libstagefright - Experimental hardware video decoding builds - Printable Version +- Kodi Community Forum (https://forum.kodi.tv) +-- Forum: Development (https://forum.kodi.tv/forumdisplay.php?fid=32) +--- Forum: Kodi Application (https://forum.kodi.tv/forumdisplay.php?fid=93) +---- Forum: Android Development (https://forum.kodi.tv/forumdisplay.php?fid=184) +---- Thread: libstagefright - Experimental hardware video decoding builds (/showthread.php?tid=152005) |
RE: libstagefright - Experimental hardware video decoding builds - CruNcher - 2013-02-07 Yep RockChip is sadly not Amlogic though i hope they see how great it would be to have support for XBMC in comparison to their competition i guess that can make some impact in their minds, lets wait and see Btw Koying the 4.0.4 ics source is released if you mean the Android source you can find it here http://service.i-onik.de/a09_source_1.5/ics/ and here are their patches for CTS passing http://service.i-onik.de/a09_source_1.5/%5BRK3066_ANDROID4.0.4-SDK_V2.23_CTS%5DPatch_V1.1/ i think the changes are pretty low level hardware generic and should work on any RK 3066 Device no matter which ODM base RockChip also already ported 4.1.1 though they don't released the source yet and who knows if they ever will Also the Kernel Sources are already in Public in different versions RE: libstagefright - Experimental hardware video decoding builds - zpadguy - 2013-02-07 Hi, I tested 20130202 on my system. Creative HanZpad Android 4.0.4 Kernel 3.0.8 Initially I ran h264_720p_hp_5.1_6mbps_ac3_planet.mp4 (on sdcard) using the Gallery app and that played fine with smooth video and good audio. I then tried to play the same file with XBMC 20130202, the video didn't start properly, the screen was completely black but the audio played for about 10 seconds and then went off. Here is the logcat :- http://xbmclogs.com/show.php?id=33715 Many thanks, zpadguy RE: libstagefright - Experimental hardware video decoding builds - CruNcher - 2013-02-07 Hui nice a Zii ZMS-40 tab I/OMXCodec( 2841): [OMX.ZiiLABS.H264.Decoder] AVC profile = 100 (High), level = 51 I/OMXCodec( 2841): [OMX.ZiiLABS.H264.Decoder] video dimensions are 320 x 240 I/OMXCodec( 2841): [OMX.ZiiLABS.H264.Decoder] Crop rect is 320 x 240 @ (0, 0) the container/bitstream parser seems to get something wrong here also try some standard youtube and or apple .mov file for testing first see my link above for a 1080p youtube file zpadguy Antutu benchmark result ? and Creative says it can do H.265 decoding is there an OMX.ZiiLABS.H265.Decoder or a OMX.ZiiLABS.HEVC.Decoder Could you also release a XBMC Render Performance Result System->System Info RE: libstagefright - Experimental hardware video decoding builds - chandankp - 2013-02-08 (2013-02-07, 21:14)Koying Wrote: I've received the rockchip device (a minix neo x5) and ran a quick test. Rockchip engineer Herman Chen has relesed user mode vpu drivers in slatedroid forum for rk3066 devices. http://www.slatedroid.com/topic/41654-pre-alpha-03-ubuntu-linux-for-mk802-iii-ug802-mk808-ug007-imito-mx1/page__view__findpost__p__532370 Hope this might help you. /Chandan RE: libstagefright - Experimental hardware video decoding builds - Koying - 2013-02-08 (2013-02-08, 06:31)chandankp Wrote: Rockchip engineer Herman Chen has relesed user mode vpu drivers in slatedroid forum for rk3066 devices.Thanks. Very interesting stuff... RE: libstagefright - Experimental hardware video decoding builds - martinm_76 - 2013-02-08 Hi there, I have followed this thread with great interest and recently tried the 02.02 build on a NEO G4 RK3066 stick. I have only tested it with streaming video, but all of them have been scrambled with hardware accelleration active. So, I have experimented a bit with the other settings under video while I await a working HW accel version, and the settings in this screenshot gives me a pretty decent experience: http://dl.dropbox.com/u/2933487/Screenshot_2013-02-08-03-12-52.png Audio is nearly always in sync with the picture. Once in a while the picture