Use ffmpeg as an Android stagefright plugin
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS). My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs. Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot! -- Chih-Wei Android-x86 project http://www.android-x86.org
I was very glad to hear the news. it not a difficult thing to update stagefright-plugins to JB. but you should cleanup these code, if feel a bit chaotic. On Sun, Aug 11, 2013 at 4:25 PM, Chih-Wei Huang <cwhuang@android-x86.org>wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
2013/8/13 Michael Chen <omxcodec@gmail.com>:
I was very glad to hear the news. it not a difficult thing to update stagefright-plugins to JB. but you should cleanup these code, if feel a bit chaotic.
On Sun, Aug 11, 2013 at 4:25 PM, Chih-Wei Huang <cwhuang@android-x86.org> wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Hi ffmpeg-dev, I've finished item 3 so now I'm able to build ffmpeg libs with the AOSP 4.3 tree. I've succeeded to build different archs including x86, mips, armv7-a and armv7-a-neon. http://git.android-x86.org/?p=platform/external/ffmpeg.git;a=shortlog;h=refs... Is it possible to merge it in the ffmpeg upstream? About the stagefright-plugins, I've also succeeded to update it to build with android 4.3. With some hacks to frameworks/av, now I'm able to load the ffmpeg plugins and play full hd video by the built-in Gallery2 and Youtube. But the performance is not so good as expected. If anyone has experience to tune the android multimedia framework, please give me a hand. Anyway, will publish it once I clean it up. Regards, -- Chih-Wei Android-x86 project http://www.android-x86.org
On Fri, Aug 16, 2013 at 10:39:27PM +0800, Chih-Wei Huang wrote:
2013/8/13 Michael Chen <omxcodec@gmail.com>:
I was very glad to hear the news. it not a difficult thing to update stagefright-plugins to JB. but you should cleanup these code, if feel a bit chaotic.
On Sun, Aug 11, 2013 at 4:25 PM, Chih-Wei Huang <cwhuang@android-x86.org> wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Hi ffmpeg-dev, I've finished item 3 so now I'm able to build ffmpeg libs with the AOSP 4.3 tree. I've succeeded to build different archs including x86, mips, armv7-a and armv7-a-neon.
http://git.android-x86.org/?p=platform/external/ffmpeg.git;a=shortlog;h=refs...
Is it possible to merge it in the ffmpeg upstream?
i see lots of hardcoded config.mak/h files Consider that new options could be added to configure and would then be missing from this, also a user might want to build with different codecs / (de)muxers than the ones hardcoded in these files [...] Thanks -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB The misfortune of the wise is better than the prosperity of the fool. -- Epicurus
Michael Niedermayer said:
Is it possible to merge it in the ffmpeg upstream?
i see lots of hardcoded config.mak/h files
Consider that new options could be added to configure and would then be missing from this, also a user might want to build with different codecs / (de)muxers than the ones hardcoded in these files
That's right. This is the easier way to create a compatible makefile for android build system. Considering the target (android) is well-known, there is no real benefit to configure codecs/(de)muxers dynamically. I also considered to generate the android config.* from configure script on the fly. However, the output config.mak of configure is not friendly to android build system. It exports some variables like CC, CFLAGS that will conflict with the android build system. (Note Android doesn't use sub-make -- all sub-makefiles are included in from the main makefile and all vars in sub-makefiles are exported) If you can modify the output of configure, say, splitting it into two files to let ARCH_*, HAVE_*, CONFIG_* vars in one file and others in another file, that would be more friendly to android build system. Then I could generate the android config.* on the fly. -- Chih-Wei Android-x86 project http://www.android-x86.org
On Tue, Aug 20, 2013 at 11:51:51AM +0800, Chih-Wei Huang wrote:
Michael Niedermayer said:
Is it possible to merge it in the ffmpeg upstream?
i see lots of hardcoded config.mak/h files
Consider that new options could be added to configure and would then be missing from this, also a user might want to build with different codecs / (de)muxers than the ones hardcoded in these files
That's right. This is the easier way to create a compatible makefile for android build system. Considering the target (android) is well-known, there is no real benefit to configure codecs/(de)muxers dynamically.
it can make sense to disable code that is very rarely used to reduce the size of the resulting binary.
