3 ms·
> Well that world would suck immensely, with horrible integration between hardware and software like exists right now outside of the Apple world. Why? The hard
by mschuster91 3y ago
> Well that world would suck immensely, with horrible integration between hardware and software like exists right now outside of the Apple world.
Why? The hardware people at companies like NVIDIA or Apple already have to write detailed documentation for their stuff anyway so that the software people can do their side.
Assuming they do a good enough job at writing documentation and testing their hardware, it should not be a problem for Linux developers to write appropriate drivers. Hell there are a very few select people able to write high quality drivers even with zero documentation (see Asahi Linux).
The only ones in trouble would be the vendors of crap SoCs who ship extremely buggy hardware and make up for the lack of QA in software quirks instead.
- pca006132 3y ago1. Hardware companies need some software to test their product, so this will be wasting their effort. 2. Good documentation and testing is probably not as common as you think. 3. It takes a huge amount of effort to write high quality drivers by reverse engineering. The thing about Asahi Linux is that the hardware does not have many variants, but there are a lot more different hardware for PCs.
- mschuster91 3y ago> 1. Hardware companies need some software to test their product, so this will be wasting their effort. They can also open source their testing bench code. The only ones who have to be afraid are those who don't care about quality. "Sunlight is the best disinfectant" also applies here - I'd really like to be able as a consumer to choose the product that demonstrates the better engineering! > 2. Good documentation and testing is probably not as common as you think. Oh, I'm aware of that, and the consequences of that are becoming more and more dire as electronic waste from defunct companies or unsupported and unsecure devices piles up, together with hacked IoT devices causing mayhem on the wide Internet in botnets. It's high time for that to stop. > 3. It takes a huge amount of effort to write high quality drivers by reverse engineering. The thing about Asahi Linux is that the hardware does not have many variants, but there are a lot more different hardware for PCs. So what? Ordinary PC hardware at its core still conforms to decades old standards that can be used as a fallback. Given a sufficient supply of beers, coffee and pizza, every CS student worth their money can throw up a basic graphical OS that runs on any mainstream (i.e. non-embedded) x86 system. Input, output (video, serial, audio, keyboard, mice, accessories) and storage all have standardized interfaces - the only thing that does not is networking. There is no reason at all to say that the ARM world can't standardize on basic components or at least interfaces. Apple can - their UART implementation, according to rumors, dates back to the early iPod/iPhone days, while Samsung has something like five or six completely different implementation, and Mediatek and all the other SoC vendors run their completely own stuff.
- judge2020 3y ago> 2. Good documentation and testing is probably not as common as you think. Not to mention that documentation != the actual behavior of the card. both nvidia and AMD release a ton of day-one driver fixes for new games being released that provide stability and sometimes double-digit performance gains over a pre-patched driver. The only way this works in a disintegrated ecosystem is if game devs become pros at creating driver bug workarounds.
- mschuster91 3y ago> The only way this works in a disintegrated ecosystem is if game devs become pros at creating driver bug workarounds. They already are, they have to because that is the reality on mobile games. IIRC the devs of Real Racing (or another Android car race game) made a blog post on that.
- scarface_74 3y agoEvery year, Apple releases new hardware and software in tandem. How long does it take Android manufacturers and Windows manufacturers to take advantage of the latest OS features? How are those Android updates working out compared to iPhones?
- mschuster91 3y ago> How long does it take Android manufacturers Way way too long, but to be fair a huge part of the issue are carriers and their insistence to pack the phones full of bloatware that needs to be tested against each update. > and Windows manufacturers to take advantage of the latest OS features? Windows device manufacturers tend to do what Microsoft demands of them. > How are those Android updates working out compared to iPhones? Samsung is fairly decent, although I'd be happier if they'd ... consolidate their lineup a bit and instead offer longer availability for spare parts.
- scarface_74 3y agoFairly decent? The iPhone 5s from 2013 just got an update and it’s the carriers is a poor excuse. Apple runs on the same carriers. Google’s first party phones don’t have third party bloat ware and still have a piss poor history of updates
- mschuster91 3y agoAnd guess why? Because ~3-ish years used to be what the SoC vendors provided as BSP support timeframe, and there was no mandate that they open up their specifications. Most don't even care about following the legal minimum aka provide the kernel and bootloader source code, especially not all these fly-by-night gongkai "brands".
- scarface_74 3y agoSo it’s what happens when hardware and software is not integrated…
- pandaman 3y agoHave you wrote drivers? Writing a driver off documentation is: a) very hard, how do you debug anything without JTAG ports and simulators? and b) not correct on principle because the documentation is not the hardware, there are errata in the hardware so what documentation says is not always what the hardware does and there are mistakes in the documentation for the same outcome. >The only ones in trouble would be the vendors of crap SoCs who ship extremely buggy hardware and make up for the lack of QA in software quirks instead. I.e. the every single vendor currently on the market...
- mschuster91 3y ago> Writing a driver off documentation is: a) very hard, how do you debug anything without JTAG ports and simulators? and b) not correct on principle because the documentation is not the hardware, there are errata in the hardware so what documentation says is not always what the hardware does and there are mistakes in the documentation for the same outcome. I agree with you, no system (particularly not something as complex as an entire SoC) is free of bugs. > I.e. the every single vendor currently on the market... And yet, we have Apple, whose hardware quality alone is good enough for a dedicated team to reverse engineer and write drivers for, with almost zero documentation available. The problem, as I expanded below, is the complete lack of competition in the SoC space. Everyone gets away with mediocrity because the largest players in town (Apple and Samsung) refuse to sell their chips to externals. The result is that the mobile ecosystem as a whole suffers, because that is what happens in markets where entities get too large.
- pandaman 3y ago>And yet, we have Apple, whose hardware quality alone is good enough for a dedicated team to reverse engineer and write drivers for, with almost zero documentation available. Apple own software is pretty janky, my current MacBook cannot bring up GPU after sleep about once a month and needs a reboot to begin drawing GUI correctly again, used to be dying completely waking up before it was fixed couple OS updates back. My previous MacBook could not stay on WiFi for more than a couple hours straight, never had been fixed. Also, reversing an existing working driver is different from writing a driver from scratch. In one case you just need to repeat the functionality, produce the same outputs for the same inputs, the quality of hardware is orthogonal here since it's already worked around in the existing driver. It's orders of magnitude easier than writing the thing from scratch.
- andromeduck 3y agoDocumenting this in detail is neigh impossible. Ashai Linux has the benefit of having working code to RE - this wouldn't work for silicon or board bring up.