GStreamer on TI DaVinci and OMAP
Each release of the TI GStreamer DMAI plug-in should have as good or better perfmance characteristics than the previous version. The performance characteristics include:
- ARM CPU usage
- DSP CPU usage (if the chip has a user accessible DSP)
- Dropped frames
In order to characterize performance, appropriate tools and test cases are required; neither of which are in place at this time. As performance analysis efforts make progress, this page will be updated. Performance analysis is a good opportunity for community contributions.
Performance Analysis Tools
GStreamer Tracing Library
The gst-tracelib library hooks into some gstreamer key functions and logs the behavior. When the application exits it displays some general statistics. Further analysis can be done based on the data written to the logfile
The GStreamer tracing library hasn't been integrated into the TI GStreamer DMAI plug-in build process yet.
Example GStreamer Tracing Library Execution
Early performance analysis is promising. Here is an example run:
LD_PRELOAD=/usr/lib/gst-tracelib/libgsttracelib.so gst-launch filesrc location = /RRunInterview.mp4 ! qtdemux ! dmaidec_mpeg4 ! fakesink
The generated log file was process using the command (assuming log file is /tmp/gstlt.log) :
./src/gsttl_splitlog.py ./src/gsttl_plot.sh --format=png | gnuplot
The file with the output plots is attached (with example graph shown to the right).
|Item ID||Associated Item||Comment|
|No Associated Items Found|