<html><body><div style="color:#000; background-color:#fff; font-family:HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif;font-size:12pt"><div class="" style=""><pre class="" style="">I'm using libavcodec to perform decoding of H.264 frames.
I'm on a Linux environment (14.04 Xubuntu), and Intel Haswell (Pentium grade) CPU.
My program decodes the frames, without rendering them on the screen. With 4 simultaneous decodes of 1080p resolution and 15 fps, the CPU utilization is around 90% (not bad), but the load average shown by 'top' is 20+
This looks excessive. I don't think I'm using the hardware decoding ability of the Haswell CPU. Could someone please advise how to find out if I'm effectively using the decoder?
The API calls I'm making are pretty standard -
avcodec_register_all();
avcodec_alloc_context3(NULL);
avcodec_find_decoder(AV_CODEC_ID_H264);
av_parser_init(AV_CODEC_ID_H264);
avcodec_open2();
The above calls are made during initialization. The below calls are made on each frame -
av_parser_parse2
avcodec_decode_video2
The following points are probably relevant -
a) The default context allocated by avcodec_alloc_context3() does not have any hw_accel associated with it. I allocated a h264_vaapi accelerator, but this doesn't lead to any improvement.
b) The codec capabilities does not have the HW_ACCEL bit set.</pre><pre class="" style=""><span style="font-size: 12pt;" class="">c) The same code worked better when used with Ubuntu 11.10 on a Sandybridge Pentium grade CPU and FFMPEG 0.9</span><br style="">
Any help would be appreciated.
Thanks,
Prashanth</pre><pre class="" style=""><br></pre></div><div style="color: rgb(0, 0, 0); font-size: 16px; font-family: HelveticaNeue, 'Helvetica Neue', Helvetica, Arial, 'Lucida Grande', sans-serif; background-color: transparent; font-style: normal;" class=""> </div></div></body></html>