Visible to Intel only — GUID: GUID-B4B5BD72-394B-4464-BDA9-5A23AE5EB2B9
Use Intel® oneVPL Core API for the Best Portability
Intel(R) Media SDK API Features Removed from oneVPL
Key Features Added to oneVPL 2.x API
AV1 Encode Features Added to Intel® oneVPL
Other Differences Between Intel(R) Media SDK and oneVPL
oneVPL Hardware Support Details
Why oneVPL Dispatcher Is Safer
Visible to Intel only — GUID: GUID-B4B5BD72-394B-4464-BDA9-5A23AE5EB2B9
Use Intel® oneVPL Core API for the Best Portability
It is optional to convert to complete 2.x enhanced API features to use oneVPL.
As you can see in the tutorial in the Transition Tutorial, applications compiling with VPL API headers or linking to VPL dispatcher may still use 1.x MFXInit* session initialization with compile-time deprecation warnings.
Intel(R) Media SDK code is fully functional in oneVPL, excepting removed features. The 1.x API limitation with MFXInit* is not a concern if the code is only written to use 1.x API features.
The steps covered in the tutorial are intended to leave the code for the rest of the application intact. New features can be adapted to your schedule. By simply updating to the oneVPL API and linking to the dispatcher, your project benefits from continuing maintenance and support.
The use of the core API provides several advantages:
Clear assurance by the compiler that there are no hidden removed feature pathways that can cause problems later.
Codeworks across Intel(R) Media SDK and oneVPL Intel(R) GPU implementations.
Departing from the core API using 2.x enhanced API features is not required. However, consider internal memory, that is the primary 2.x enhanced API feature. Internal memory makes new code development easier since oneVPL handles the complexities of video surface management. It can also make interoperability with other graphics APIs easier.
NOTE:
Internal memory is not supported by legacy GPU hardware. The use of internal memory makes your project incompatible with hardware supported by the Intel(R) Media SDK runtime implementation.
External surface allocations continue to be supported by oneVPL 2.x API. Intel(R) has no plans to deprecate them.