I also considered to generate the android config.* from configure script on the fly. However, the output config.mak of configure is not friendly to android build system. It exports some variables like CC, CFLAGS that will conflict with the android build system. (Note Android doesn't use sub-make -- all sub-makefiles are included in from the main makefile and all vars in sub-makefiles are exported)
If you can modify the output of configure, say, splitting it into two files to let ARCH_*, HAVE_*, CONFIG_* vars in one file and others in another file, that would be more friendly to android build system. Then I could generate the android config.* on the fly.
You should be able to change CFLAGS / CC in the "# OS specific" section of configure if they have incorrect values configure is using the set cc/cflags to test various things like compiler features so they should be set correctly. But if i misunderstand and theres a real need to omit fields from the config files that can be done as well or they could be placed under ifdef [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Good people do not need laws to tell them to act responsibly, while bad people will find a way around the laws. -- Plato
2013/8/21 Michael Niedermayer <michaelni@gmx.at>:
On Tue, Aug 20, 2013 at 11:51:51AM +0800, Chih-Wei Huang wrote:
You should be able to change CFLAGS / CC in the "# OS specific" section of configure if they have incorrect values
Android build system has already defined them. (in its way) What I need in config.mak is just the HAVE_*, CONFIG_* stuff.
configure is using the set cc/cflags to test various things like compiler features so they should be set correctly.
But if i misunderstand and theres a real need to omit fields from the config files that can be done as well or they could be placed under ifdef
Yes, I want to omit them. So ifdef is a good approach. -- Chih-Wei Android-x86 project http://www.android-x86.org
On Wed, Aug 21, 2013 at 12:12:07PM +0800, Chih-Wei Huang wrote:
2013/8/21 Michael Niedermayer <michaelni@gmx.at>:
On Tue, Aug 20, 2013 at 11:51:51AM +0800, Chih-Wei Huang wrote:
You should be able to change CFLAGS / CC in the "# OS specific" section of configure if they have incorrect values
Android build system has already defined them. (in its way)
To speak plainly: Yes, and the settings it uses are _shit_. You should be much better of just using what FFmpeg wants to use. Actually both things apply to the whole Android build system, it's not needed and causes more issues than it solves compared to using your own/FFmpeg's, properly designed one.
2013/8/21 Reimar Döffinger <Reimar.Doeffinger@gmx.de>:
On Wed, Aug 21, 2013 at 12:12:07PM +0800, Chih-Wei Huang wrote:
2013/8/21 Michael Niedermayer <michaelni@gmx.at>:
On Tue, Aug 20, 2013 at 11:51:51AM +0800, Chih-Wei Huang wrote:
You should be able to change CFLAGS / CC in the "# OS specific" section of configure if they have incorrect values
Android build system has already defined them. (in its way)
To speak plainly: Yes, and the settings it uses are _shit_.
Well, sounds like a religious war. But sorry I don't want to involve.
You should be much better of just using what FFmpeg wants to use.
Surely I use what ffmpeg wants to use. But they should be put to LOCAL_CFLAGS etc defined by the android build system.
Actually both things apply to the whole Android build system, it's not needed and causes more issues than it solves compared to using your own/FFmpeg's, properly designed one.
I didn't ask you change your design. I just hope the output of configure script could be more friendly to other build system including but not limited to android. This could be easily be done by separating the output into two parts (by ifdef or anything else) If you don't like the idea, that's fine. I'll maintain my own branch. -- Chih-Wei Android-x86 project http://www.android-x86.org
On Wed, Aug 21, 2013 at 12:50:25PM +0800, Chih-Wei Huang wrote:
2013/8/21 Reimar Döffinger <Reimar.Doeffinger@gmx.de>:
On Wed, Aug 21, 2013 at 12:12:07PM +0800, Chih-Wei Huang wrote:
2013/8/21 Michael Niedermayer <michaelni@gmx.at>:
On Tue, Aug 20, 2013 at 11:51:51AM +0800, Chih-Wei Huang wrote:
You should be able to change CFLAGS / CC in the "# OS specific" section of configure if they have incorrect values
Android build system has already defined them. (in its way)
To speak plainly: Yes, and the settings it uses are _shit_.
