> Both of these really suck because they are policy, not technical, decisions.

They are policy decisions that kind of make sense for a device like a tablet or phone though. Even though you could technically allow installing a complete development toolchain on an iPad, I can't imagine what the process would look like in practice. Download and install a complete *nix userland through the app store? Plus a compiler toolchain and each and every tool used in the build phase for your product? Who is going to maintain and distribute all these parts if the whole ecosystem is designed around the idea that apps are sandboxed and distributed through a curated app store? Imagine the customer support burden if you are the maintainer of some app that depends on external tools that can be used in a zillion different build/deploy configurations.

You could of course argue that the iOS ecosystem should not be based around a curated app store and sandboxed applications, but that would make it a MacBook...

Maybe we should put the whole idea of having one device that does everything to rest and accept that there are advantages to have a split between 'real computers' and tablets/phones. That's just my opionion though...

Edit: ah great, an immediate -3 because apparently people here think it is absolutely required to downvote straight away because they disagree with some opinion that is not their own.

Goodbye Hacker News, after ~10 years I'm finally done with the comment sections here and will deactivate my account and ask for it to be deleted

But they do all these things for obvious reasons. Reasons you and I may not agree with or be happy about, but still obvious reasons. In the case of repair/replacement it's just because they want you to use expensive replacement parts, they want to lure you into their Apple stores, and they don't want any liability/accountability for repairs with 'unofficial' parts.

I don't see how providing specifications about how their GPU's work so someone can make a Linux driver out of it hurts their commercial interests or liability though. Yes people may screw up their system if they install Linux on a Mac and it doesn't boot anymore, but as long as you can still take it into an Apple store and they can restore it to MacOS, why would Apple actively fight the extremely small minority of people who want to do that? And even if more people (developers/enthusiasts) would buy M1 hardware and immediately slap Linux on it, why would they care about that? They still made the sale, and these people will still walk around with a machine with a big fat Apple logo on it?

They previously spent a lot of effort accomodating people who wanted to run Windows on Macs using Boot Camp, so why would they be worried about people running Linux on M1 macs?

Edit: I can imagine Apple want to protect their IP and hence don't want to disclose anything about out it, period. Much like NVidia and most other GPU manufacturers do. But if AMD and Intel can be OSS-friendly, Apple could be too, apparently IP protection does not have to be a deal-breaker.

If you would use your 2fps streaming browser to read, say, hacker news, every scroll operation would be hideously slow and pull in another ~60KB per second, even though the page data itself is only a few KB and never changes. Your ‘streaming solution’ only makes sense if the total amount of data to fetch for the page itself outweighs the total amount of data for all the frames you need to stream while you are using the page. Which is probably almost never, unless you always look at static single-page applications which continuously pull in data on the backend without presenting anything new at the front end. Highly unlikely.

What I don't really see is why this service needs to exist to solve that particular problem (browser gets slow because too many tabs), because IMO that problem has already been solved very well by most decent browsers. They just swap out the inactive tabs and are able to restore them fast enough even on low-end systems, as long as they have an SSD. Inactive tabs that are not swapped out don't take a lot of CPU resources either. This service sells you a cloud browser with 16GB of RAM, which is pretty much the norm for laptops and desktops now, so it's not going to save you much if 'too many tabs' is causing slowness.

Two fps at 34KB each is ~500kbps by the way, not 60 kbps

As far as I know Dolby Atmos can be seamlessly mapped onto to 5.1 or 7.1, so from the developer perspective there should be no effort/cost to provide Dolby Atmos audio. I might be wrong about this but I assume the licensing cost would be for the playback device and not for the 'right' to bundle an Atmos audio track with your game?

What I don't understand is why Dolby Atmos is not used much more for games. Xbox One and PC support it but only very few titles use it. PS4 and PS5 don't support it all for games (only for video content), despite all Sony's bragging about their dedication to PS5 audio. Dolby Atmos seems perfect for games, for developers because it effortlessly maps audio directly to any 3D position in space, and for users because it scales all the way from headphones to soundbars to full 7.4.2 setups.

