I used "Media SDK 2013 R3 LinuxServers" in ubuntu 12.04 and cpu is i3-3225.
Today, I run "sample_vpp_drm" in samples fold to convert a YUY2 file to NV12.
In "top" result, I saw that cpu is too high.
In "intel_gpu_top" result, "render busy & bitstream busy & blitter busy" are all in idle.
Is VPP hardware accelerate supported in linux ?
If yes, how can I do about it?
While most of the basic functionality in VPP is hardware accelerated for 3rd and 4th Generation Core processors (Ivybridge and Haswell), many of the new features in the 16.3 release are hardware accelerated only for 4th Generation/Haswell. More details are in the release notes.
Another thing to keep in mind is that the VPP samples read and write raw frames. The biggest bottleneck by far for individual pipeline components is raw file I/O. If you're interested in VPP effects on transcode performance a better place to start may be sample_multi_transcode, which will give a better sense of how VPP functions without I/O overhead.