VDPAU allows video programs to access the specialized video decoding ASIC on the GPU to offload portions of the video decoding process and video post-processing from the CPU to the GPU. Currently, the portions capable of being offloaded by VDPAU onto the GPU are motion compensation, inverse discrete cosine transform, VLD and deblocking for MPEG-1, MPEG-2, MPEG-4 ASP, H.264/MPEG-4 AVC and VC-1, WMV3/WMV9 encoded videos. Which specific codecs of these that can be offloaded to the GPU depends on the generation version of the GPU hardware.
History
VDPAU was originally designed by Nvidia for their PureVideo SIP block present on their GeForce 8 series and later GPUs. On March 9, 2015, Nvidia released VDPAU version 1.0 which supports High Efficiency Video Coding decoding for the Main, Main 4:4:4, Main Still Picture, Main 10, and Main 12 profiles.
VDPAU is implemented in X11 software device drivers, but relies on acceleration features in the hardware GPU. All Nvidia graphic cards for which the driver implements VDPAU are listed in Nvidia PureVideo. S3 Graphics added VDPAU to the Linux drivers of their Chrome 400video cards. As of version 14.02.17 of its Linux device driver, VDPAU is available with the S3 Chrome 430 GT, S3 Chrome 440 GTX, S3 Chrome 530 GT and the S3 Chrome 540 GTX hardware. ATI/AMD released an open source driver for Radeon HD 4000+ graphic cards featuring VDPAU acceleration. Intel does not offer VDPAU drivers, they only support their VA-API. It is, however, possible to use Intel's VA-API drivers by way of [|libvdpau-va-gl]. Nvidia hopes other GPU designers will make their products compatible with the open source VDPAU library and provide drivers with VDPAU acceleration by mentioning example names of hardware specific drivers for Intel and ATI: libvdpau_intel.so and libvdpau_ati.so. Intel has stated they are considering VDPAU. sunxi SoCs have experimental VDPAU implementation. Mesa as of v8.0 includes VDPAU for video cards that utilize Gallium3D.
Generic VDPAU driver
As of late 2013, there is an independently developed back-end driver that in turn uses OpenGL, and VA-API if available. It has been reported to work on some Intel graphics and Adobe Flash Player. The accelerated scaling with just OpenGL functionality is needed mostly because of Flash player, which uses un-accelerated scaling if VDPAU is unavailable. Almost all other video software that runs on GNU/Linux or FreeBSD supports Xv. It's essential for full-screen video on slower computers without native VDPAU support. The VDPAU to VA-API translation for HW decoding is useful with recent Intel graphics hardware, as some software supports HW decoding through VDPAU but not VA-API.
VDPAU can also be used as a backend for VA-API and OpenMAX IL, which themselves cover a subset of the VDPAU capabilities; so any software that uses the VA-API or OpenMAX IL is also partly capable of using VDPAU.
Nvidia VDPAU Feature Sets
Nvidia VDPAU Feature Sets are different hardware generations of GPU's supporting different levels of hardware decoding capabilities. For feature sets A, B and C, the maximum video width and height are 2048 pixels, minimum width and height 48 pixels, and all codecs are currently limited to a maximum of 8192 macroblocks. Partial acceleration means that VLD decoding is performed on the CPU, with the GPU only performing IDCT, motion compensation and deblocking. Complete acceleration means that the GPU performs all of VLD, IDCT, motion compensation and deblocking. ; Feature Set A ; Feature Set B ; Feature Set C ; Feature Set D ; Feature Set E ; Feature Set F ; Feature Set G ; Feature Set H
libvdpau standalone VDPAU library
The libvdpau standalone VDPAU library is distributed by Nvidia independently of their proprietary Linux graphics driver in an effort to help the adoption of VDPAU by those outside of Nvidia. This open source library package contains a wrapper library and a debugging library allowing other manufacturers to implement VDPAU in their device drivers.
