AlienBBC on ARM based NAS

[bpa]

Without fix pt - I think on an 266MHZ ARM mplayer was about 25 sec CPU for 1 sec of audio - not realtime. 500Mhz still would be too slow for Realtime so I think you have fix pt running.

I think 49% is a bit high just for mplayer on a live stream - is that with slimserver as well ?

I am surpised mms worked - IIRC WMA code still uses floating pt.

[flipflip]

No, 49% only mplayer (in a shell, not via Slimserver or Alien).

The mms stream showes as "mp3lame codec" or something else with mp3. I guess I'll have to try a real wma stream. Gotta find one :-)

flip

[flipflip]

../bin/mplayer-stdout -cache 64 -vc null -vo null -ao pcm:fast:file=- rtsp://a1355.l784539966.c7845.e.lr.akamaistream.net/live/D/1355/7845/v0001/reflector:39966 > rtsp-2.wav

Opening audio decoder: [ffmpeg] FFmpeg/libavcodec audio decoders
AUDIO: 44100 Hz, 2 ch, s16le, 32.0 kbit/2.27% (ratio: 4005->176400)
Selected audio codec: [ffcook] afm: ffmpeg (FFmpeg COOK audio decoder)

50.3 / 2.4
50.0 / 2.4
48.0 / 2.4


../bin/mplayer-stdout -cache 64 -vc null -vo null -ao pcm:fast:file=- mms://202.7.7.90/rnz-natrad > mms-wma-1.wav

Opening audio decoder: [ffmpeg] FFmpeg/libavcodec audio decoders
AUDIO: 16000 Hz, 2 ch, s16le, 20.0 kbit/3.91% (ratio: 2500->64000)
Selected audio codec: [ffwmav2] afm: ffmpeg (DivX audio v2 (FFmpeg))

22.0 / 2.5
22.5 / 2.5
21.9 / 2.5


../bin/mplayer-stdout -cache 64 -vc null -vo null -ao pcm:fast:file=- mms://dms-cl-021.skypro-media.net/radiox-64 > mms-mp3-1.wav

Opening audio decoder: [ffmpeg] FFmpeg/libavcodec audio decoders
AUDIO: 48000 Hz, 2 ch, s16le, 64.0 kbit/4.17% (ratio: 8001->192000)
Selected audio codec: [ffwmav2] afm: ffmpeg (DivX audio v2 (FFmpeg))

55.2 / 2.7
55.6
54.7

The numbers are 60 second averages CPU / memory.

[bpa]

I must re-run my test on FSG3 - my memory could be faulty. Burt if 107+ is 500MNhz and 128Mbytes RAM - I would expect the CPU load to be lower but perhaps the Marvel ARM core has cut some corners (e.g. cache) compared to Intel Xscale chip in FSG3.

When I run my test I send output to /dev/null as I think interface in NAS can be slow.

[flipflip]

I re-run the tests, too:

Results (60 seconds CPU average, measurement started after mplayer started playback):

MPlayer       Stream   Output          Readings           Average
---------------------------------------------------------------------
stdout-fixpt  RTSP 1   /dev/null       50.8  49.6  49.5     50.0
stdout-fixpt  RTSP 1   stdout > file   50.6  49.8  49.9     50.1
standard      RTSP 1   /dev/null       48.8  47.8  47.3     49.5
standard      RTSP 1   file            47.8  49.5  48.1     48.5

stdout-fixpt  WMA 1    /dev/null       28.8  28.4  24.2     27.1
stdout-fixpt  WMA 1    stdout > file   29.6  28.4  28.4     28.8

stdout-fixpt  WMA 2    /dev/null       55.5  55.3  53.9     54.9
stdout-fixpt  WMA 2    stdout > file   56.0  55.3  52.6     54.6


RTSP stream 1
  rtsp://a1355.l784539966.c7845.e.lr.akamaistream.net/live/D/1355/7845/v0001/reflector:39966
  Opening audio decoder: [ffmpeg] FFmpeg/libavcodec audio decoders
  AUDIO: 44100 Hz, 2 ch, s16le, 32.0 kbit/2.27% (ratio: 4005->176400)
  Selected audio codec: [ffcook] afm: ffmpeg (FFmpeg COOK audio decoder)
  AO: [pcm] 44100Hz 2ch s16le (2 bytes per sample)

WMA stream 1:
  mms://202.7.7.90/rnz-natrad
  Opening audio decoder: [ffmpeg] FFmpeg/libavcodec audio decoders
  AUDIO: 16000 Hz, 2 ch, s16le, 20.0 kbit/3.91% (ratio: 2500->64000)
  Selected audio codec: [ffwmav2] afm: ffmpeg (DivX audio v2 (FFmpeg))
  AO: [pcm] 16000Hz 2ch s16le (2 bytes per sample)

WMA stream 2:
  mms://dms-cl-021.skypro-media.net/radiox-64
  Opening audio decoder: [ffmpeg] FFmpeg/libavcodec audio decoders
  AUDIO: 48000 Hz, 2 ch, s16le, 64.0 kbit/4.17% (ratio: 8001->192000)
  Selected audio codec: [ffwmav2] afm: ffmpeg (DivX audio v2 (FFmpeg))
  PCM: Samplerate: 48000Hz Channels: Stereo Format s16le

Hardware:
  Synology Diskstation DS-107+.128, 128 MB RAM,  80 GB SATA
  500 MHz Marvell 88F5281 (CPU (ARM9E processor family, ARMv5TEJ architecture)