skips several frames, but it is definately watchable with the things I have tried: DR NU video add-on cnet video add-on Project Free TV video add-on TWIT video add-on (some videos a little more choppy than others) I hope this may be of use to someone else as well. I have my NEO G4 set to 1280x720 60 Hz, so I'm not trying full 1080p playback here... Regards, /Martin. RE: libstagefright - Experimental hardware video decoding builds - bluepeter - 2013-02-08 (2013-02-08, 12:36)Koying Wrote:(2013-02-08, 06:31)chandankp Wrote: Rockchip engineer Herman Chen has relesed user mode vpu drivers in slatedroid forum for rk3066 devices.Thanks. Very interesting stuff... I take back what I said, very interesting indeed. RE: libstagefright - Experimental hardware video decoding builds - bluepeter - 2013-02-08 I assuming folk read further on in the thread and noted Hermans reply on the 26th, and I quote: Hi all, this is user mode vpu driver, but no On2 HW codec driver. If someone wanna to use rk3066 Vidoe Process Unit (VPU) hardware for acceleration, On2 HW codec driver is needed RE: libstagefright - Experimental hardware video decoding builds - Koying - 2013-02-08 For rk3066 users, it actually seems *some* 1080p work, while others basically crashes the vpu. I'll work on that this WE. RE: libstagefright - Experimental hardware video decoding builds - bluepeter - 2013-02-08 Koying what I've noticed - I re-ripped a dozen of my blu rays last weekend especially to run tests on my 2 RK 3066 TV sticks (UG 802/MK802-III) - ripped using MakeMKV to DD 5.1 or DTS 5.1, (no hi def audio) blu rays sizes ranged from 19gig to over 34gig.... I removed all crap (subs/titles, etc) so just retained movie & English sound track. Uploaded to my Qnap NAS.... both sticks struggle. I then ran these rips through Handbrake, left at default.... (MKV/high profile) resulting in 5gig to 9gig files... these all stream fine from my NAS. Though no 5.1 just stereo. As yet no pass-thru on HDMI with the RK3066, just down mixed. I've measured my network thru put, my wireless network is well capable of streaming 35gig ISO's that play @ 40mbps ... (dune/netgear neo media streamers) so I'm struggling to understand why the RK3066 is struggling with straight rips using MakeMKV but having transcoded them with handbrake they work fine? RE: libstagefright - Experimental hardware video decoding builds - CruNcher - 2013-02-08 (2013-02-08, 16:16)Koying Wrote: For rk3066 users, it actually seems *some* 1080p work, while others basically crashes the vpu. I'll work on that this WE. yet have to find a 1080p bitstream that works without freeze.crash or reboot somewhere @ playback but it's nice that you trying to debug this you should look @ the 18.01 build and the changes after it Btw where those 50 ms delay on screen rendering results with downscaling involved (1080p->720p) ? RE: libstagefright - Experimental hardware video decoding builds - Wolfsteen - 2013-02-08 (2013-02-08, 16:16)Koying Wrote: For rk3066 users, it actually seems *some* 1080p work, while others basically crashes the vpu. I'll work on that this WE. Thanks for trying to fix issues on rk3066 devices... RE: libstagefright - Experimental hardware video decoding builds - Koying - 2013-02-09 (2013-02-08, 18:57)CruNcher Wrote:Well, tried the 18.01 version and it doesn't work any better (on Monsters, at least -> freeze).(2013-02-08, 16:16)Koying Wrote: For rk3066 users, it actually seems *some* 1080p work, while others basically crashes the vpu. I'll work on that this WE.yet have to find a 1080p bitstream that works without freeze.crash or reboot somewhere @ playback but it's nice that you trying to debug this you should look @ the 18.01 build and the changes after it I've found some kind of patterns, though. All 1080p 4.1 I've tried with ReFrames = 5 do work (I didn't find one with > 5 in my lib). All < 5 basically crashes the vpu... You can check the value of a given file with MediaInfo. Correction, the pattern seem to be that full size frames, i.e. 1920x1080, vs. cropped frames, e.g. 1920x800, make the vpu crash. Furthermore, there something strange in the logcat: Code: I/OMXCodec( 1851): [OMX.rk.video_decoder.avc] video dimensions are 1920 x 1080 No idea where the 1088 come from RE: libstagefright - Experimental hardware video decoding builds - CruNcher - 2013-02-10 (2013-02-09, 10:12)Koying Wrote:(2013-02-08, 18:57)CruNcher Wrote:Well, tried the 18.01 version and it doesn't work any better (on Monsters, at least -> freeze).