According to an AMD engineer I asked at the time, when they evaluated Ryzen/K12, it was "maybe" a 15% advantage for ARM depending on scenarios.
The efficiency came solely from the frontend which is a lot heavier on x86, and stay up longer because decoding is way more complex. The execution units were the same (at least mostly, I think, might be misremembering) so once you are past the frontend there's barely any difference in power efficiency.
That's super cool, congrats on releasing it ! It's a feature that some people periodically ask me to add in Aerial, but I never got to it. Piping from yt-dlp to AVFoundation is definitely the way to go.
I was gonna warn you about a bug in macOS 15+ where your screensaver stays around after you go back to the desktop, but for some reason your code seems to avoid that issue. I'm not quite sure how, as you don't hook stopAnimation or any event apart from the deinit. But it works, so, massive kudos, I'll have to try and understand why !
So fyi, the way you hook it there is what makes the preview flicker in System Settings.
You check if the screen is locked, and if not, kill the host. But screen is not locked in System Settings. So basically, you're killing the host process every 2 seconds (and macOS, at least in Tahoe, restarts it, it doesn't in previous macOS versions).
That's also what causes your issues with "Options" not working (because you killed the instance that was linked to that button). The way we workaround it usually is to hook a system event.
Ha, you do the exit trick too then, I just missed it.
FYI that works 99% of the time, but for some people it sometimes crashes (because we exit our host container - legacyScreenSaver.appex - and sometimes if you do it at a wrong time things just hang).
In Spain, in a legal context, it's either condenado (condemned) or sentenciado (sentenced) more or less indistinctly. I have the impression we use a lot of words without much care for details.
Out of that context, it's usually condenado the one used.
I've been trying to make a library/cli to set the wallpaper/screensaver to use in the next version of Aerial (https://github.com/AerialScreensaver/PaperSaver) on individual screens and been toying around a lot with that whole WallpaperAgent subsystem (and obviously everything Aerial like the manifests, etc, before that), so I may have some insights/questions if you have time ?
From what I've seen there are multiple parts to the way that macOS subsystem works :
- Apple fetches the manifest (json file) with their own videos in (only) 240FPS
- It gets ingested in '/Library/Application Support/com.apple.idleassetsd/Aerial.sqlite' for some reason
- Apple pulls the videos in '/Library/Application Support/com.apple.idleassetsd/Customer/4KSDR240FPS' but renames them through a UUID (despite them having individual keys in the JSON)
Adding videos and sections in System Settings, you can do by manipulating that sqlite and killing WallpaperAgent (or maybe something else) before doing it, but as far as I remember (I only toyed with that part last year during Sequoia beta, so probably misremembering), macOS will periodically pull the manifest again and (fairly often) erase all your changes.
As far as I know, what you select then gets saved per screen/space in `~/Library/Application\ Support/com.apple.wallpaper/Store/Index.plist` (with a lovely Base-64 coding thrown in for fun).
The last part to this is a SystemWallpaperURL key stored in `~/Library/Preferences/com.apple.wallpaper.plist` which I believe (not 100% sure), contains the "special" video that gets played on a cold boot login (as far as I know it's a separate state from the "classic" Lock Screen).
So if I may :
- Do your videos show up on a cold boot too (that separate state I mentionned?) or just the "classic" Lock Screen? My rough guess is the cold boot lock screen can only display videos that are on the System Volume that's mounted before the user volume, so that one is probably fully out of reach.
- How hard did you have to workaround working with restricted paths ? Apple (for some good reasons) restricts hard access to files in user folders, and at that point the only safe place I can reasonably find (outside of containers, but that's a whole other story with screensavers) is
`/Users/Shared`. Are you using that folder too?
- Are you messing with the sqlite db, or are you injecting via a reversed engineered api?
- Did you try editing `~/Library/Application\ Support/com.apple.wallpaper/Store/Index.plist` to set your video wallpapers or are you just relying on them being integrated in System Settings?
Since Sequoia, right now Apple broke the way we could set a screensaver via terminal. I got that part working (setting per screen/space) in PaperSaver, but the wallpaper part (basically just switching to another user selected image, not even a video, but this has to be done per space for which we don't have a public api for), I can't seem to get quite right yet, so any insight you have on that would be welcome. Take care and again awesome effort on your launch, this is a non trivial system with so many pitfalls, it takes a bunch of dedication to make it work with so many subtle problems in every corner.
You could say "Linux was CREATED out of thin air", and I wouldn't argue with you.
But creation only counts for so much -- without support, Linux could still be a hobby project that "won't be big and professional like GNU"
I'm saying Linux didn't APPEAR out of thin air, or at least it's worth looking deeper into the reasons why. "Appearing" to the general public, i.e. making widely useful software, requires a large group of people over a sustained time period, like 10 years.
