Hi,
I’ve got a library I’d like to build for the Xavier. It’s ported from a system using cuda 9, afaik. I’ve got jetpack 4.1.1 installed and as far as I can see tensorflow, numpy, etc… all got installed correctly. nvcc reports:
Cuda compilation tools, release 10.0, V10.0.117
My C source requires:
#include <cuda_runtime.h>
#include <cuda_gl_interop.h>
which I can find under /usr/local/cuda-10.0/targets/aarch64-linux/include/ but also requires:
#include “dynlink_cuda.h”
Where should this file be, or where should it come from ?
the including file is
nvidia@jetson:~/path$ head dynlink_cuviddec.h
/*
- This copyright notice applies to this header file only:
- Copyright (c) 2010-2016 NVIDIA Corporation
…
EDIT: after some investigation, it seems the codebas I’m porting is relying on
/usr/local/cuda/include/dynlink_nvcuvid.h://! \fn CUresult CUDAAPI cuvidCreateVideoParser(CUvideoparser *pObj, CUVIDPARSERPARAMS *pParams)
/usr/local/cuda/include/dynlink_nvcuvid.h://! \fn CUresult CUDAAPI cuvidDestroyVideoParser(CUvideoparser obj)
/usr/local/cuda/include/dynlink_nvcuvid.h://! \fn CUresult CUDAAPI cuvidParseVideoData(CUvideoparser obj, CUVIDSOURCEDATAPACKET *pPacket)
/usr/local/cuda/include/dynlink_nvcuvid.h:typedef CUresult CUDAAPI tcuvidCreateVideoParser(CUvideoparser *pObj, CUVIDPARSERPARAMS *pParams);
/usr/local/cuda/include/dynlink_nvcuvid.h:typedef CUresult CUDAAPI tcuvidDestroyVideoParser(CUvideoparser obj);
/usr/local/cuda/include/dynlink_nvcuvid.h:typedef CUresult CUDAAPI tcuvidParseVideoData(CUvideoparser obj, CUVIDSOURCEDATAPACKET *pPacket);
/usr/local/cuda/include/dynlink_nvcuvid.h:typedef void *CUvideoparser;
and those can’t be found on Xavier AGX cuda 10.0 install. Is there a porting guide, or ar those deprecated ?
Thanks,
J.