Sapozhnik posted:

Vlc works at a basic level, but "works" in 2017 should mean that it uses all available hardware acceleration for playing video, which afaict no linux media player does right now.

Switching to wayland is a necessary component of making this happen though.

spankmeister posted:

fedora has the best gnome and wayland implementation

Sapozhnik posted:

Vlc works at a basic level, but "works" in 2017 should mean that it uses all available hardware acceleration for playing video, which afaict no linux media player does right now.

Rooney McNibnug posted:

Fedora is pretty great but lol at trying to get media stuff like VLC working since the switch to wayland, such as https://ask.fedoraproject.org/en/question/98107/vlc-picture-freezes-on-fedora-25/

It will be even more cool and good when they're able to support that stuff better and bring it out of rpmfusion.

Captain Foo posted:

Just don't play any weird animes and you'll be GOLDEN

Sapozhnik posted:

Vlc works at a basic level, but "works" in 2017 should mean that it uses all available hardware acceleration for playing video, which afaict no linux media player does right now.

Sapozhnik posted:

Switching to wayland is a necessary component of making this happen though.

Notorious b.s.d. posted:

every linux video player has done hardware acceleration since the 1990s, using Xv. naturally wayland breaks the poo poo out of Xv, because wayland is not an X11 server.

Apocadall posted:

optimus card is acting odd, i can start up using the intel graphics card and booting a game or something that uses more resources will engage the nvidia card but after they close out it'll stay on instead of switching back to the intel graphics

still need to figure out power management on this too, in windows lenovo had a power management driver and i could get 6-8 hours out of my battery, now i'm lucky to get 3 so i know my power management stuff is screwed up

Sapozhnik posted:

Codec acceleration is at least as important as YUV presentation.

Apocadall posted:

optimus card is acting odd, i can start up using the intel graphics card and booting a game or something that uses more resources will engage the nvidia card but after they close out it'll stay on instead of switching back to the intel graphics

Apocadall posted:

still need to figure out power management on this too, in windows lenovo had a power management driver and i could get 6-8 hours out of my battery, now i'm lucky to get 3 so i know my power management stuff is screwed up

Sapozhnik posted:

outputs are hardwired to the intel gpu in all but the very earliest hybrid graphics laptops. the discrete gpu renders into shared video memory which is then scanned out by the intel gpu. this is how it is supposed to work.

i hope you've learned your lesson and don't buy a hybrid graphics laptop ever again though because why do you even need any of that nonsense. a discrete gpu that can fit into a laptop's thermal profile is still going to sukkkkkk

Sapozhnik posted:

outputs are hardwired to the intel gpu in all but the very earliest hybrid graphics laptops. the discrete gpu renders into shared video memory which is then scanned out by the intel gpu. this is how it is supposed to work.

i hope you've learned your lesson and don't buy a hybrid graphics laptop ever again though because why do you even need any of that nonsense. a discrete gpu that can fit into a laptop's thermal profile is still going to sukkkkkk

Sapozhnik posted:

outputs are hardwired to the intel gpu in all but the very earliest hybrid graphics laptops. the discrete gpu renders into shared video memory which is then scanned out by the intel gpu. this is how it is supposed to work.

Sapozhnik posted:

i hope you've learned your lesson and don't buy a hybrid graphics laptop ever again though because why do you even need any of that nonsense. a discrete gpu that can fit into a laptop's thermal profile is still going to sukkkkkk

Notorious b.s.d. posted:

at least under X11, the "switching" on linux doesn't actually switch any outputs. the intel chipset always controls the graphics outputs. on-demand, the nvidia chip can be powered on, renders to an off-screen buffer, and the intel chip displays the buffer

this has a modest performance overhead but it makes everything a lot simpler for X11


re: power management, it sounds like your nvidia chip is being powered on continuously. that's about the only fuckup that could kill battery life that badly

James Baud posted:

So I thought I'd try a Ubuntu 17.04 beta LiveCD tonight on my Windows laptop (some manner of Asus ZenBook with a high DPI screen) just to see how the hardware support was looking.

Booted well enough, but the network experience was a little lackluster compared to all previous experience with these LiveCDs, going back to 2004 or so. It seems Ubuntu is using systemd-resolved and systemd-timesyncd these days (since 2016.10?) and have failed to provide any manner of fallback for what I think is one of the most likely non-developer use cases:

1) I said this was a Windows system. Windows stores local time to the hwclock, not UTC.
2) The LiveCD sets system time from the hwclock, assuming UTC.
3) systemd-resolved is set to perform DNSSEC validation.
4) Every page you access in a browser that provides DNSSEC info (that's a fair chunk of the web, these days) fails to load due to DNSSEC failures.
5) Every fairly official NTP server it (or you) might try to access definitely does provide DNSSEC validated answers and thus doesn't resolve, failing to update the clock.
6) If you change /etc/resolve.conf to route around this local resolution failure, too bad, systemd-timesyncd still seems to use systemd-resolved (unconfirmed, maybe it just didn't​ try to update while I had done that and was trying to figure out what I could run on Ubuntu without ntpdate present to do the update)

So, really, to fix this problem, you must realize (without Google) that it's all broken because your time zone defaulted to UTC and that your "correct" clock is wrong and lower yourself to manually fixing the time to some inaccurate "close enough" value. Good luck with that, anyone who doesn't already know all about DNSSEC.


(A hardcoded fallback IP for the default NTP server would be sensible, but no..... Or perhaps they could fallback on setting the clock the way I think Chromebooks do, by trusting the time on a https server that matches a given hardcoded certificate)

Anyway, I assume 16.10 shipped this way, too, but I sure didn't get any hits during a couple minutes of looking for other people who encountered the same problem - possibly my terminology since I actually understand what happened.

Tankakern posted:

sudo timedatectl set-local-rtc 1

carry on then posted:

linux for human beings

James Baud posted:

So I thought I'd try a Ubuntu

Tankakern posted:

sudo timedatectl set-local-rtc 1

The_Franz posted:

code:

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\TimeZoneInformation]
"RealTimeIsUniversal"=dword:00000001

fix windows legacy dos poo poo instead of breaking things that store time correctly

The_Franz posted:

code:

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\TimeZoneInformation]
"RealTimeIsUniversal"=dword:00000001

fix windows legacy dos poo poo instead of breaking things that store time correctly

Tankakern posted:

sudo timedatectl set-local-rtc 1

Sapozhnik posted:

Run powertop and make sure the on-die SATA controller and various other on-die PCIe peripherals have power management enabled. SATA is the big one though: if the SATA controller doesn't power down the SATA links then the controller itself cannot go into a low-power state, which inhibits power saving for the entire SoC.

The reason that this isn't the default is because the current Linux power management code does not do precisely the exact same thing that Windows does in every single situation and Windows is the only thing that hardware manufacturers ever test.

James Baud posted:

Sapozhnik posted:

Run powertop and make sure the on-die SATA controller and various other on-die PCIe peripherals have power management enabled. SATA is the big one though: if the SATA controller doesn't power down the SATA links then the controller itself cannot go into a low-power state, which inhibits power saving for the entire SoC.

The reason that this isn't the default is because the current Linux power management code does not do precisely the exact same thing that Windows does in every single situation and Windows is the only thing that hardware manufacturers ever test.

hifi posted:

https://lkml.org/lkml/2017/3/31/641

Cybernetic Vermin posted:

yeah, loving nvidia, clearly the party to blame in not keeping a driver up to date on a platform where the kernel development team has decided that defining an abi to write the driver against is something that can be successfully replaced by a process wherein the developer instead spends his time smearing his poo poo over the walls and complaining

ratbert90 posted:

Jesus Nvidia, get your poo poo together already.

Fedora updated to kernel 4.10.6 which promptly broke the Nvidia driver.

4.10 has been out since February 26th, and Nvidia STILL has yet to update the driver to work with it. I found a patch, but seriously, I am starting to understand why Torvalds hates Nvidia.

Notorious b.s.d. posted:

codec acceleration is a hot mess with few/no standards and licensing minefields

nvidia vdpau is very well-supported across the board, and once you get out of that safe space, it's a shitshow