----
i.e. Right NOW there are probably hundreds of projects like Linux that you haven't heard of, which don't necessarily align with funders
I would actually make the comparison to GNU -- GNU is a successful project, but there are various efforts underneath it that kind of languish.
I'm saying that VIDEO CODECS might be structurally more similar to these projects, than they are to the Linux kernel.
i.e. making a freely-licensed kernel IS aligned with Red Hat, Intel, Google, but making an Intelligent Personal Assistant is probably not.
Somebody probably ALREADY created a good free intelligent personal assistant (or one that COULD BE as great as Linux), but you never heard of them. Because they don't have hundreds of companies and thousands of people aligned with them.
My point was, a lot of the early corporate support were smallish companies built specifically around Linux. RedHat is the perfect example of that, it started as a university project to make a distro.
It took a while (and a lot of pain) to get a lot of driver vendors to come fully into the project, yet Linux was already gaining a bunch of traction at that time (say last half of 90s).
I'll give you that Intel was always more or less a good actor though! But Google didn't exist when Linux already mattered. And when Google was created, they definitely benefited a lot from it, basing much of their infra on it.
Marketing needs (and laywer approval) can bring support faster than most things. Opus for audio is a good example of that too.
Definitely a cool idea ! Some feedback on your page :
The "See it in action" section doesn't really show it in action ? There's a picture but it's abstract, is it representative ? Are there multiple styles ? Is it just an image or a video ? Just looking at this I don't really know. Consider putting a gif/video preview in there so we get a better idea and/or add multiple examples ?
It got more confusing when I got below and saw the "How to use" > "visualize" section below where you have a mountain picture ? So which is it ?
Also maybe add video somewhere when you mention mp4, for non tech users. Visualization (which you use a lot in your text) doesn't convey that to me, but just my 2 cents!
About the swap, I can see how it happened with the new design code, make layers, put the transparent one below, and they didn't want to have the left side higher than the right one for reasons?
To me the bigger issue is by not having it go to the edge, it really breaks the one face/two face original design.
And it looks worse somehow in dark mode (but to be fair, everything is worse in dark mode right now in beta1, but Safari is the one that needs discussing) : https://ibb.co/CLhJ4XM
Installed iOS, iPad and macOS yesterday, some things are quickly obvious :
- In general, it always looks worse on dark mode
- The glass transparency effect is too local. It looks only at what's exactly below, so if you have two icons side by side in Control center on iPhone, one may show dark and the next one light, making you think one is active and the other one is inactive. It's pretty clear they wrestled with icons being too transparent so they blurred them a bunch, but it just makes it worse in those cases.
- It does have sensible defaults for (most) 3rd party icons that are flat, by adding some reticule on the flat logo to make it pop and look less out of place.
- The textfield contrasts can be horrendous. If you try to add a sky background to macOS messages (the first choice), the textfield is white text on lightly colored background. In Safari, if you have one of the default desktop background, you can get grey text on blue grayish background. There's absolutely no contrast and it's clear that they will have to address it.
- Safari for macOS takes the contrast issue above and pushes it to 11. It tries to reintroduce the universally hated concept of "the webpage takes over your browser window" but makes it worse. It's horrible enough to have your tabs and icons change color from white to black if you tab from say hacker news to github, but they've added a very slow (and buggy) animation for the UI on top. So while the tab switches immediately, the UI on top slowly morphs from white to black. Absolutely infuriating (and can't be disabled in beta 1). You also can't really see the selected tab in dark mode on a webpage with a black background.
In summary, some things look ok but in general it's really rough. The finder icon sums it best, they had a concept (transparent layers), and tried hard to shove everything through it, never stopping to question if maybe the concept needs adjusting when it clearly didn't work. I expect a bunch of changes, as is it's really rough.
Not op but having played Rebirth, while overall very good, it suffers from the classic case nowadays of adding repetitive "chores" to do around maps, to artificially increase the length of the game with little purpose.
So far (only 6 hours in, but some friends who went further confirmed), Expedition 33 seems to steer away from that, being a lot more story driven.
It also has, by far, the greatest prologue I've seen in a game.
Worse thing so far, the UI in the menu does take a second to get. Particularly the selected state is way too subtle and a bit confusing at first.
Agreed, I've only played FF7 Remake (which I'd waited patiently for years while it was being developed), and just did not enjoy playing the game. It felt like it 20 hour game stretched out to 70 hours with repetitive fetch quests. It lacked the charm and fun of the original, and it sounds like this Clair Obscur game has learnt from this experience.
The efficiency came solely from the frontend which is a lot heavier on x86, and stay up longer because decoding is way more complex. The execution units were the same (at least mostly, I think, might be misremembering) so once you are past the frontend there's barely any difference in power efficiency.