Martin Atkins

he/him; I live in Portland, OR and get paid to write software. Some time ago I worked on federated social web protocols but I've not been actively involved for several years. Increasingly skeptical about technology. 🙃

Pronouns: he/him

Public Key

npub1ej2ahxcsam4vw2h06j2sfv3p2e8wv2m0ue46hvatm6u4fnn90gaq068d64

NIP-05 Address

apparentlymart_at_mastodon.online@momostr.pink

Profile Code

nprofile1qqsve9wmnvgwa6k89thaf9gykgs4vnhx9dh7v6atkw4aaw25eejh5wspz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsjzujjq

Publishing to

yabu.me

nos.lol

relay.damus.io

Author Public Key

npub1ej2ahxcsam4vw2h06j2sfv3p2e8wv2m0ue46hvatm6u4fnn90gaq068d64

Show more details

Last Notes

2026-02-24 01:51:03 UTC

- reply

With how things are currently set up, folks who make that much money through wages today will probably use them to buy stocks and similar, making them also owners.
And at some point the returns from those securities become a significant portion of ongoing income, which is perhaps part of why this category feels like a grey area? They are in transition to being owners rather than workers. 🤔

2026-01-24 18:46:23 UTC

I don't want to hear politicians talking about going back to some ill-specified better time in the past, regardless of what party they belong to.
I want to hear them talking about creating a world that's better than it ever has been before. With specifics.
Tell me exactly how you intend to create the best world we've ever had, and ensure that it keeps getting better.

2025-01-05 19:37:51 UTC

- reply

I am inner northeast.
Rarely a day goes by that I get across the street outside my house without putting myself directly in the path of a vehicle to get the driver to actually stop.
On some extra special days they even cross into the opposing lane to pass me even though I'm already walking across the street.
Despite some investment in basic bike lanes a couple decades ago, this city is still pretty car-centric and drivers don't seem any less entitled here to me. 🤷🏻‍♂️

2025-01-05 18:11:34 UTC

- reply

I guess you live an entirely different Portland, Oregon than I do. 🙃

2024-11-26 15:23:46 UTC

- reply

It seems to me like the memory access itself is the main bottleneck either way, so it might not make much difference to speed but does free up a CPU core to be doing some other work.
But of course you might not have any other useful work to do, which is what I meant in my opening about it potentially not being worth it.

2024-11-26 05:34:13 UTC

- reply

I dunno if it's worth the trouble when you have four CPU cores just sitting there anyway 😀 but FWIW the SoC's DMA controllers are quite useful for this sort of copying. Some of the channels have a "2D mode" that supports copying multiple chunks with a specified stride, exactly as needed for copying dirty rects.
Circle seems to have a helper CDMA4Channel::SetupMemCopy2D for getting the control blocks correctly configured for that kind of operation, which is neat.

2024-11-23 02:48:25 UTC

- reply

I keep distracting myself messing with everything else except this so I've only played with it superficially, but I thought this looked promising:
https://gitlab.com/bztsrc/scalable-font2

2024-11-22 16:54:27 UTC

- reply

When I was still doing a bare metal thing (before I switched to running atop a Linux kernel) I had things set up so that the extra cores were essentially pretending to be separate "chips":
Core 2 dealt with async blitting/composition and some other misc graphics work
Core 3 dealt with audio mixing/etc
Core 4 dealt with other misc hardware stuff, like interrupts from the USB controller
This was more for fun than utility though. Aesthetically similar to custom chips but as software. 🤷🏻‍♂️

2024-11-11 21:12:08 UTC

- reply

One thing I've run into on my project:
Some applications want the abstraction of a "pointer", with an absolute position in the same coordinate system as the graphics.
But other applications just want to know how much a mouse/trackball/etc has moved relative to the last time they asked.
I have an explicit "mouse mode" in my API which let's the application choose which of those treatments it wants for the mouse, hiding the pointer when in relative mode.
I have no window system, tho

2024-11-09 21:18:51 UTC

- reply

Admittedly I made my life a bit harder by using Rust for my bare metal bits, so it took some fiddling to even get my program to start running. Starting with Circle would probably have been the better option, but... I dunno, I just feel like writing Rust right now. 🤷🏻‍♂️

2024-11-09 21:14:36 UTC

- reply

That JTAG/FTDI HAT I'm using is my own design, here:
https://github.com/apparentlymart/raspi-jtag-hat
I originally made it for a Pi3 (hence the caption on the silkscreen) but it seems to work just fine on Pi4 too.
A little fiddling with config.txt gets the JTAG pins set up by the firmware so you can halt early in the OS.

