Problem occurred when I use multiprocessing with VPF encoding, need help! #161
Description
Describe the bug
Problem occurred when I run two encoding process on one GPU (GeForce RTX 3090) as follows:
Traceback (most recent call last):
File "./multiprocessing/ending_data_manager.py", line 39, in flush_avaliable_frames
video_writer.write(output_img)
File "./util/video.py", line 25, in write
success, encByteArray = self.convert_numpy_to_nv12(bgr_frame)
File "./util/video.py", line 45, in convert_numpy_to_nv12
success = self.nvEnc.EncodeSingleSurface(cvtSurface, encFrame, sync=False)
RuntimeError: NvEncoder : m_nvenc.nvEncOpenEncodeSessionEx(&encodeSessionExParams, &hEncoder) returned error 10
Description: EncodeAPI Internal Error.
at ./VideoProcessingFramework/PyNvCodec/TC/src/NvEncoder.cpp:84
It is worth mentioning that the problem only appear on my GeForce RTX 3090 GPU, but not appear on my Tesla V100 GPU.
I notice the error "EncodeAPI Internal Error." appears near the code "nvEncOpenEncodeSessionEx" in VPF source code, so I look up the NvEncodeAPI doc. It seems that the error means that a process cannot open an encoding session.
Further, I find the number of NVENC/CHIP is different between Tesla V100 and RTX 3090. As the website shows, Tesla V100 has 3 NVENC/CHIP, and RTX 3090 only has 1. I wonder whether the difference caused this error in my mutiprocessing application.
Please give me some advice, thanks a lot.
Desktop (please complete the following information):
- OS: ubuntu 20.04
- Nvidia driver version: 460.32.03
- CUDA Version: 11.1
- Video Codec SDK Version: 9.1.23
- Python Version: 3.8.5