LeopardBoard DM368 HD H.264 encode

With DVSDK working on LeopardBoard DM368, Now let's trying video encode function.

1.Check kernel log to see if the dsplink related modules are loaded.

CMEMK module: built on Jun 14 2011 at 18:02:26
  Reference Linux version 2.6.32
  File /home/gigijoe/LeopardBoard/DM368/original/ti-dvsdk_dm368-evm_4_02_00_06/linuxutils_2_26_01_02/packages/ti/sdo/linuxutils/cmem/src/module/cmemk.c
allocated heap buffer 0xc6000000 of size 0x4400000
heap fallback enabled - will try heap if pool buffer is not available
CMEM Range Overlaps Kernel Physical - allowing overlap
CMEM phys_start (0x1000) overlaps kernel (0x80000000 -> 0x83c00000)
cmemk initialized
IRQK module: built on Jun 14 2011 at 18:02:29
  Reference Linux version 2.6.32
  File /home/gigijoe/LeopardBoard/DM368/original/ti-dvsdk_dm368-evm_4_02_00_06/linuxutils_2_26_01_02/packages/ti/sdo/linuxutils/irq/src/module/irqk.c
irqk initialized
EDMAK module: built on Jun 14 2011 at 18:02:30
  Reference Linux version 2.6.32
  File /home/gigijoe/LeopardBoard/DM368/original/ti-dvsdk_dm368-evm_4_02_00_06/linuxutils_2_26_01_02/packages/ti/sdo/linuxutils/edma/src/module/edmak.c

2.If there are errors check /etc/init.d/loadmodule-rc. Below is mine

#!/bin/sh

# Default DM368 EVM Memory Map
#
# Start Addr    Size    Description
# -------------------------------------------
# 0x00001000    32K     ARM TCM memory
# 0x80000000    48 MB   Linux
# 0x83000000    12 MB   Video driver memory (Linux)
# 0x83C00000    68 MB   CMEM
# 0x88000000    BOTTOM  ADDRESS

load () {
    modprobe cmemk phys_start=0x83C00000 phys_end=0x88000000 allowOverlap=1 phys_start_1=0x00001000 phys_end_1=0x00008000 pools_1=1x28672 useHeapIfPoolUnavai
lable=1
    modprobe irqk
    modprobe edmak
    modprobe dm365mmap
    mknod /dev/dm365mmap c `awk "\\$2==\"dm365mmap\" {print \\$1}" /proc/devices` 0
}

unload () {
    rmmod cmemk 2>/dev/null
    rmmod irqk  2>/dev/null
    rmmod edmak 2>/dev/null
    rmmod dm365mmap 2>/dev/null
    rm -f /dev/dm365mmap
}

case "$1" in
      start)
            load
             ;;
       stop)
            unload          
             ;;
       restart)
            unload          
            load
             ;;
        *)
             echo "$0 "
             ;;
esac

3.Decode test to verify if DSP works.

$ cd /usr/share/ti/ti-dmai-apps
$ ./video_decode_io2_dm368.x470MV -c h264dec -e decode -i /usr/share/ti/data/videos/davincieffect.264 -n 30 -o output.yuv
Starting application...
Loader created and primed, entering main loop..
Loaded Frame 1: frame size 401
Loaded Frame 2: frame size 36
Loaded Frame 3: frame size 804
Loaded Frame 4: frame size 844
Loaded Frame 5: frame size 863
Loaded Frame 6: frame size 933
Loaded Frame 7: frame size 1189
Loaded Frame 8: frame size 1253
Loaded Frame 9: frame size 1509
Display Frame 1: Wrote 420P frame size 1382400 (1280x720) to disk
Loaded Frame 10: frame size 1573
Display Frame 2: Wrote 420P frame size 1382400 (1280x720) to disk
Loaded Frame 11: frame size 1829

4.Now, it's time to encode video from camera. I got mt9p031 (LI-BCM5M03).
If you got kernel log below, the camera is ready

...
Leopard: switch to HD imager-MT9P031 video input
mt9p031 1-0048: Detected a MT9P031 chip ID 1801
mt9p031 1-0048: mt9p031 1-0048 decoder driver registered !!
vpfe-capture vpfe-capture: v4l2 sub device mt9p031 registered
...

5.Test if camera works

$ cd /usr/share/ti/ti-dmai-apps
$ ./video_loopback_dm368.x470MV -O component -I camera -y5
davinci_resizer davinci_resizer.2: RSZ_G_CONFIG:0:1:124
vpfe-capture vpfe-capture: IPIPE Chained
vpfe-capture vpfe-capture: Resizer present
Leopard: switch to HD imager-MT9P031 video input
-----Exposure time = 456
-----Exposure time = 456
vpfe-capture vpfe-capture: width = 1920, height = 1080, bpp = 1
vpfe-capture vpfe-capture: adjusted width = 1920, height = 1080, bpp = 1, bytesperline = 1920, sizeimage = 3110400
vpfe-capture vpfe-capture: width = 1920, height = 1080, bpp = 1
vpfe-capture vpfe-capture: adjusted width = 1920, height = 1080, bpp = 1, bytesperline = 1920, sizeimage = 3110400
davinci_v4l2 davinci_v4l2.1: Before finishing with S_FMT:
layer.pix_fmt.bytesperline = 1920,
 layer.pix_fmt.width = 1920,
 layer.pix_fmt.height = 1080,
 layer.pix_fmt.sizeimage =3110400
davinci_v4l2 davinci_v4l2.1: pixfmt->width = 1920,
 layer->layer_info.config.line_length= 1920

6.There's a application can do H.264 RTP which included in my dmai_2_20_00_15 patch called video_encode_rtp.

This application get frames from camera through VPFE then encoded in H.264 format packeted with H264RTP packet then send to network.

From LeopardBoard DM368

$ cd /usr/share/ti/ti-dmai-apps
$ ./video_encode_rtp_dm368.x470MV -I camera -y 4 -c h264enc -i /dev/video0 -o 192.168.168.72 -p 1234 -e encode -b 2000000

-I camera = camera input
-y 4 = 720P,
-o 192.168.168.72 = Send to 192.168.168.72
-p 1234 = port 1234
-b 2000000 = Bit rate 2000000

I'm decoding the video stream with gstreamer on PC Ubuntu.

$ gst-launch udpsrc port=1234 caps="application/x-rtp,media=(string)video,clock-rate=(int)90000,encoding-name=(string)H264" ! gstrtpjitterbuffer latency=100 ! rtph264depay ! ffdec_h264 ! xvimagesink

If the video frame sometimes droped, try larger gstrtpjitterbuffer latency = 200 / 300 / 500.

7.Full HD

From LeopardBoard DM368

-y 5 = 1080P

$ ./video_encode_rtp_dm368.x470MV -I camera -y 5 -c h264enc -i /dev/video0 -o 192.168.168.72 -p 1234 -e encode -b 2000000