diff options
| author | Chuanbo Weng <chuanbo.weng@intel.com> | 2015-11-11 17:22:29 +0800 |
|---|---|---|
| committer | Yang Rong <rong.r.yang@intel.com> | 2015-11-17 11:20:50 +0800 |
| commit | 804bf86528eb80abd76c1e7af4cfdc79fb49a53d (patch) | |
| tree | ab6401b5399e9446a030d5dc5074eb7a901932ce /docs | |
| parent | 03abc25a2cc4adde23e54d340a9c9c33886f21a5 (diff) | |
| download | beignet-804bf86528eb80abd76c1e7af4cfdc79fb49a53d.tar.gz | |
Full support of cl_intel_motion_estimation extension.
The following items are supported in this commit:
1. Return residuals.
2. All types of mb_block_type, subpixel_mode, sad_adjust_mode in
cl_motion_estimation_desc_intel.
After this commit, cl_intel_motion_estimation is fully supported.
Signed-off-by: Chuanbo Weng <chuanbo.weng@intel.com>
Reviewed-by: "Yang, Rong R" <rong.r.yang@intel.com>
Diffstat (limited to 'docs')
| -rw-r--r-- | docs/howto/video-motion-estimation-howto.mdwn | 26 |
1 files changed, 9 insertions, 17 deletions
diff --git a/docs/howto/video-motion-estimation-howto.mdwn b/docs/howto/video-motion-estimation-howto.mdwn index d9edc9b9..8deaa61a 100644 --- a/docs/howto/video-motion-estimation-howto.mdwn +++ b/docs/howto/video-motion-estimation-howto.mdwn @@ -1,21 +1,15 @@ Video Motion Vector HowTo ========================== -Beignet now supports cl_intel_accelerator and part of cl_intel_motion_estimation, which -are Khronos official extensions. It provides a hardware acceleration of video motion +Beignet now supports cl_intel_accelerator and cl_intel_motion_estimation, which are +Khronos official extensions. It provides a hardware acceleration of video motion vector to users. -Supported hardware platform and limitation ------------------------------------------- +Supported hardware platform +--------------------------- -Only 3rd Generation Intel Core Processors is supported for vme now. And now we just -implement this part of cl_intel_motion_estimation for motion vector computation(residuals -can not be returned yet) on 3rd Generation Intel Core Processors: - mb_block_type = CL_ME_MB_TYPE_16x16_INTEL - subpixel_mode = CL_ME_SUBPIXEL_MODE_INTEGER_INTEL - search_path_type = CL_ME_SEARCH_PATH_RADIUS_2_2_INTEL / CL_ME_SEARCH_PATH_RADIUS_4_4_INTEL - / CL_ME_SEARCH_PATH_RADIUS_16_12_INTEL -We will fully support cl_intel_motion_estimation in the future. +Only 3rd Generation Intel Core Processors is supported for vme now. We will consider +to support more platforms if necessary. Steps ----- @@ -23,15 +17,13 @@ Steps In order to use video motion estimation provided by Beignet in your program, please follow the steps as below: -- Create a cl_accelerator_intel object using extension API clCreateAcceleratorINTEL, with - the following parameters: +- Create a cl_accelerator_intel object using extension API clCreateAcceleratorINTEL, like + this: _accelerator_type_intel accelerator_type = CL_ACCELERATOR_TYPE_MOTION_ESTIMATION_INTEL; cl_motion_estimation_desc_intel vmedesc = {CL_ME_MB_TYPE_16x16_INTEL, CL_ME_SUBPIXEL_MODE_INTEGER_INTEL, CL_ME_SAD_ADJUST_MODE_NONE_INTEL, - CL_ME_SEARCH_PATH_RADIUS_16_12_INTEL( - or CL_ME_SEARCH_PATH_RADIUS_2_2_INTEL - or CL_ME_SEARCH_PATH_RADIUS_4_4_INTEL) + CL_ME_SEARCH_PATH_RADIUS_16_12_INTEL }; - Invoke clCreateProgramWithBuiltInKernels to create a program object with built-in kernels |