<div dir="ltr">Dear Experts,<div><br></div><div>we're using x264 CLI to transcode 1080p video to 240p, using default setting [--vf resize:432,240]</div><div>we noticed that the transcode was slow, achieving only ~70fps on an 2-cpu, 8-core E5520 "Gainestown" machine, while utilizing only a very small portion of the available CPU.</div>
<div>my analysis led me to think that the bottleneck is the down-scaling stage.</div><div><br></div><div>to increase the encoding speed, we ran 4 instances of unmodified x264 in parallel, each working on a fragment of the original source media, using the exact same parameters and settings.</div>
<div>after all x264 instances have finished, we concatenated the output files using avconv, while taking care of [hopefullly] correctly setting PTS and DTS for the combined file.</div><div>this achieved the desired 100% cpu utilization.</div>
<div><br></div><div>Unfortunately we still have an issue when playing the file, where a small "jump" or "lag" is just noticeable at the exact location where these separately encoded files were sewn together.</div>
<div>as far as we can tell, all PTSs and DTSs are correctly increasing.</div><div><br></div><div>is there some information in the first SPS, PPS or keyframe that could cause this effect? maybe by causing the decoder to reset itself somehow?</div>
<div>we would be happy to share a small input/output sample [through <a href="http://datafilehost.com">datafilehost.com</a> or any other method] along with full command-line and logs</div><div><br></div><div>Many thanks,</div>
<div><div><div><br></div>-- <br>Aviad Rozenhek<br>
</div></div></div>