‼️ = Must read!
🧐 = Interesting, curio.
Retrocomputing (might contain emulation).
- A 7xxx-series NVidia GPU on Win 98.
Required some patches around IRQ handling and memory banking - interesting that there are people writing them. This video also shows an interesting fact I've noticed - you might think that's possible to run old software at crazy speeds on modern hardware, and in general that is not actually the case. GPUs though are an exceptions, as the programming model implies lots of good parallel execution, and thus performance is less bottlenecked by old code, interfaces et al.
- Sprite graphic tricks explained. 🧐
Analysis of some nice tricks in vintage games.
- Vic20 port of Ultima.
- Bendix G-15 and other weird architectures.
If you have not watched Usagi's electronics series on the Bendix G-15, you're in for a treat! His latest video made me curious of looking up the architecture more in detail - which in turn jogged some memories of the mill "belt" CPU (seemed crackpot back then - still now - at least if we want to compare it with "big" CPUs, not DSPs etc) - which in turn made me find Burroughs Large Systems stack architecture. Any of these would make for some cute zachlike games.
Miscellanea.
- #rtr ! A virtual goniophotometer for BRDFs.
This is something I always wanted to do! There is some code as well, so I might play around with it. Neat stuff.
- #rtr Teardown teardown.
see also this one... Old but gold. It's particularly interesting (to me) how relatively simple teardown's rendering ended up being. The power of a bespoke engine, done with purpose.
- #coding Documentary on the history of Python. 🧐
From academic failure to success, starting from a small but real (and close to you) use-case. Also shows the importance of being able to access a community. And fun! Open-source but not too restrictive. Matrices one of the first packages, Blender implemented it very early on! "because the people were nicer".
- #coding Life of an instruction in LLVM.
- Aras-p random unity thoughts.
Almost a year old now but I just managed to read it. Good stuff, I think Aras is right on the money there - and I think his vision would be lovely, but at the same time as he recognizes, it probably is not the kind of stuff VC is made of (or now public markets) - which is the huge conundrum Unity was/is stuck with, one that has no (easy) solution.
- TinyTapeout. ‼️
Crazy. A wasm gate-level simulator, plus a visualizer of the "tapeout", for small (but not that small) VGA hardware "toys".
- Dicing an Onion, the Mathematically Optimal Way.
Quite far from optimal. A good example of the perils of simplifying too much, which should be familiar even to an elementary student. Most obvious is that it does not consider that this is a statistical problem, layers and cuts have variability and the optimal technique is one that minimizes sensitivity to this variance. Also, 2d in this case is not the same as 3d (or more).
- =?utf-8?Q?#math_Intuition_for_Pick=E2=80=99s_Theorem?=.
Feels like this might be useful one day for graphics
- On the "naturalness" of FoV. 🧐
Got triggered by "Moments, just as your eye sees them" - and many reviewers parroting how 43mm is close to how we see. It is most definitely not - rectilinear images are "natural" when you view them at the same FOV as they captured (i.e. the display or printed paper occupies in your visual field the same that was captured) - so it's all a matter of perspective. That said, this led me again into a rabbit hole of visual perception. I first wondered if anyone created an accurate simulator of how scenes project onto the retina (assuming some reasonable parametrization of its surface). I didn't find much about that, but I did find simulators of the retina itself, which seem neat. Of course, eyes are weird and vision is super-weird, and there's really no way to create an image that in general would be "the most natural". At best you can decide which aspect of sensation/perception you want to prioritize, and create something natural in respect to that. Or, I guess, you can ask people what feels most natural in a variety of viewing conditions and go from there. Chances are that you'd learn what artists did. This is super interesting - and everything else from Aaron Hertzmann.
- Some thoughts on LLMs and Software Development.
Some good points. All tech innovations come with bubbles - that doesn't make them useless. Hallucinations are a feature, not a bug
- Graph Theory in State-Space.
Fantastic visualizations of the state space of a 2d puzzle game, bringing some insight on the structure of its solutions.