I was royally pissed off to learn PS5 would not natively support Dolby Atmos, I have a full 7.4.1 home-theater setup with height speakers and movies and Dolby Atmos demo's sound absolutely awesome. Yet if I play games the best I can get is 7.1 which is nice, but the height speakers go totally unused. It's probably related to licensing costs, but it is extremely disappointing having waited for the PS5 for so long and not seeing any kind of upgrade to the audio.

I know I had to jump through a lot of hoops to make it work though, my X470 motherboard didn't isolate the USB controller without a BIOS update for example, and after that the USB controller exhibited USB FLR (function level reset) problems causing it to hang the VM. This required blacklisting it's PCI ID from the Linux kernel and patching the kernel to disable FLR (fortunately these changes were later merged into the mainline kernel). I also had problems with the second GPU, if I plugged it into any slot other than the bottom x1 slot, the motherboard BIOS would reshuffle the IOMMU groups making it impossible to pass through the NVME and USB controller, or (if I put it in the second x16 slot) it would halve the PCIe bandwidth to the RTX3080.

All in all it took me the better part of a weekend to get everything working, but if I had to do it again from scratch and did some research into (in particular) the motherboard and BIOS, I would be able to set it everything up again in less than an hour.

Performance is indistinguishable from native, e.g. I can easily drive the screen at the max 144 Hz refresh rate, G-sync works, etc. I did put some effort in figuring out how to pin CPU threads to cores, optimize for the CPU core topology so Windows only gets cores on the same CCX (it's an AMD Zen 3, before that it was Zen 2), etc. But all of this is documented in many places.

Do note that depending on your motherboard not all of this is possible if the chipset & BIOS do not provide enough MMIO isolation, you might not be able to isolate a USB controller or a separate NVME drive, or you might get only half the PCIe lanes for the passthrough GPU if you need to use the other full-lane slot for something like a second GPU.

It's kind of weird structured bindings where not captured with [=](){} before, actually. I'm still stuck at C++11 for most of my work so I cannot use structured bindings at all, but I would not have expected to have to write that kind of monstrosity in C++17

Now I'm curious. Can you give a small code example of the kind of thing this solves and how it will change your life? ;-)

The litho step could in theory be optimized by skipping incomplete fields at the edges, but the reduction in exposure time would be relatively small, especially for smaller designs that fit multiple chips within a single image field. I imagine it would als introduce yield risk because of things like uneven wafer stress & temperature, higher variability in stage move time when stepping edge fields vs center fields, etc.

I still use it at work though because we don't have CLion there and that code base does not use CMake, but only because it is just slightly better than plain VIM with some plugins. But compared to CLion + CMake it's just one small step beyond a glorified text editor.

This way of using Wayland and VM's seems very interesting and useful. What I'm wondering though, is whether it would be realistically possible to also expose some kind of GPU acceleration to clients in VM's, for example by means of a virtual OpenGL adapter such as VirGL, which itself uses virtio on top of the host GPU driver. This would not get you anything near the performance required for games, but it should be way better than software rendering inside the clients. Do you think something like this would be possible already?

Well most of the time linked lists are used for things that have to be kept in some order that cannot (easily) be determined by a sorting predicate, and/or when re-sorting is not allowed.

>> In that case, why not store the elements in a unique_ptr (and then everything in a vector)?

But why? What's the benefit of doing that? The objects will still be scattered around the heap so iterating and dereferencing still trashes the cache, you lose benefits like not invalidating iterators on inserts/removals, and you complicate the code.

I'm not saying linked lists should be your go-to data structure, but IMO the 'linked lists are bad, dont use them' meme is a little overused.

It's not hard to come up with plenty of valid use cases for linked lists. Any time you need some algorithm that does a lot of non-predictable insertions/deletions, on a collection that is not regularly iterated in full or used to find elements (so pointer-chasing/cache effects are irrelevant). And/or when the elements are prohibitively expensive reorder/move on insertions/deletions, or some other piece of code is known to keep direct references to the elements that should not be invalidated when some unrelated element is inserted/removed and. Maintaining validity of iterators while an algorithm is operating on the list is another important benefit.

Some people will come up with all kinds of clever workarounds or more complicated datastructures/bookkeeping so you can achieve the same thing without a linked list, but in most cases I don't see the point of it. Especially not in cases that are not performance-sensitive at all. In such cases you're just increasing code complexity and obfuscating intent because of some ideological objection to using linked lists.