Intel Updates VA-API Video Acceleration Code

Intel's Video Acceleration API (VA-API) has seen updates to its core library as well as to the Intel-specific VA-API Intel driver.

The libva-1.0.16 that was announced by Haihao Xiang of Intel adds API support for handling data structures for JPEG baseline decoding, clearing up another bit of the API, and adding a test case for VA-API JPEG decoding. The announcement of this generic VA-API library update can be found on the libva mailing list.

At the same time, Haihao announced the VA-API Intel driver 1.0.18 update in this mailing list message. The VA-API driver implementation for Intel "GenX" hardware brings a JPEG decoding implementation for Intel Ivy Bridge hardware, support for a new Ivy Bridge chipset, support for new API calls, fixed decoding of MPEG-2 with implicit IQ matrices, fixing concurrent creation of VA objects (multi-threading safety), and fixed decoding of large resolution videos. VA-API can handle decoding of 4K resolution videos with this latest generation Ivy Bridge hardware. The newly-added Ivy Bridge chip for the Intel VA-API driver is a server chip with GT2 graphics that carries a PCI ID of 0x016a.

VA-API video encode/decode support continues to be one of the advantages of Intel's open-source Linux graphics stack. AMD has yet to document or publish code on their UVD video encode/decode engine for their Radeon hardware and Nouveau developers are still doing reverse-engineering and driver development, so those Linux desktop users interested in video playback with open-source mainline Linux graphics drivers are best off with Intel and using the widely-supported VA-API for now. If you don't mind using proprietary Linux graphics drivers, NVIDIA graphics cards with their binary blob backed by VDPAU continues to be another excellent choice.