Well, sounds like a religious war. But sorry I don't want to involve.
No, rather frustration (about its complexity and obfuscatedness) from when I tried to use it. And maybe a bit of ignorance.
Actually both things apply to the whole Android build system, it's not needed and causes more issues than it solves compared to using your own/FFmpeg's, properly designed one.
I didn't ask you change your design. I just hope the output of configure script could be more friendly to other build system including but not limited to android. This could be easily be done by separating the output into two parts (by ifdef or anything else)
If you don't like the idea, that's fine.
Personally I wouldn't object to a small and easy to maintain change if there are some people who think it's better, even if I'd consider it not useful personally. Of course I'd need to see a patch to see if it's small and easy to maintain :-) I am also (more out of personal interest) curious why you want to integrate FFmpeg into the Android build system that way. Because I looked into doing that and concluded for myself that it's only asking for pain for no benefit, compared to just building it separately and only linking to it. But as said, I see that I might simply be ignorant... Regards, Reimar
Hi, I'd like to ask for your generic advice in back-porting your work to my build, which is a slightly customized 4.1.2 AOSP (found here: https://github.com/Renesas-EMEV2). I think the pieces are all there already, but I'm wondering whether there's something preventing that, or making it too complex. Thanks in advance Fabio Rome - Italy On Sunday, 11 August 2013 10:25:25 UTC+2, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
Hi, Oh, NO, it is easy that Compatible with older Android versions， like 4.1.2 or 4.0.4 AOSP On Mon, Oct 13, 2014 at 6:24 PM, Fabio Fumi <ffumi68@googlemail.com> wrote:
Hi,
I'd like to ask for your generic advice in back-porting your work to my build, which is a slightly customized 4.1.2 AOSP (found here: https://github.com/Renesas-EMEV2).
I think the pieces are all there already, but I'm wondering whether there's something preventing that, or making it too complex.
Thanks in advance
Fabio Rome - Italy
On Sunday, 11 August 2013 10:25:25 UTC+2, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
Thanks for confirming that. Could you please explain a bit this step, how it could be applied to the AOSP case: "you should merge my android_frameworks_native(branch: cm_maguro-10.1) and android_frameworks_av(branch: cm_maguro-10.1) code" thanks Fabio On Mon, Oct 13, 2014 at 4:42 PM, Michael Chen <omxcodec@gmail.com> wrote:
Hi, Oh, NO, it is easy that Compatible with older Android versions， like 4.1.2 or 4.0.4 AOSP
On Mon, Oct 13, 2014 at 6:24 PM, Fabio Fumi <ffumi68@googlemail.com> wrote:
Hi,
I'd like to ask for your generic advice in back-porting your work to my build, which is a slightly customized 4.1.2 AOSP (found here: https://github.com/Renesas-EMEV2).
I think the pieces are all there already, but I'm wondering whether there's something preventing that, or making it too complex.