2024-11-09 21:12:29 UTC

- reply

FWIW, I found bare metal on the Pi4 generation considerably more efficient once I reconfigured the bootloader to netboot (from a tftp server on my dev machine) and attached a JTAG interface to it so I could step-debug my early boot code using OpenOCD.
I've never used a Pi400 so I don't know if all of this is transferrable but attached is my CM4 setup (with Ethernet unplugged right now just because I'm working with a RISC-V board currently)
https://files.mastodon.online/media_attachments/files/113/455/036/201/952/654/original/07bedecc26037a57.jpg

2024-11-05 23:50:40 UTC

- reply

I previously said that it seemed like you and I were both working on fantasy computers with different goals, but actually it seems like our goals are remarkably similar. 😀
The great thing about fantasy computers is that there are so many different ways to arrange the details! I'm excited to try out Capi and see what's similar and what's different.
(I will share more about my project at some point; I've just not had so much time to work on it lately because I just started a new job.)

2024-10-26 14:30:30 UTC

- reply

I might still be misunderstanding, since the math involved is a little beyond my current capacity for quick learning 😀 but the sense I am left with is:
In principle an optical flow sensor could sense rotation, but to avoid too much extra ambiguity it would need to be instead of rather than in addition to translation. So I guess you could get both with two sensors and different analysis of each, but you could just compare two translation sensors at that point anyway.

2024-10-26 14:22:05 UTC

- reply

I guess what was nagging at me is that if the optical flow sensor starts with a grid of light intensities then it already has "multiple sensors" in some sense.
After some more reading today I've learned about the "aperture problem", which involves an effect similar to a spinning "barber shop pole" appearing to move downwards because you can't see the edges of the stripes for reference.
So I guess the sensor is too localized to reliably distinguish rotation from translation.

2024-10-26 03:10:09 UTC

- reply

Unfortunately I was not able to find any definitive answer. I have seen some claims that it's hard to distinguish rotation from translation in optic flow, but couldn't get a sense of why. 🤔

2024-10-26 02:55:01 UTC

- reply

That way of describing an optical mouse made me wonder:
Could an optical mouse sense rotation in addition to translation, as a different way to manipulate objects?
(I guess you'd need to deal with what happens if you need to press a mouse button while it's rotated 180deg. 🤔)
I'm gonna go try to learn the answer and I'll report if I learn anything interesting. 😀

2024-10-25 19:59:50 UTC

- reply

We've already seen that my motivations are a little different than yours, but FWIW on my own fantasy computer I'm using 640x360 for the terminal and code editor (though the latter is currently _total_ fantasy) but game software defaulting to 320x180 for the sake of the very chunky aesthetic.
(It also kinda reminds me of developing on my Amiga back in the day, where my editor was on a hires screen while my game was typically lores. But without the rectangular pixels in hires. 😀)

2024-10-24 01:56:29 UTC

- reply

FreeBASIC is written in FreeBASIC but I believe it can transpile to C and its native code output uses the C ABI so maybe possible to adapt its compiler to be a library instead of an executable by one of these two methods. 🤔
(I have not tried, yet.)

2024-10-23 03:16:18 UTC

- reply

In particular I made exactly the same screen resolution tradeoffs. It's funny that it's called "widescreen" and not "shortscreen" since it typically means removing rows, not adding columns.
I intend to have it running as pid 1 on top of a Linux kernel first for driver reasons, but trying to design it to run on real RISC-V hardware later on.

2024-10-23 03:14:11 UTC

- reply

I hinted in a response to a question you asked a couple weeks ago that I was working on something and what you've described overlaps with it considerably. 😀
The main exception is that the "native language" of my thing is RISC-V machine code. But coincidentally Lua was the very first real program I ported to it, so I guess I could make some bindings to the graphics, sounds, and input APIs to be more like the other fantasy computers/consoles.
Not much time for it right now though. 😖

2024-10-13 20:28:46 UTC

- reply

Since you said "distribution" I assume you mean something a little more substantial than that... I'm curious where you would draw the line between distro and application in your ideal solution to this!

2024-10-13 20:27:17 UTC

- reply

I am in the early stages of building something like this, though what I'm building is tailored just for my one application so I don't think it would actually quite meet your description.
My program _is_ "init" and it runs a few other small programs, like wpa_supplicant and a DHCP client, as child processes of itself.
My program is a "fantasy" homebrew computer using the Linux kernel mainly just for its drivers, since that's usually the closest thing to hardware docs these days.

2024-10-11 14:46:20 UTC

- reply

Ah, I just found your other post that confirmed using USB, and now I also spotted it in your photo! Sorry for the noise.

2024-10-11 14:45:06 UTC

- reply

Do you have the scope connected with USB?
Looks like we both have the same scope and last time I tried similar unfortunately the Sigrok driver for it seemed quite crashy, but that was quite a few years ago now. Your positive report has inspired me to try again soon. 😀