for the game you want to create would've provided guidance for how to address this issue. Then "videogame consoles" would take the popular techniques & move them into peoples' homes.

Monitors of the day shot electrons towards the viewer to hit a phosphorescent coating, lighting it up. Magnets would direct these electrons to every point onscreen. Or maybe instead you hook it up to a control unit interpreting some "vector graphics" paths!

2/?

Over the next few days I think I'll write some threads about how computer graphics works. (I've started some videos on the topic, but those are a lot of work...)

Let's start with some of the early techniques!

It is the 1980s, maybe 1970s. You want to play with some of these new-fangled digital "computers" as a medium for entertainment. One problem: memory is expensive!

You need to generate a lot of data 30+ times a second in a way we're used to interpreting, & have nowhere to store it.

1/?

Now I've got generic height computation compiling successfully...

Next to implement positioning & & a high level abstraction, as well unittests...

I mean I can say a lot of things that are wrong, or frustrating, or nearly impossible to do well on Android. But dag nabbit, I have never had really awful bugs where a whole section of a pretty important page in settings or the Play Store won’t read. I mean I’m not even saying the grass is greener. The grass on Android is still very wild and uncut. But at least I can manage my subscriptions there without having to OCR the screen. I don’t know. I just think both companies could use an accessibility person in just about every team.

Ugh my gosh I can’t wait till Mastodon at least supports freaking Markdown so people can stop using mathematical latin italic capital u mathematical Latin italic lowercase n Mathematical Latin italic lowercase i Mathematical Latin italic lowercase c Mathematical Latin italic lowercase o Mathematical Latin italic lowercase d Mathematical Latin italic lowercase e Mathematical Latin italic lowercase t Mathematical Latin italic lowercase e Mathematical Latin italic lowercase x Mathematical Latin italic lowercase t.
Yeah. You feel that? The need to concentrate and pick out each individual letter and then reconstruct the word? That’s how I feel whenever I hear that in a username or post. You know what I do when I hear that? I skip it. Social media simply needs formatting tools. At least italics, bold, and underline. At least then screen reader will just read the post. Oh it won’t speak the formatting unless forced to cause blind people don’t need to know formatting /s but it’ll be bearable to listen to.

CatTrap is progressing! I'm now working on implementing tree-traversal & dispatching to the other layout formulas. So far I've got widths theoretically working...

#ArgonautStack

@fatboy The Pantheon desktop, elementary OS!

it's sometimes really annoying that I have to submit reproductions for bugs to open source maintainers in order to be believed.

I 100% understand why and it is very necessary to help them fix stuff.

What I find irritating is that some will insist that bugs I'm looking at on my screen *do not exist*.

Of course I'm the one filing bug reports and of course I'm aware of the discourse around all the free labor oss maintainers do. #theVent

@cancel @wim_v12e @csepp I do think the video addressed function-size as something that doesn't necessarily yield more optimal code in C++, but ultimately I agree.

I think OO got over-hyped leading to some bad software-engineering decisions which we're still living with.

P.S. @cancel I would appreciate your source to adjust my understanding. Unfortunately I've lot track of mine.

@cancel @csepp @wim_v12e I'm aware. Hence why I'd want to know before taking this too seriously...

@wim_v12e @csepp @cancel The biggest one that helps JITers is small function sizes, which are heavily reused.

Then again, polymorphism is especially tricky for JS to optimize away. Not that that stops JS engines from trying!

@cancel My point was that this is a terrible example to apply OOP. Because there is no benefit to the added complexity. So I'm not sure why you assume I did not pay attention…

Judging “Clean code” on a few multiplications is somewhat like judging a fish on its ability to climb a tree. Expect bad result.

I also suggested the quadrilateral example because it would make slightly more sense to have polymorphism there as it won't be described by just a height and a width.

@csepp @cancel I happen to know that not following clean code guidelines does deter JITers (like the major JS engines) from applying their optimizations, so caveat there.

And beyond that like @wim_v12e I'd ask about compiler settings in these experiments since both GCC & LLVM put a lot of work into ensuring don't slow down your program.

Finally as a computer scientist I'd advise that these sorts of wins aren't the biggest bang for your buck!

There's a support routine for looking up strings from the font, & amongst others. It implements Murmur Hashing itself. There's a couple lookup tables.

There's a methodtable for looking up glyph names, wrapping what I've described yesterday. And there's a methodtable for looking up fonttables from the parsed datastructures, this one exposes the face-loading method defined elsewhere.

3/3 for today. I thought I'd finish today, but... Should be tomorrow!

That accessor wraps one I've described yesterday, & directly exposes it as a method.

There's a method for looking up postscript font names, possibly (build & runtime) consulting the font-variation. Looks up an index then prefers Windows stringtable to Apple's, since it appears they couldn't agree in the early days.
This font-variation lookup is implemented in this same file, checking a few different places & running callback methods before checking hashes.

2/?

Hopefully finishing my studies of FreeType's SFNT support today, somewhat auxiliary I see amongst it's simpler subcomponents:

* Web Open Font Format decoding, with index-sorting & closing logic.
* Hook between LibPNG & FT_Streams, whilst premultiplying alphas.
* Loading SVG fonts with caching.

This SFNT subsystem as per usual exposes a methodtable to FreeType's submodule infrastructure. This exposes methods for retrieving the charset encoding & registry (everyone uses UTF8 now).

1/?

Talon is amazing - Xe: https://xeiaso.net/blog/voice-control-talon

L.A.’s Scoring System for Subsidized Housing Gives Black and Latino People Experiencing Homelessness Lower Priority Scores - Colin Lecher & Maddy Varner @ The Markup: https://themarkup.org/investigation/2023/02/28/l-a-s-scoring-system-for-subsidized-housing-gives-black-and-latino-people-experiencing-homelessness-lower-priority-scores

How We Investigated L.A.’s Homelessness Scoring System: https://themarkup.org/show-your-work/2023/02/28/how-we-investigated-l-a-s-homelessness-scoring-system

@theruran Ok , but I love unix/linux lol but I rather program in smalltalk, or Elixir, or something else. C is fun to reverse engineer but its a horrible language.