Thanks in advance
Fabio Rome - Italy
On Sunday, 11 August 2013 10:25:25 UTC+2, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
Hi Chih, I've completed the merge of frameworks/av and framewroks/native and added the stagefright-plugins and ffmpeg to my project. I enabld the USES_NAM and built new system. Startup wnet fine, but on YouTube test playback (which went just fine before), I get a crash (below). Note how I haven't enabled he new ffmpeg codecs in media_codecs.xml and I'm using Google SW codec for h264: <MediaCodec name="OMX.google.h264.decoder" type="video/avc"/> Suggestions? thanks Fabio ... D/H264Dec ( 707): H264SwDecInit# D/H264Dec ( 707): H264SwDecInit# decInst 0x2a055c08 noOutputReordering 0 D/H264Dec ( 707): H264SwDecInit# OK: return 0x2a0b1238 I/OMXCodec( 707): [OMX.google.h264.decoder] AVC profile = 66 (Baseline), level = 30 I/OMXCodec( 707): [OMX.google.h264.decoder] video dimensions are 320 x 240 I/OMXCodec( 707): [OMX.google.h264.decoder] Crop rect is 320 x 240 @ (0, 0) ... D/H264Dec ( 707): H264SwDecDecode# D/H264Dec ( 707): H264SwDecDecode# decInst 0x2a0b1238 pInput 0x4201ed88 pOutput 0x4201ed7c D/H264_decoder( 707): Access unit boundary D/H264_decoder( 707): SEQ PARAM SET D/H264Dec ( 707): H264SwDecDecode# OK: DecResult 1 D/H264Dec ( 707): H264SwDecDecode# D/H264Dec ( 707): H264SwDecDecode# decInst 0x2a0b1238 pInput 0x4201ed88 pOutput 0x4201ed7c D/H264_decoder( 707): Access unit boundary D/H264_decoder( 707): PIC PARAM SET D/H264Dec ( 707): H264SwDecDecode# OK: DecResult 1 D/H264Dec ( 707): H264SwDecDecode# D/H264Dec ( 707): H264SwDecDecode# decInst 0x2a0b1238 pInput 0x4201ed88 pOutput 0x4201ed7c D/H264_decoder( 707): Access unit boundary D/H264_decoder( 707): IDR D/H264_decoder( 707): SLICE HEADER D/H264Dec ( 707): H264SwDecDecode# OK: DecResult 4 D/H264Dec ( 707): H264SwDecGetInfo# D/H264Dec ( 707): H264SwDecGetInfo# decInst 0x2a0b1238 pDecInfo 0x4201eda8 D/H264Dec ( 707): H264SwDecGetInfo# OK D/H264Dec ( 707): H264SwDecDecode# D/H264Dec ( 707): H264SwDecDecode# decInst 0x2a0b1238 pInput 0x4201ed88 pOutput 0x4201ed7c D/H264_decoder( 707): IDR D/H264_decoder( 707): SLICE HEADER D/H264_decoder( 707): (null) D/H264Dec ( 707): H264SwDecDecode# OK: DecResult 2 D/H264Dec ( 707): H264SwDecNextPicture# D/H264Dec ( 707): H264SwDecNextPicture# decInst 0x2a0b1238 pOutput 0x4201ed98 flushBuffer 0 D/H264Dec ( 707): H264SwDecNextPicture# OK: return H264SWDEC_PIC_RDY F/libc ( 707): Fatal signal 11 (SIGSEGV) at 0x00000008 (code=1), thread 1818 (le.h264.decoder) ... I/DEBUG ( 70): #00 pc 00011ed0 /system/lib/libc.so I/DEBUG ( 70): #01 pc 0000ad6b /system/lib/libstagefright_omx.so (android::OMX::CallbackDispatcher::post(android::omx_message const&)+14) I/DEBUG ( 70): #02 pc 0000b5e5 /system/lib/libstagefright_omx.so (android::OMX::OnEmptyBufferDone(void*, OMX_BUFFERHEADERTYPE*)+30) I/DEBUG ( 70): #03 pc 0000ebd9 /system/lib/libstagefright_omx.so (android::SoftOMXComponent::notifyEmptyBufferDone(OMX_BUFFERHEADERTYPE*)+16) I/DEBUG ( 70): #04 pc 00009f60 /system/lib/libstagefright_soft_h264dec.so (android::SoftAVC::onQueueFilled(unsigned long)+508) On Sunday, 11 August 2013 01:25:25 UTC-7, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
