raphlinus

I'm also noodling on an emulator, shooting for cycle accuracy (to the point where you'd be able to run vapor lock demos). Mine passes the 6502 functional tests but not yet the Tom Harte suit…

They are related, thanks for the pointer! The representation of the tree is very similar, as are the techniques to get better than O(log n) - the work efficiency stuff. I'll add a reference to th…

The stack monoid is related to the bicyclic semigroup (which got its name before "monoid" was more common, though I do see both names), but I am drawing a distinction: the bicyclic semigroup…

Not sure why this is popping up now, but people should know there's been quite a bit of followup work. One is a draft paper (not yet polished for publication), on arXiv[1]. I have an issue on my …

This is a good reason to bit-bang the protocol. It is in fact possible to get HDMI including audio out of an RP2040, as is demonstrated by infones[1], a NES emulator. [1]: https://github.co…

It needs to be done in the renderer. I think it's doable though, the FidelityFX library looks like it can be ported, it'll just run a bit slow because of the lack of subgroups. This particul…

I can perhaps refine (1) and provide some insight in support of (2). It does make intuitive sense for handles to be 1/3 of the length of the curve, as that makes the "speed" close to co…

The goals look quite similar to my hyperbezier explorations. There's an online demo of the first draft[1], and a Zulip thread on some progress toward a second draft[2]. The tl;dr is that I think …

I'm also looking forward to it. One of the big challenges is the sorting, for which I'm unaware of a good WebGPU implementation. I have some more notes on this question in a Zulip thread[1].…

Right, in relatively recent versions of LLVM, there is a "freeze" operation. Making it available to programmers, possibly as a method off MaybeUninit, would potentially be useful. The safety…

I suspect it may be: https://www.youtube.com/watch?v=5I4YPkVU7mY I'm interested in this because these linearized tree representations lend themselves well to processing on GPU. W…

Last I checked[1], Swift JSON performance was terrible, well over an order of magnitude slower than serde. It's possible it's been improved since then but if so I didn't find any announ…

Amen. When someone does the math and adds up the winners and losers in all this, one clear winner will be this video from Asianometry, entitled The History of Superconductors (Before LK-99)[1]. It onl…

So the HP 2645 did use a half-bit shift and it's quite clever. It's shown in this video[1] by CuriousMarc (I've set the timestamp to the relevant part). The document shown in the vide…

To a first approximation, Kompute[1] is that. It doesn't seem to be catching on, I'm seeing more buzz around WebGPU solutions, including wonnx[2] and more hand-rolled approaches, and IREE[3]…

I think we'll get there but it will take years to arrive at a robust solution, because of the complexity of the ecosystem. The cooperative matrix extension, the portable and standard way to do wh…

I'll give some context. Shaping is converting a string of unicode code points into a sequence of glyphs, each with an advance and also finer positioning. For typical Latin fonts, shaping is strai…

I think the suitable followup to this is the Richard Feldman talk, "Why Static Typing Came Back."[1] It gives the major reasons why one would prefer one over the other, and makes the case th…

I love flattened ASTs. One of my favorite uses of them is for inline markup in pulldown-cmark (see [1] for a brief description). Here's a sketch of the problem being solved. The raw input is brok…

https://en.wikipedia.org/wiki/Johnston_(typeface) Agreed, it's extremely distinctive.…

This is something that fundamentally can't work, unfortunately. One showstopper (and there may be others) is subgroup size. Nvidia hardware has a subgroup (warp) size of 32, while Intel's su…

It really depends on your workload. Prefix sum is pretty important, but you can also work around the missing barrier by doing extra dispatches. You'll want to do that if your goal is portable cod…

My personal feeling is that higher level layers should be built on top of strong lower level layers, and we don't have that yet. Scan is of course useful, and there are a ton of them in Vello, bu…

That's correct, it's the memory scope that I expect to be device-scoped. GPUs tend not to have execution barriers in the shader language beyond workgroup scope; generally the next coarser …

If you're doing advanced compute work (including lock-free data structures), then it's best effort. https://github.com/linebender/vello/issues/42 is an issue …

The most complete documentation is in the applegpu repo[1] by dougallj showing a great deal of recent activity (including by alyssarosenzweig). Last I checked, the documentation of barrier instruction…

Vello is progressing well, it's most of what I spend my time on these days. I can't promise a date for a stable version, but I would say that for the most part we're on our roadmap. I&#…

Toastytech[1] has a great many early Mac screenshots, as well as many others from around that time. [1]: http://toastytech.com/guis/macos1.html …