Memory issue for NVENC in linux

Hi,
Recently, I modified the source code in nvidia video codec sdk(version 9.0.20 and 11.1.5) for stress test,
specifically added a sentence in AppEncCuda.cpp, which is “fpIn.seekg(nRead, std::ios::beg)” after “std::streamsize nRead = fpIn.read(reinterpret_cast<char*>(pHostFrame.get()), nFrameSize).gcount();”, so that the same frame data can be encoded in a cycle. When I run cat /proc/(pid)/status in the process of cyclic coding, I find that the memory will grow every once in a while, and this growth process is continuous. Do I need to worry about the problem of continuous memory growth.

Hi,
I tried another stress test method, and kept cycling the process of init, encode and close. In this case, the memory kept growing. After about 20 times of cycling, the program exited abnormally and returned the error code of NV_ENC_ERR_UNSUPPORTED_DEVICE