Also after adding new ffmpeg codecs to I still get a crash on H264 (avc) playback. Added ffmpeg codecs to device/renesas/emev/media_codecs.xml Rebuilt... test... still crashing: V/AwesomePlayer( 758): track of type 'video/avc' does not publish bitrate V/AwesomePlayer( 758): mBitrate = -1 bits/sec V/AwesomePlayer( 758): haveAudio:1, haveVideo:1 V/AwesomePlayer( 758): initVideoDecoder flags=0x0 I/EV2OMXPlugin( 758): ***run OMFPlugin ------- makeComponentInstance !!!!!!!! I/OMXCodec( 758): [OMX.RENESAS.VIDEO.DECODER.H264] AVC profile = 66 (Baseline), level = 30 I/OMXCodec( 758): [OMX.RENESAS.VIDEO.DECODER.H264] video dimensions are 640 x 344 E/OMXNodeInstance( 758): OMX_GetExtensionIndex (index:'7fffffff') failed I/EV2OMXPlugin( 758): ***run OMFPlugin ------- destroyComponentInstance !!!!!!!! E/OMXMaster( 758): Invalid OMX component name 'OMX.ffmpeg.h264.decoder' E/MediaPlayer( 996): error (1, -2147483648) E/MediaPlayer( 996): Error (1,-2147483648) W/YouTube ( 996): (unknown) MediaPlayer error during prepare [what=1, extra=-2147483648] W/YouTube ( 996): (unknown) Retrying MediaPlayer error [retry=1, max=3] F/libc ( 758): @@@ ABORTING: INVALID HEAP ADDRESS IN dlfree addr=0x2a023ad0 F/libc ( 758): Fatal signal 11 (SIGSEGV) at 0xdeadbaad (code=1), thread 794 (Binder_2) ... I/DEBUG ( 70): #00 pc 000137b8 /system/lib/libc.so I/DEBUG ( 70): #01 pc 00015b95 /system/lib/libc.so (dlfree+1628) I/DEBUG ( 70): #02 pc 00016d43 /system/lib/libc.so (free+10) I/DEBUG ( 70): #03 pc 0007a853 /system/lib/libstagefright.so I/DEBUG ( 70): #04 pc 0007ac0d /system/lib/libstagefright.so (android::TimedEventQueue::~TimedEventQueue()+36) I/DEBUG ( 70): #05 pc 000538ad /system/lib/libstagefright.so (android::AwesomePlayer::~AwesomePlayer()+288) On Thursday, 30 October 2014 03:09:35 UTC-7, Fabio Fumi wrote:
Hi Chih,
I've completed the merge of frameworks/av and framewroks/native and added the stagefright-plugins and ffmpeg to my project. I enabld the USES_NAM and built new system.
Startup wnet fine, but on YouTube test playback (which went just fine before), I get a crash (below).
Note how I haven't enabled he new ffmpeg codecs in media_codecs.xml and I'm using Google SW codec for h264:
<MediaCodec name="OMX.google.h264.decoder" type="video/avc"/>
Suggestions?
thanks Fabio
... D/H264Dec ( 707): H264SwDecInit# D/H264Dec ( 707): H264SwDecInit# decInst 0x2a055c08 noOutputReordering 0 D/H264Dec ( 707): H264SwDecInit# OK: return 0x2a0b1238 I/OMXCodec( 707): [OMX.google.h264.decoder] AVC profile = 66 (Baseline), level = 30 I/OMXCodec( 707): [OMX.google.h264.decoder] video dimensions are 320 x 240 I/OMXCodec( 707): [OMX.google.h264.decoder] Crop rect is 320 x 240 @ (0, 0) ... D/H264Dec ( 707): H264SwDecDecode# D/H264Dec ( 707): H264SwDecDecode# decInst 0x2a0b1238 pInput 0x4201ed88 pOutput 0x4201ed7c D/H264_decoder( 707): Access unit boundary D/H264_decoder( 707): SEQ PARAM SET D/H264Dec ( 707): H264SwDecDecode# OK: DecResult 1 D/H264Dec ( 707): H264SwDecDecode# D/H264Dec ( 707): H264SwDecDecode# decInst 0x2a0b1238 pInput 0x4201ed88 pOutput 0x4201ed7c D/H264_decoder( 707): Access unit boundary D/H264_decoder( 707): PIC PARAM SET D/H264Dec ( 707): H264SwDecDecode# OK: DecResult 1 D/H264Dec ( 707): H264SwDecDecode# D/H264Dec ( 707): H264SwDecDecode# decInst 0x2a0b1238 pInput 0x4201ed88 pOutput 0x4201ed7c D/H264_decoder( 707): Access unit boundary D/H264_decoder( 707): IDR D/H264_decoder( 707): SLICE HEADER D/H264Dec ( 707): H264SwDecDecode# OK: DecResult 