(2013-02-08, 16:16)Koying Wrote: For rk3066 users, it actually seems *some* 1080p work, while others basically crashes the vpu. I'll work on that this WE.yet have to find a 1080p bitstream that works without freeze.crash or reboot somewhere @ playback but it's nice that you trying to debug this you should look @ the 18.01 build and the changes after it i wouldn't count on that you should first try to get the Zelda Medley (Film Framerate) stable as it is much lower complexity and the chances if it runs are good that anything else will be more stable overall Complete name : G:\dropbox\Zelda Medley - YouTube.mp4 Format : MPEG-4 Format profile : Base Media / Version 2 Codec ID : mp42 File size : 177 MiB Duration : 4mn 8s Overall bit rate mode : Variable Overall bit rate : 5 996 Kbps Encoded date : UTC 2011-11-16 05:29:03 Tagged date : UTC 2011-11-16 05:29:03 gsst : 0 gstd : 248290 gssd : BADC21241HH1349561247919422 gshh : o-o---preferred---sn-4g57ln7s---v21---lscache2.c.youtube.com Video ID : 1 Format : AVC Format/Info : Advanced Video Codec Format profile : [email protected] Format settings, CABAC : Yes Format settings, ReFrames : 1 frame Codec ID : avc1 Codec ID/Info : Advanced Video Coding Duration : 4mn 8s Bit rate : 5 841 Kbps Maximum bit rate : 14.1 Mbps Width : 1 920 pixels Height : 1 080 pixels Display aspect ratio : 16:9 Frame rate mode : Variable Frame rate : 23.976 fps Minimum frame rate : 23.762 fps Maximum frame rate : 24.341 fps Color space : YUV Chroma subsampling : 4:2:0 Bit depth : 8 bits Scan type : Progressive Bits/(Pixel*Frame) : 0.117 Stream size : 173 MiB (97%) Tagged date : UTC 2011-11-16 05:29:06 Audio ID : 2 Format : AAC Format/Info : Advanced Audio Codec Format profile : LC Codec ID : 40 Duration : 4mn 8s Bit rate mode : Variable Bit rate : 152 Kbps Maximum bit rate : 181 Kbps Channel(s) : 2 channels Channel positions : Front: L R Sampling rate : 44.1 KHz Compression mode : Lossy Stream size : 4.50 MiB (3%) Title : IsoMedia File Produced by Google, 5-11-2011 Encoded date : UTC 2011-11-16 05:29:06 Tagged date : UTC 2011-11-16 05:29:06 and yes 1080 is mostly handled as 1088 but this especially for broadcast and blu-ray hardware encoder you will see that the bitstreams are 1088 but the 8 pixels aren't shown @ playback Tough what makes me curios about this 1 user here said he could playback this without problems with his finless rom based mk808 So here it is again and i please everyone with RK 3066 to test it and report results if it crashes or doesn't, the firmware you use and if it's HDMI out and which output resolution and screen refresh rate. https://dl.dropbox.com/s/18ys53m6npnmbju/Zelda%20Medley%20-%20YouTube.mp4?dl=1 also you should maybe first fix the SD stride issue first (squished output) from non width MOD8 bitstreams by padding those additional pixels @ playback so the renderer/decoder thinks it's mod8/mod16 and not mod2 which he seems can't handle RE: libstagefright - Experimental hardware video decoding builds - Koying - 2013-02-10 (2013-02-10, 01:00)CruNcher Wrote:I'm speaking about frame size, not framerate. Zelda is full 1920x1080 frame size and crashes. QED(2013-02-09, 10:12)Koying Wrote: Correction, the pattern seem to be that full size frames, i.e. 1920x1080, vs. cropped frames, e.g. 1920x800, make the vpu crash.i wouldn't count on that you should first try to get the Zelda Medley (Film Framerate) stable as it is much lower complexity and the chances if it runs are good that anything else will be more stable overall BTW, Zelda also shows this Code: I/ ( 90): Width 1920 Height 1088 |