Notices by Adrian Cochrane (alcinnz@floss.social), page 26

@garritfra Most of them land at: https://adrian.geek.nz/docs

Other popular techniques involves filming a starkly-lit surface texture, computing a Fast Fourier Transform to minimize the data being processed, & aligning the two images to see how far they shifted. And thus what data should be sent to the CPU.

Another popular technique is to have a grid of translucent wires, which might be placed atop the screen such that powering each row lights up the columns where a finger's capacitance overlaps it. Again requires (clustering) preprocessing.

2/3?

Having spent numerous threads describing how computers render visual graphics we can understand, how do they let us interact with those graphics?

The way pointers work at a hardware level is varied. Early models rotated a pair of wheels, hooked up to counters of the number of slots rolling past. Your scrollwheel still often uses this technique. That pretty much directly gives us the data we want but by now we've largely moved away from mechanical techniques to pure electronic techniques.

1/?

Some of those macros wraps Assembly code.

4.1/4.1 Fin for today! Tomorrow: Mini-GMP & probably MPF!

* Assertion support routines
* Convert from doubles (takes a fair bit of code to decompose the double)
* A C++ wrapper, to take advantage of operator overloading
* Fraction comparator*s*
* `2n%m` repeated p times
* `n == 0 ? 0 : m-n`
* Computing bitmasks to test for square-modulos of given number
* Fairly sophisticated iterator over (small) prime numbers, for internal needs
* Compute bitmask to detect prime numbers
* Headerfiles, include macros & inline routines covering many operators

3/3.1

@theruran Yeah, especially with the large amount of Assembly code it has!

Furthermore, I see:

* Raising SIGFPE exceptions.
* Default allocator callbacks.
* Scripts to generate lookuptables for division, factorials, jacobi symbol, fibonacci, etc.
* Wrappers around the division routines.
* Prime number testing (fastpath checks for 1, 3, 5, & 7 factors, otherwise uses slowpath loop for factors > 11).
* Another handful of routines to raise SIGFPE signals.
* Logarithms by testing against repeated bitshifts.
* Couple variations of dividing 1 by a number ("inversion").

2/?

Sometimes you want to avoid bugs caused by the finite number of bits a CPU-native number consists of. GNU provides a handful of libraries for dealing with this, & today I'll start studying GNU MultiPrecision (LibGMP).

Amongst the main codebase (I'll go over the subsystems starting tomorrow), I see:

* A handful of globals specifying memory-layout & version-numbers.
* Lookuptables.
* Utilities to get or set memory-allocation callbacks, offering reentrant or debugging values for them.

1/?

How Video Games Inspire Great UX - Scott Jenson: https://jenson.org/games/

A bug fix in the 8086 microprocessor, revealed in the die's silicon - Ken Shirriff: http://www.righto.com/2022/11/a-bug-fix-in-8086-microprocessor.html

Adventures in REPL Implementation - Nikita Prokopov: https://tonsky.me/blog/clojure-sublimed-3/

@tim1724 @TechConnectify From the sound of it, I don't think I'd blame Microsoft here.

I think the "firmware" code in some chip or other is to blame. There's a heap of it in any computer nowadays (since late 90s if I recall correctly), & there's bound to be bugs.

A rant on web font licenses - Manuel Moreale: https://manuelmoreale.com/@/page/AczvIzKXXPUQjhIz

Tech-last - Robin Rendle: https://robinrendle.com/notes/tech-last/

Very much my attitude!

I'll learn new tech when I see a personal need for it, I resent feeling pressured to learn them simply because "they're modern development practice".

Email - Explained from First Principles: https://explained-from-first-principles.com/email/

Took me a while to read this!

Commentary: Cory Doctorow: End to End - Cory Doctorow @ Locus: https://locusmag.com/2023/03/commentary-cory-doctorow-end-to-end/
Pluralistic: End to End (07 Mar 2023) - Cory Doctorow: https://pluralistic.net/2023/03/07/disenshittification/

I promise you ChatGPT can't access the internet, even though it really looks like it can - Simon Willison: http://simonwillison.net/2023/Mar/10/chatgpt-internet-access/#atom-everything

How to Under Engineer Your JavaScript - Go Make Things: https://gomakethings.com/how-to-under-engineer-your-javascript/

Looking up MX records for 203 million domains yielded around 30 million unique MX servers in around 21 million second-level domains.

But not every domain has an MX record. In fact, 119 million (58% of all) domains are lacking MX records.

Now many domains that include alt1.aspmx.l.google.com. as an MX will likely also include alt2.aspmx.l.google.com., so let's flatten these numbers by MX domain frequency, which breaks down our data set to 21 million unique domains.

The top 20 are: