|
|
#73 | |
|
Registered User
Join Date: Jun 2006
Posts: 678
|
Actually it's not that rosy.
When I try to replay some embedded youtube clips, I hear audio, but the picture remains static - only a full webpage refresh fixes the problem. |
|
|
|
|
|
|
#74 | |
|
Registered User
Join Date: Oct 2008
Posts: 114
|
I experience pixelation and the thing artem mentions too (sometimes only audio but no video, having to restart it to play again).
And sometimes the whole systems seems to freeze completely. When that happens it's enough to ssh into the system and kill the flash process (in Firefox it's "plugin-container" IIRC). But meh, what can we expect from Adobe...? |
|
|
|
|
|
|
#75 |
|
The god of fertility
Join Date: Apr 2006
Posts: 24
|
Adobe makes a plugin for a user process. It can only use interfaces provided by the system. If a user process can freeze your system, this is a system problem and a security threat.
|
|
|
|
|
|
#76 | |
|
Registered User
Join Date: Oct 2008
Posts: 114
|
Quote:
Edit: I noticed however, that in order to be able to play videos normally again I have to restart the comp (I haven't tried to unload the driver or restart X tho). When the "hang" happens, the CPU is divided 50% Xorg, 50% the flash process. I suspect flash uses VDPAU in a wrong way and leaves it in a bad state. When other apps try to use VDPAU again they can't for this reason. |
|
|
|
|
|
|
#77 |
|
Registered User
|
Adobe Flash Player 11.0.1.129 (a.k.a. 11.0 RC1 from 06/09/2011) 32bit, FireFox 6.0.2 32bit (official linux binary from Mozilla.com), GTS-250 GPU with 1024MB VRAM, nVIDIA driver v.280.13.
Having the following in the mms.cfg (using number and not verbose boolean values is recommended in the official Adobe's Flash Player administrators guide): Code:
EnableLinuxHWVideoDecode = 1 OverrideGPUValidation = 1 * Looks like Adobe had decided to disable OpenGL and/or presentation part of VDPAU with 11.x flash player series. As a result the "best" I can get when playing videos on youtube is "Software video rendering, Accelerated video decoding" and jerky performance when playing fullscreen 1080p. Performance was way better with version 10.3 of flash player and I would call how it was "almost ideal" unless that annoying "overlay" problems that are all well-know to the thread posters. * "Molehil" acceleration doesn't work as expected. Using testcase at http://labs.adobe.com/technologies/f.../molehill.html I get about 150% CPU usage while it should be around 3-5% in case "Molehil" were working correctly using OpenGL to accelearate animation. "Zombies" techdemo is also extremely slow and CPU hungry. Unfortunately I had missed times when flash player incubation 11.0 d0 version had been offered for downloads from Adobe Labs so I can't test "Molehil" techdemos with it. There were reports on forums that acceleration had been working as expected with that version of flash player. * Enabling VDPAU for flash player results in periodical "freezes" of the entire X server followed by quick black "flash" across the entire screen and lines like following written to the kernel log (dmesg): Code:
[765362.445808] NVRM: Xid (0000:02:00): 13, 0001 00000000 00005097 000015e0 00000000 00000080 [767129.336102] NVRM: Xid (0000:02:00): 8, Channel 00000003 [767131.336076] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context [767141.345079] NVRM: Xid (0000:02:00): 13, 0001 00000000 00005097 000015e0 00000000 00000080 [767275.622086] NVRM: Xid (0000:02:00): 8, Channel 00000003 [767277.622055] NVRM: os_schedule: Attempted to yield the CPU while in atomic or interrupt context [767287.632145] NVRM: Xid (0000:02:00): 13, 0001 00000000 00005097 000015e0 00000000 00000080 To summarize the above: with 11.0 RC1 using flash player with VDPAU had become more stable than it was in prior versions but this stability was achieved not by really fixing the bugs but rather by disabling output acceleration through GLX thus skipping renderpaths that triggered the bugs. While I beleive that most of the crappy code is inside flash player there are obviously things to fix on nVIDIA side either. The fact that userland app is able to bring the system - X server part of it - into totally unusable state (and - under some circumstances - it even may lead to a hard hang at kernel level and/or kernel panic) by messing with HW through VDPAU interface isn't acceptable. Userland interfaces providing access to the HW features must be foolproof, it would be a huge security risk otherwise. |
|
|
|
|
|
#78 | |
|
User
Join Date: Mar 2005
Posts: 366
|
Here's another bug report, this time on Nvidia Ion + Atom. (attached)
X sometimes freezes with 100% CPU usage when playing youtube flash in firefox. |
|
|
|
|
|
|
#79 |
|
The god of fertility
Join Date: Apr 2006
Posts: 24
|
As lexa2 observed in RC1, final release of Flash 11.0.1.152 also disables hardware acceleration even if it's turned on in the settings. Guess that's better than leaking through black.
I'm not trying VDPAU since you've confirmed it breaks the system just as with Flash 10. I'm a Linux user, gotta cherish my precious uptime! ;) |
|
|
|
|
|
#80 |
|
Registered User
Join Date: Nov 2008
Posts: 95
|
It's not just a problem with vdpau, though. The "Attempted to yield the CPU while in atomic or interrupt context" error causing X to slow (and usually lock the kernel up completely) has been around for ages (and is still in the 290.03 driver). I get it in 3d apps like wine games.
|
|
|
|
|
|
#81 | |
|
Registered User
Join Date: Jan 2007
Posts: 119
|
Quote:
https://bugbase.adobe.com/index.cfm?...bug&id=3002806 If you guys care about flash performance, please, vote for the bug. |
|
|
|
|
|
|
#82 | |
|
Registered User
|
Quote:
|
|
|
|
|
|
|
#83 |
|
Registered User
Join Date: Mar 2007
Posts: 47
|
Well, strange thing, there is new beta: http://labs.adobe.com/downloads/flashplayer11-2.html with this one I have both accelerated decoding and rendering, but, also pretty high CPU usage ;/ - about 40% on one core (and about 2x50% while the video is buffering - is this an issue on windows?) Just like they said in the note, they can now use both (for me) cores, but not very well:
Code:
Unknown sequence number while processing queue [xcb] Most likely this is a multi-threaded client and XInitThreads has not been called [xcb] Aborting, sorry about that. chromium-browser/chrome --type=plugin --plugin-path=/opt/Adobe/flash-player/plugin/libflashplayer.so --lang=pl --channel=14312.0x7f1c2727bd80.1807095856: /var/tmp/portage/x11-libs/libX11-1.4.4/work/libX11-1.4.4/src/xcb_io.c:273: poll_for_event: Assertion `!xcb_xlib_threads_sequence_lost' failed. |
|
|
|
|
|
#84 | |
|
Registered User
|
Quote:
Actually I'm so bored with Flash Player issues under Linux that even don't want to spend my time checking new beta just to run into a bunch of old and new bugs. For now I resort to watching flash videos under Google Chrome on my linux workstation as it is the only semi-bug-free way to do it. Same flashplayer binary under Firefox 7.0.1 works slowly as hell and almost hangs Firefox each time I try to open up more than three pages with embedded flash content. God bless HTML5, I wish this cross-platform technology a great success and would be happy to see Adobe Flash to be burned in flames. |
|
|
|
|
![]() |
| Thread Tools | |
|
|