4 D/H264Dec ( 707): H264SwDecGetInfo# D/H264Dec ( 707): H264SwDecGetInfo# decInst 0x2a0b1238 pDecInfo 0x4201eda8 D/H264Dec ( 707): H264SwDecGetInfo# OK D/H264Dec ( 707): H264SwDecDecode# D/H264Dec ( 707): H264SwDecDecode# decInst 0x2a0b1238 pInput 0x4201ed88 pOutput 0x4201ed7c D/H264_decoder( 707): IDR D/H264_decoder( 707): SLICE HEADER D/H264_decoder( 707): (null) D/H264Dec ( 707): H264SwDecDecode# OK: DecResult 2 D/H264Dec ( 707): H264SwDecNextPicture# D/H264Dec ( 707): H264SwDecNextPicture# decInst 0x2a0b1238 pOutput 0x4201ed98 flushBuffer 0 D/H264Dec ( 707): H264SwDecNextPicture# OK: return H264SWDEC_PIC_RDY F/libc ( 707): Fatal signal 11 (SIGSEGV) at 0x00000008 (code=1), thread 1818 (le.h264.decoder) ... I/DEBUG ( 70): #00 pc 00011ed0 /system/lib/libc.so I/DEBUG ( 70): #01 pc 0000ad6b /system/lib/libstagefright_omx.so (android::OMX::CallbackDispatcher::post(android::omx_message const&)+14) I/DEBUG ( 70): #02 pc 0000b5e5 /system/lib/libstagefright_omx.so (android::OMX::OnEmptyBufferDone(void*, OMX_BUFFERHEADERTYPE*)+30) I/DEBUG ( 70): #03 pc 0000ebd9 /system/lib/libstagefright_omx.so (android::SoftOMXComponent::notifyEmptyBufferDone(OMX_BUFFERHEADERTYPE*)+16) I/DEBUG ( 70): #04 pc 00009f60 /system/lib/libstagefright_soft_h264dec.so (android::SoftAVC::onQueueFilled(unsigned long)+508)
On Sunday, 11 August 2013 01:25:25 UTC-7, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
second example attachement *20141106_163805.jpg *was missing ... On Sunday, 11 August 2013 10:25:25 UTC+2, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
hints audio: 1. E/OMXCodec( 2556): [OMX.RENESAS.AUDIO.DECODER.AAC] setParameter('OMX_IndexParamAudioAac') failed (err = -1010) 2. E/OMXCodec( 2556): [OMX.RENESAS.AUDIO.DECODER.AAC] setAACFormat() failed (err = -1010) video: 1. E/SoftFFmpegVideo( 2556): Deinterlacing failed On Fri, Nov 7, 2014 at 12:13 AM, Fabio Fumi <ffumi68@googlemail.com> wrote:
second example attachement *20141106_163805.jpg *was missing ...
On Sunday, 11 August 2013 10:25:25 UTC+2, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
Btw - testing downloaded files (as suggested in https://github.com/omxcodec/stagefright-plugins), playback is fine, at east with respect to image quality. Similar for both the test .mov and .mp4 files. Playback hangs and locks once in a while, for a few seconds, with logcat showing lots of: V/AwesomePlayer( 2090): we're late by 125631 us (0.13 secs) V/AwesomePlayer( 2090): we're late by 125631 us (0.13 secs) dropping one after 0 frames V/AwesomePlayer( 2090): we're late by 125382 us (0.13 secs) V/AwesomePlayer( 2090): we're late by 125382 us (0.13 secs) dropping one after 0 frames V/AwesomePlayer( 2090): we're late by 120819 us (0.12 secs) V/AwesomePlayer( 2090): we're late by 120819 us (0.12 secs) dropping one after 0 frames ... This happens in particular when there's extesive image changing. Less when video is more static. Seems like a lack of CPU bandwidth... Mediaserver sets CPU above 50% (though never 100%). PID PR CPU% S #THR VSS RSS PCY UID Name 2090 0 52% S 16 64160K 17060K fg media /system/bin/mediaserver 3175 0 1% R 1 1072K 476K root top 257 0 0% S 73 549956K 38280K fg system system_server 3009 0 0% S 15 492776K 37872K fg u0_a69 com.mxtech.videoplayer.ad ... Anyway, with respect to image quelity I don't see any significant difference in the video attribute, with repsect to those streamed though YouTube. See also logcat in http://pastebin.com/tnrt9NB6 On Sunday, 11 August 2013 01:25:25 UTC-7, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