Mplayer flags used:
  stdout-fixpt, /dev/null: -cache 64 -vc null -vo null -ao pcm:fast:file=/dev/null $url
  stdout-fixpt, stdout   : -cache 64 -vc null -vo null -ao pcm:fast:file=- $url > $file
  standard, /dev/null    : -cache 64 -vc null -vo null -ao pcm:fast:file=/dev/null $url
  standard, file         : -cache 64 -vc null -vo null -ao pcm:fast:file=$file $url

Mplayer version:
  MPlayer SVN-r23339 (C) 2000-2007 MPlayer Team (mplayer-export-snapshot.tar.bz2, 2007-05-18)

Mplayer configure flags:
  CFLAGS="-I/volume1/SSODS/include -march=armv5te" ./configure \
  --target=armv5tel-linux-gnu --enable-armv5te --prefix=/volume1/SSODS \
  --disable-mencoder --language=en,de,fr --disable-jpeg  --disable-gif \
  --disable-tga  --disable-liblzo --disable-tv  --disable-dvb --disable-sdl \
  --disable-pnm --disable-xvid --disable-toolame --disable-mad --disable-png \
  --disable-vm  --enable-md5sum --disable-menu --charset=US-ASCII --disable-dvdread \
  --mandir=/volume1/SSODS/man --disable-tv  --disable-bitmap-font --disable-freetype \
  --disable-fontconfig --disable-inet6

And yes, I am 99.99999% sure that I have your Fixpt patch included. At least the libavcodec/*cook* files are there and there is #define COOK_FIXPOINT in cook.c.

Regards,
flip

[bpa]

Interesting results. What tool is used to measure CPU load ? I used the one inbuilt in to FSG3.

Another measure on optimisation - I measured CPU time taken to "play" a rm file to /dev/null. I used this to compare how fast fix vs floating pt.

I must also try WMA on my system.

[flipflip]

I used "top" from procps and set the update interval to 60 seconds. The readings are cpu_time_for_mplayer/wallclock_time*100.

Btw, somebody pointed me to this:

http://lists.arm.linux.org.uk/piperm...ay/039745.html

It seem like the 88F5281 has a floating point unit. But you need a recent GCC to support it. See also http://oinkzwurgl.org/ds107p for the links I received.

Unfortunately I don't have time in the near future to investigate this and try it.

Have a nice weekend!

[dxzk7]

Recently,I attempted transplant the MPlayer to the ARM platform.
After Compiled.I can use the player to broadcast mp3 on the ARM platform.
But cannot broadcast real audio.It seems that the cook decoder cannot work
in ARM platform.

Here's the console information:
/=====================system info==============/

/mnt/udisk$./mplayer lx.rm
...
CPU: ARM
Creating config file: /tmp/.mplayer/config

Failed to open /dev/rtc: No such file or directory (it should be readable by the
user.)
Playing lx.rm.
REAL file form
Stream description: Audio Stream
Stream mimetype: audio/x-pn-realaudio
Stream mimetype: logical-fileinfo
Clip info:
name: 11111
author: 11111
copyright: 11111
================================================== ========================
Opening audio decoder: [ffmpeg] FFmpeg/libavcodec audio decoders
AUDIO: 44100 Hz, 1 ch, s16le, 32.audio_open
prescaler = 2
prescaler = 2
s3c2410-uda1341-superlp: audio_set_dsp_speed:44100 prescaler:66
audio_clear_buf
buf 0: start ffc2b000 dma 839254016
buf 1: start ffc2d000 dma 839262208
buf 2: start ffc2f000 dma 839270400
buf 3: start ffc31000 dma 839278592
buf 4: start ffc33000 dma 839286784
buf 5: start ffc35000 dma 839294976
buf 6: start ffc37000 dma 839303168
buf 7: start ffc39000 dma 839311360
prescaler = 2
prescaler = 2
s3c2410-uda1341-superlp: audio_set_dsp_speed:44100 prescaler:66
0 kbit/4.54% (ratio: 4005->88200)
Selected audio codec: [ffcook] afm: ffmpeg (FFmpeg COOK audio decoder)
================================================== ========================
AO: [oss] 44100Hz 1ch s16le (2 bytes per sample)
Video: no video
Starting playback...
mixer ioctl 2147766295 unknown
Debug_A: -0.0 (unknown) of 267.0 (04:27.0) ??,?%
Debug_A: -0.0 (unknown) of 267.0 (04:27.0) ??,?%
Debug_A: -0.0 (unknown) of 267.0 (04:27.0) ??,?%
Debug_A: -0.0 (unknown) of 267.0 (04:27.0) ??,?%
Debug_A: -0.0 (unknown) of 267.0 (04:27.0) ??,?%
......

/=====================end info======================/


I have insert a break point in avcodec_decode_audio().the Application just
implement once. In normal condition.This function should be circulates the
execution. so I confuse very much this.

All in the Maillist group is expert.could you help me.
Any help is very appreciated!

[bpa]

Did you read the first post in this thread ?

It tells you how to change the mplayer code so that it will run on an ARM processor.

[dxzk7]

which toolchain are you used? arm-linux-gcc-3.3.2 or others?

here is my system information:
CPU samsung2410 arm920 200MHZ
OS linux-2.6.*
toolchain arm-linux-gcc-3.3.2
MPlayer-1.0pre8.tar

I replaced Cook.c&cook.h and insert the fixptcook.
of course. the transplant work defeated.
I get the the lastest MPlayer.but the error as before.