Linux VAAPI: Nuc, Chromebox, HSW, IVB, Baytrail with Ubuntu 14.04 - Printable Version +- Kodi Community Forum (https://forum.kodi.tv) +-- Forum: Support (https://forum.kodi.tv/forumdisplay.php?fid=33) +--- Forum: General Support (https://forum.kodi.tv/forumdisplay.php?fid=111) +---- Forum: Linux (https://forum.kodi.tv/forumdisplay.php?fid=52) +---- Thread: Linux VAAPI: Nuc, Chromebox, HSW, IVB, Baytrail with Ubuntu 14.04 (/showthread.php?tid=165707) Pages:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
|
RE: vaapi-sse4: Deinterlacing Testing - uomiarz - 2014-05-09 (2014-05-03, 17:50)uomiarz Wrote: I am runnung asus chromebox on Celeron® 2955U Here is a fix for tearing Code: sudo nano /usr/share/X11/xorg.conf.d/20-intel.conf More details here: http://forum.xbmc.org/showthread.php?tid=185329&pid=1699218#pid1699218 RE: vaapi-sse4: Deinterlacing Testing - lelouch - 2014-05-09 (2014-05-07, 20:21)illiac4 Wrote:(2014-05-06, 19:57)lelouch Wrote:(2014-05-05, 14:57)lelouch Wrote: Hi, I have been using this ppa for a while and I must say it is really good. Specially for IPTV. Yes IPTV streams in most of the cases are multicast streams. I totally agree with you that streams can be slow because of packet flooding or broadcast storm or anything which can choke the network with unnecessary packets. However I have verified that my network is not flooded. It was verified using tcpdump and wireshark. Even I like to convert my udp stream to http stream using udpxy. However in this network, in which I experienced very slow channel switching time, problem actually is in how the IPTV headend is sending the transport stream [Mpegts]. In PMT table pcr_id value of the PES packet elementary stream is used. [Generally pcr_id value of the first elementary transport video stream is used] VLC is able to handle this stream pretty good so I am sure xbmc can also be tweaked to handle such streams. This IPTV stream can be tested using this file tcpdump file. https://www.dropbox.com/s/4apjw3aq73gg4jm/outfile.pcap Following is the example of replaying this file tcprewrite --enet-smac=00:26:6c:bc:70:33 --enet-dmac=01:00:5e:03:02:02 --dstipmap=0.0.0.0/0:239.3.2.2 --srcipmap=0.0.0.0/0:172.24.0.33 --infile=/outfile.pcap --outfile=/outfilewr.pcap tcprewrite --infile=/udpzero.pcap --outfile=/udpzero_chk.pcap --fixcsum tcpreplay --loop=0 --intf1=eth0 /udpzero_chk.pcap please note in the first tcprewrite command use your own mac addresses & ip addr to make this replaying work Once you replay this file, then you can start xbmc and open this link udp://239.3.2.2:1234. Opening this stream takes about 20 secs. Developers please try this out. This might help us make xbmc's mpegts library more better. Thank you RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-05-09 (2014-05-09, 04:58)uomiarz Wrote:(2014-05-03, 17:50)uomiarz Wrote: I am runnung asus chromebox on Celeron® 2955U No, No and no again. This will introduce Triple Buffering and we absolutely don't want that! Cause we are using swapBuffers to update the images on screen. Use the /etc/init/xbmc.conf I provide and you won't have a single bit of tearing. It's all a backing store issue. As you are not running this howto anyways, fix your lightdm.conf RE: vaapi-sse4: Deinterlacing Testing - gebnix - 2014-05-12 since a few days i have several problems with skins, others than confluence, on this build : 1. if i select "change content" in filemanager's context menu of a folder to add the folder to my library, xbmc is crashing 2. video and audio setting menus in OSD during video and Live-TV playback are empty both issues occur on Aeon MQ5, Aeon Nox and Amber. no problems with confluence. i can't reproduce these errors on my Raspberry Pi with official Openelec 4.0 using Amber. is there something special in this build, which could cause these issues ? RE: vaapi-sse4: Deinterlacing Testing - George - 2014-05-13 (2014-05-12, 19:09)gebnix Wrote: since a few days i have several problems with skins, others than confluence, on this build : Same here. I suspect it has something to do with these changes: http://forum.xbmc.org/showthread.php?tid=194515 This build is built from master and those skins are not yet ready for Helix. Isn't it possible to compile a Gotham compatible version with the deinterlacing goodness included also? RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-05-13 You are free to do so. RE: vaapi-sse4: Deinterlacing Testing - George - 2014-05-13 (2014-05-13, 10:04)fritsch Wrote: You are free to do so. Somehow I knew that would be the answer :-) I fixed the Aeon MQ5 skin, changes available for download here. RE: vaapi-sse4: Deinterlacing Testing - puithove - 2014-05-13 I've just purchased my first NUC and it is my first experience with Intel graphics. I'm putting this in as an upgrade for a much older and now under-powered Nvidia box. I decided to go this way based on what