Hi,
for me is pretty extraordinary how AMD can ship a new library (OVD) saying:
"Support for UVD video hardware component through OpenCL"
without documentation exceptuating headers.. and also without a simple source code sample..
Well let's compare to nvidia similar library cuvid.. they ship headers, documentation and source code
sample using both ogl and dx interop..
so what I ask is similar to cuvid for ovd i.e. ship documentation and sample using ogl and dx interop..
ah and they support on linux also since cuda 3.2 so I would want also linux support if it's not much
to ask..
well it has something better to cuvid namely that stereo profile seems is exposed so I ask is stereo
profile exposed in ovd working on uvd 3 hardware?
Another finding well ovd seems a simple wrapper around a opencl extension: cl_amd_open_video
with following functions:
clEnqueueRunVideoProgramAMD clGetVideoSessionInfoAMD clReleaseVideoSessionAMD clRetainVideoSessionAMD clCreateVideoSessionAMD
so why can't you document/expose this function directly and avoid using ovd and linking to another dll..
Well I have from header inferred a simple test sample..
but even before trying to use the meat of the library namely the OVDecodePicture function which
requires more knowledge about codecs than I know it still fails to work..
basically it fails in OVAcquireObject which seems to not like a object created like
clCreateBuffer(cxGPUContext, CL_MEM_READ_WRITE, 1920*1080*4*16, NULL, &ciErr1)
so what's the fail here is the size of the buffer?..
also some of the release/destroy functions is returning an error seems..
please help wanted..
unsigned int num=2;
ovdecode_device_info inf[10];
OVresult res=false;
res=OVDecodeGetDeviceInfo ( &num, inf);
printf("device %d ",num);
char *frase[]={"","OVD_H264_BASELINE_41",
"OVD_H264_MAIN_41", // H.264 bitstream acceleration main profile up to level 4.1
"OVD_H264_HIGH_41", // H.264 bitstream acceleration high profile up to level 4.1
"OVD_H264_BASELINE_51", // H.264 bitstream acceleration baseline profile up to level 5.1
"OVD_H264_MAIN_51", // H.264 bitstream acceleration main profile up to level 5.1
"OVD_H264_HIGH_51", // H.264 bitstream acceleration high profile up to level 5.1
"OVD_H264_STEREO_HIGH", // H.264 bitstream acceleration stereo high profile
"OVD_VC1_SIMPLE", // VC-1 bitstream acceleration simple profile
"OVD_VC1_MAIN", // VC-1 bitstream acceleration main profile
"OVD_VC1_ADVANCED", // VC-1 bitstream acceleration advanced profile}
};
for(int i=0; i<num; i++)
{
printf("device %d ",i);
printf("max_decode_stream %d\n",inf.max_decode_stream);
//cdDevice
ovdecode_cap *caps=(ovdecode_cap *)malloc(sizeof(ovdecode_cap)*inf.decode_cap_size);
//
{
res= OVDecodeGetDeviceCap (inf.device_id, inf.max_decode_stream, caps);
for(int j=0; j<inf.max_decode_stream; j++)
{
printf("%s ",frase[caps.profile]);
switch(caps.output_format)
{
case OVD_NV12_INTERLEAVED_AMD: printf("OVD_NV12_INTERLEAVED_AMD"); break;
case OVD_YV12_INTERLEAVED_AMD: printf("OVD_YV12_INTERLEAVED_AMD"); break;
}
printf("\n");
}
}
free(caps);
ov_handle h=OVCreateOVDHandleFromOPHandle(cmDevSrcA);
ov_session ses=OVDecodeCreateSession (cxGPUContext/*:cpPlatform*/, inf.device_id,
OVD_H264_HIGH_41, OVD_NV12_INTERLEAVED_AMD, 1920, 1080);
cmDevSrcA = clCreateBuffer(cxGPUContext, CL_MEM_READ_WRITE, 1920*1080*4*16, NULL, &ciErr1);
if (ciErr1 != CL_SUCCESS)
{
shrLog("Error in clCreateBuffer, Line %u in file %s !!!\n\n", __LINE__, __FILE__);
Cleanup(EXIT_FAILURE);
}
res=OVAcquireObject(ses, 1, &h, 0, NULL, NULL);
/*res= OVDecodePicture (ses, ovd_picture_parameter *picture_parameter_1,
void *picture_parameter_2,
unsigned int picture_parameter_2_size,
ovd_bitstream_data *bitstream_data,
unsigned int bitstream_data_size,
ovd_slice_data_control *slice_data_control,
unsigned int slice_data_control_size,
h, NULL, NULL, NULL, picture_id);*/
res=OVReleaseObject(ses,1, &h, 0, NULL, NULL);
res=OVReleaseOVDHandle ( h);
res=OVDecodeDestroySession (ses);