Forgetting for the moment the image issues above (as I've restored playback through the HW H264 codec), I'm now hitting another issue... Testing on a "Live" (m3u8) streaming (.m3u8 URLs; which is the reason I'm trying to use ffmpeg SW codecs at all, as HW one failed to playback these), I got this failure: V/ACodec ( 75): onAllocateComponent V/ACodec ( 75): findMatching OMX codecs 'video/avc' (SW only) V/OMXCodec( 75): matchComponentName (null) V/OMXCodec( 75): matchComponentName (null) V/ACodec ( 75): Found matching codec: 'OMX.ffmpeg.h264.decoder' ... V/ACodec ( 75): ComponentName: OMX.ffmpeg.h264.decoder V/ACodec ( 75): nativeWindow creation V/ACodec ( 75): initNativeWindow E/OMXNodeInstance( 75): OMX_GetExtensionIndex (index:'4083345d') failed Probably ACodec need being patched for ffmpeg too? For one, I don't expect it to use native window... Added below check, in ACodec::LoadedState::onConfigureComponent: if (msg->findObject("native-window", &obj) && strncmp("OMX.google.", mCodec->mComponentName.c_str(), 11) && strncmp("OMX.RENESAS.", mCodec->mComponentName.c_str(), 12) #ifdef USES_NAM && strncmp("OMX.ffmpeg.", mCodec->mComponentName.c_str(), 11) #endif Now ACodec doesn't crash anymore and (just like the HW codec!) it goes on in logcat but no image is shown on screen: ... V/ACodec ( 75): [OMX.google.aac.decoder] calling fillBuffer 0x2a0671b8 V/ACodec ( 75): [OMX.google.aac.decoder] Now Executing V/ACodec ( 75): [OMX.google.aac.decoder] onOMXFillBufferDone 0x2a06a998 time 0 us, flags = 0x00000000, dataPtr = 0x42cef000 V/ACodec ( 75): [OMX.google.aac.decoder] calling fillBuffer 0x2a06a998 V/ACodec ( 75): [OMX.google.aac.decoder] onOMXEmptyBufferDone 0x2a0df9c0 V/ACodec ( 75): [OMX.google.aac.decoder] calling emptyBuffer 0x2a0df9c0 w/ time 92877 us ... Could something still be missing in ACodec? Thanks once again in advance for any help! Fabio On Friday, 7 November 2014 07:04:10 UTC-8, Fabio Fumi wrote:
Btw - testing downloaded files (as suggested in https://github.com/omxcodec/stagefright-plugins), playback is fine, at east with respect to image quality. Similar for both the test .mov and .mp4 files.
Playback hangs and locks once in a while, for a few seconds, with logcat showing lots of:
V/AwesomePlayer( 2090): we're late by 125631 us (0.13 secs) V/AwesomePlayer( 2090): we're late by 125631 us (0.13 secs) dropping one after 0 frames V/AwesomePlayer( 2090): we're late by 125382 us (0.13 secs) V/AwesomePlayer( 2090): we're late by 125382 us (0.13 secs) dropping one after 0 frames V/AwesomePlayer( 2090): we're late by 120819 us (0.12 secs) V/AwesomePlayer( 2090): we're late by 120819 us (0.12 secs) dropping one after 0 frames ...
This happens in particular when there's extesive image changing. Less when video is more static.
Seems like a lack of CPU bandwidth... Mediaserver sets CPU above 50% (though never 100%).
PID PR CPU% S #THR VSS RSS PCY UID Name 2090 0 52% S 16 64160K 17060K fg media /system/bin/mediaserver 3175 0 1% R 1 1072K 476K root top 257 0 0% S 73 549956K 38280K fg system system_server 3009 0 0% S 15 492776K 37872K fg u0_a69 com.mxtech.videoplayer.ad ...