I read here. Unfortunately I'm having some disappointing results with the deinterlacing and I'd really appreciate some help trying to track it down. All progressive content is playing beautifully. Interlaced content via LiveTV (or recorded) from MythTV backend is having trouble. If I set Deinterlace method to "De-Interlace", I get extremely jerky playback depending on the content. When I open the codec info panel or have debugging enabled, I can see that the FPS is dropping down under 29.97 - around 15 or so. If I use Auto Select, it chooses a different method (I think probably BOB based on the visual results), playback is smooth. I do notice in this mode the FPS is highly variable (again based on content) anywhere between 29.97 up to and including 59.94. Some content runs at a steady 59.94 and this is the same content that plays fine with "De-Interlace" method. This can change on the same channel in the same stream when going to/from commercials. The content that seems to cause the variable FPS and the jerky playback appears to be that which the original recording was film - movies, TV shows recorded on film, etc. Other content that I'm pretty sure was sourced on video seems to be the smooth playback. If I'm right about that, it seems like it's content that has been Telecined that causes the issue. I'm using VAAPI hardware acceleration with the SW Filters enabled, but I get the same results with HW Accel turned off, using multithreaded software decode. CPU not even close to being maxed and load fairly evenly spread between cores. Settings that I'm using that seem relevant: Vertical Sync: Let driver decide Render Method: Auto Detect HQ scalers: 20% Decoding method: HW VDPAU: Off VAAPI: On MPEG2: On H264: On VC1: Off Use SW Filter: On Adjust refresh rate: On Start/Stop Sync playback to display: Resample Audio Deinterlace Video: Auto Deinterlace Method: De-Interlace Scaling Method: Nearest Neighbor (Lancos3 Optimized for 20% HQ). My specs: Openelec official release build 4.0.0 (my first go with OE and I'm really liking it thus far) Intel NUC D54250WYK1 -> HDMI out -> Yamaha AVR -> HDMI In -> Panasonic VT50 Plasma MythTV backend 0.27 PVR cmyth add-on Debug log of a reboot and starting playback for a couple minutes of affected video: https://spideroak.com/storage/OB2WS5DIN53GK/shared/647068-2-1485/01_XBMC.log?5ad3a35e903abfe3274cb051cc1edd96 This repeats repeats constantly during playback: Code: 14:34:26 T:140635415770880 DEBUG: ffmpeg[3BFFF700]: [src] Flat options syntax is deprecated, use key=value pairs Debug log using "Auto Select" method a few minutes later on the same TV show: https://spideroak.com/storage/OB2WS5DIN53GK/shared/647068-2-1480/01_XBMC.log?6b2ebe8944332ab5b2cc1f5852158333 You can see there are none of the messages called out above. RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-05-13 Provide: /usr/lib/xbmc/xbmc-xrandr | pastebinit From the log I see you run most of the time in 30hz mode and later in 60hz mode. Both are really bad for 59.94 content. All fine with that sample: https://dl.dropboxusercontent.com/u/55728161/deinterlacing-test.mp4 and Deinterlace? RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-05-13 Please try the latest from here: http://xbmcnightlybuilds.com/category/openelec-generic/ RE: vaapi-sse4: Deinterlacing Testing - puithove - 2014-05-13 Yea, I forgot that's one of the issues I'm having. On each reboot the GUI switches to 30hz mode because that's the "Preferred" from EDID. It does successfully switch to 59.94 when playback starts (this is with jerky video in full effect): http://sprunge.us/hVAe So far I've just been changing the GUI to 59.94 after each reboot (If I remember). Still need to google a solution at some point. Yes, playback of that sample seems smooth(ish) with method set to "De-Interlace". It's a 50hz file and gets played with refresh at 60.0 since I don't have a 50 mode. http://sprunge.us/iKRb (2014-05-13, 21:18)fritsch Wrote: Please try the latest from here: http://xbmcnightlybuilds.com/category/openelec-generic/ Actually that was one of the things I tried earlier this morning - r18369-g320ea5e - and it didn't help. RE: vaapi-sse4: Deinterlacing Testing - FernetMenta - 2014-05-13 could you please post a log running this version? RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-05-13 Can it be convinced to have 50hz, by putting that xorg.conf to /storage/.config ? Code: Section "Device" Edit: Btw. please try with one of those nightlies. OE has patched out ffmpeg 2.2 out of the FM patches for 4.0 and "mixed" something up. RE: vaapi-sse4: Deinterlacing Testing - puithove - 2014-05-13 (2014-05-13, 21:35)FernetMenta Wrote: could you please post a log running this version? Sure, would you like it playing MythTV content, that sample, or both? RE: vaapi-sse4: Deinterlacing Testing - fritsch - 2014-05-13 Never trust myth, but go ahead. |