third_party/opencv
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fixed and updated OpenCL-VA interoperability

Browse Source
This commit is contained in:
Odianosen Ejale
2020-09-11 11:29:31 +01:00
committed by Maksim Shabunin
parent 0dc28d3446
commit 862fc06b6f
11 changed files with 204 additions and 67 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/core/include/opencv2/core/va_intel.hpp
+3 -6
View File
@@ -29,14 +29,11 @@ namespace cv { namespace va_intel {
/** @addtogroup core_va_intel
This section describes Intel VA-API/OpenCL (CL-VA) interoperability.
To enable CL-VA interoperability support, configure OpenCV using CMake with WITH_VA_INTEL=ON . Currently VA-API is
supported on Linux only. You should also install Intel Media Server Studio (MSS) to use this feature. You may
have to specify the path(s) to MSS components for cmake in environment variables:
To enable basic VA interoperability build OpenCV with libva library integration enabled: `-DWITH_VA=ON` (corresponding dev package should be installed).
- VA_INTEL_IOCL_ROOT for Intel OpenCL (default is "/opt/intel/opencl").
To enable advanced CL-VA interoperability support on Intel HW, enable option: `-DWITH_VA_INTEL=ON` (OpenCL integration should be enabled which is the default setting). Special runtime environment should be set up in order to use this feature: correct combination of [libva](https://github.com/intel/libva), [OpenCL runtime](https://github.com/intel/compute-runtime) and [media driver](https://github.com/intel/media-driver) should be installed.
To use CL-VA interoperability you should first create VADisplay (libva), and then call initializeContextFromVA()
function to create OpenCL context and set up interoperability.
Check usage example for details: samples/va_intel/va_intel_interop.cpp
*/
//! @{