Anyway, with respect to image quelity I don't see any significant difference in the video attribute, with repsect to those streamed though YouTube. See also logcat in http://pastebin.com/tnrt9NB6
On Sunday, 11 August 2013 01:25:25 UTC-7, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
Sorry posted wrong logcat fragment in my last update. During playback I keep getting messages like these below, though no image is shown: ... V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] calling emptyBuffer 0x2a06c4f0 w/ time 53232200 us V/ACodec ( 75): [OMX.google.aac.decoder] onOMXFillBufferDone 0x2a06aa48 time 53266566 us, flags = 0x00000000, dataPtr = 0x42cf7000 V/ACodec ( 75): [OMX.google.aac.decoder] onOMXEmptyBufferDone 0x2a0df9c0 V/ACodec ( 75): [OMX.google.aac.decoder] calling emptyBuffer 0x2a0df9c0 w/ time 53336233 us V/ACodec ( 75): [OMX.google.aac.decoder] calling emptyBuffer 0x2a07a038 w/ time 53359455 us V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] onOMXFillBufferDone 0x2a067240 time 52912200 us, flags = 0x00000000, dataPtr = 0x4452e000 V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] calling fillBuffer 0x2a067240 V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] onOMXEmptyBufferDone 0x2a06cfb0 V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] calling emptyBuffer 0x2a06cfb0 w/ time 53152200 us V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] onOMXFillBufferDone 0x2a0d4d48 time ... On Sunday, 11 August 2013 01:25:25 UTC-7, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
Michael Chen's good work(libstagefirght ffmepg extension) is especially useful to handle codecs which google is unwiiling to support,for the reason of license fee, such as rv,ra,wma,etc.As for those commonly used codecs h264,aac, it's better way to use original libstagefright and omx hardware codec to handle them. 2014-11-12 23:28 GMT+08:00 Fabio Fumi <ffumi68@googlemail.com>:
Sorry posted wrong logcat fragment in my last update. During playback I keep getting messages like these below, though no image is shown:
... V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] calling emptyBuffer 0x2a06c4f0 w/ time 53232200 us V/ACodec ( 75): [OMX.google.aac.decoder] onOMXFillBufferDone 0x2a06aa48 time 53266566 us, flags = 0x00000000, dataPtr = 0x42cf7000 V/ACodec ( 75): [OMX.google.aac.decoder] onOMXEmptyBufferDone 0x2a0df9c0 V/ACodec ( 75): [OMX.google.aac.decoder] calling emptyBuffer 0x2a0df9c0 w/ time 53336233 us V/ACodec ( 75): [OMX.google.aac.decoder] calling emptyBuffer 0x2a07a038 w/ time 53359455 us V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] onOMXFillBufferDone 0x2a067240 time 52912200 us, flags = 0x00000000, dataPtr = 0x4452e000 V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] calling fillBuffer 0x2a067240 V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] onOMXEmptyBufferDone 0x2a06cfb0 V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] calling emptyBuffer 0x2a06cfb0 w/ time 53152200 us V/ACodec ( 75): [OMX.ffmpeg.h264.decoder] onOMXFillBufferDone 0x2a0d4d48 time ...
On Sunday, 11 August 2013 01:25:25 UTC-7, Chih-Wei Huang wrote:
Hello ffmpeg list, I'm planning to integrate the latest ffmpeg as a stagefright OMXPlugin with Android-x86 4.3. I found the stagefright-plugins developed by Michael Chen is a good start. (repo: https://github.com/omxcodec/stagefright-plugins.git) But it's based on android 4.0 (ICS).
My plan is: * Update stagefright-plugins to work with stagefright of android 4.3. * Make it to be a libstagefrighthw.so plugin. * Create Android.mk to build ffmpeg libs.
Before I go ahead, I'd like to know if anyone is doing a similar project to avoid duplicate work. Thanks a lot!
-- Chih-Wei Android-x86 project http://www.android-x86.org
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
-- ,你好
participants (6)
-
Chih-Wei Huang -
Fabio Fumi -
Michael Chen -
Michael Niedermayer -
Reimar Döffinger -
Xiong Tang