raphlinus

Thanks, this is helpful. The specific method that's causing me trouble right at the moment is "computeCommandEncoder"[1], which is a method off MTLCommandBuffer, and I think not in the …

I totally get where you're coming from, Metal is much easier to program than raw Vulkan. I think a Metal-like API over "modern" APIs could be a good thing. There are things like the Vul…

Right. Would you happen to know if there's a way to make a retained object actually deallocate on last "release," other than putting an autorelease pool around it? Would you also happen…

"Subset" is not the right word here. Vulkan has pointers now, but (as was discussed in a recent thread), there are serious limitations compared with "real" C++. At the same time, M…

Along similar lines, from my skim of the README, it also seems like deallocated things are not actually released until the return from the nearest autorelease block, which is not exactly convenient fo…

+1 to GPGPU specifically. It feels like there is enormous opportunity here, as existing languages have serious limitations (CUDA is effectively Nvidia-only, choices for compute shaders are extremely l…

This is consistent with my experience as well, as a researcher in industry with a fairly high open source profile. I get a trickle of emails along the lines of, "your work is awesome, can I be yo…

Let's talk if and when you want to build something. I make no promises that piet-gpu-hal is suitable for other workloads (right now I would say it barely meets the needs for piet-gpu), but on the…

Yeah, sometimes atomics perform way better than you expect them to. Check out the linkedlist benchmark in my suite, 12.1 G elements/s on AMD 5700 XT using DX12. That's a respectable fraction…

Vulkan can't reliably do parallel dispatches, certainly not with any kind of scheduling fairness guarantee. CUDA has cooperative groups, which is a huge advantage. Okay, I see your point about de…

> They want to walk a LoD/culling tree DAG, my friend, not tree. Seriously, I recommend people watch the talk, it's one of the more impressive demonstrations of how to use GPU compute pow…

Workgroup in Vulkan/WebGPU lingo is equivalent to "thread block" in CUDA speak; see [1] for a decoder ring. > Using atomics to solve this is rarely a good idea, atomics will make thi…

Thanks for the detailed explanation, that makes sense and is useful information. My understanding was that they were more like real pointers, but I see that is not the case.

So it depends a lot on the goals. CUDA is a very good developer experience, and Vulkan compute shaders is (at the moment) a very bad one. But if the goal is to ship real compute on a wide variety of d…

This doesn't sound right to me. Vulkan 1.2 has support for pointers through an extension[1], and that's getting more widely available (certainly a lot more cards than can run ROCm). There&#x…

There's a lot of good stuff to be had from disassembling the ROM as well, don't ask me how I know.

Gah thanks, I will fix this. Originally I had an SGI Indigo, but I switched away from it because there isn't anywhere nearly as good a description of its architecture as copetti's on the cla…

What would you suggest instead, as a well-documented system? I considered a few, but went with PS2 because of the excellent analysis. My understanding is that it's asynchronous from a programming…

That looks really interesting, thanks! I hadn't seen it before. It certainly looks like it has most of the goodies we want, including present identifiers, the ability to target a specific time, e…

I think it doesn't help that much - if your system is well engineered, you should be able to hit the frame deadline reliably - and it also makes the situation with compositing more complex. My pe…

I talked about beam racing a bit in the "compositor is evil" post, but this time around wanted to focus on techniques that are practical to ship and can work reliably rather than depending o…

The chart is based loosely on the Call of Duty presentation (linked in the article), which does employ parallelism to reduce latency. There's also a lot of interest in ECS engines that can parall…

That's fair. The Cochrane review links to many of the same studies, for what it's worth.

Your link is an oft-debunked misinformation campaign[1]. In my opinion, the best current review is the Cochrane review[2]. And indeed, I think "anyone who looked into it never saw it work, ever. …

I call bullshit. Antibodies wane but the memory cell response is robust and holds up over what looks to be a pretty long period (studied intensively for 6 months but with no sign of decline). Here…

A really excellent survey on the prospects for mRNA technology from Derek Lowe: https://www.science.org/content/blog-post/what-mrna-good-and... …

Here's the actual paper: https://www.pnas.org/content/118/43/e2108728118 There's a lot of detailed discussion about different types of cells. Vero cells don&#…

I'm starting to get pretty optimistic. There seem to be a bunch of antivirals with very different mechanisms of action - this one blocks entry of the virus to cells, Molnupiravir fucks up accurat…