GPU Acceleration

OvenMediaEngine supports GPU-based hardware decoding and encoding. Currently supported GPU acceleration devices are Intel's QuickSync and NVIDIA. This article explains how to install the drivers for OvenMediaEngine and set up the configuration to use your GPU.

1. Install Drivers

NVIDIA

1. Install NVIDIA GPU Driver

If you are using an NVIDIA graphics card, please refer to the following guide to install the driver. The OS that supports installation with the provided script are CentOS 7/8 and Ubuntu 18/20 versions. If you want to install the driver in another OS, please refer to the manual installation guide document.

CentOS environment requires the process of uninstalling the nouveau driver. After uninstalling the driver, the first reboot is required, and a new NVIDIA driver must be installed and rebooted. Therefore, two install scripts must be executed.

(curl -LOJ https://github.com/OvenMediaLabs/OvenMediaEngine/archive/master.tar.gz && tar xvfz OvenMediaEngine-master.tar.gz)
OvenMediaEngine-master/misc/install_nvidia_driver.sh

How to check driver installation

After the driver installation is complete, check whether the driver is operating normally with the nvidia-smi command.

$ nvidia-smi
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.288.01             Driver Version: 535.288.01   CUDA Version: 12.2     |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |
|                                         |                      |               MIG M. |
|=========================================+======================+======================|
|   0  NVIDIA GeForce GTX 1050        On  | 00000000:3B:00.0  On |                  N/A |
| 20%   39C    P8              N/A /  75W |    204MiB /  2048MiB |      0%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+
|   1  NVIDIA RTX 4000 SFF Ada ...    On  | 00000000:AF:00.0 Off |                  Off |
| 30%   38C    P8               5W /  70W |    171MiB / 20475MiB |      0%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+
                                                                                         
+---------------------------------------------------------------------------------------+
| Processes:                                                                            |
|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |
|        ID   ID                                                             Usage      |
|=======================================================================================|
|    0   N/A  N/A    802940      C   ...prise/src/bin/DEBUG/OvenMediaEngine       40MiB |
|    1   N/A  N/A    802940      C   ...prise/src/bin/DEBUG/OvenMediaEngine      158MiB |
+---------------------------------------------------------------------------------------+

2 . Prerequisites

If you have finished installing the driver to use the GPU, you need to reinstall the open source library using Prerequisites.sh . The purpose is to allow external libraries to use the installed graphics driver.

OvenMediaEngine-master/misc/prerequisites.sh --enable-nv

NVIDIA with Docker

Netint VPU Ni Logan

2. Build & Run

Please refer to the link for how to build and run.

Getting Started

Intructions on running Docker

you must include the --gpus all option when running Docker

docker run -d ... --gpus all airensoft/ovenmediaengine:dev

3. Configuration

To use hardware acceleration, set the HardwareAcceleration option to true under OutputProfiles. If this option is enabled, a hardware codec is automatically used when creating a stream, and if it is unavailable due to insufficient hardware resources, it is replaced with a software codec.

<OutputProfiles>
    <HWAccels>
        <!-- 
        Setting for Hardware Modules.
            - nv : Nvidia Video Codec SDK
            - xma :Xilinx Media Accelerator
            - nilogan: Netint VPU

        You can use multiple modules by separating them with commas.
        For example, if you want to use xma and nv, you can set it as follows.

        <Modules>[ModuleName]:[DeviceId],[ModuleName]:[DeviceId],...</Modules>
        <Modules>nv:0,xma:0</Modules>
        -->
        <Decoder>
                <Enable>true</Enable>
                <Modules>nv</Modules>
        </Decoder>
        <Encoder>
                <Enable>true</Enable>
                <Modules>nv</Modules>
        </Encoder>
    </HWAccels>
    
    <OutputProfile>
        ...
    </OutputProfile>
</OutputProfiles>

Configuration

Appendix. Support Format

The codecs available using hardware accelerators in OvenMediaEngine are as shown in the table below. Different GPUs support different codecs. If the hardware codec is not available, you should check if your GPU device supports the codec.

Device	H264	H265	VP8	VP9
NVIDIA	D / E	D / E	-	-
Docker on NVIDIA Container Toolkit	D / E	D / E	-	-
Xilinx U30MA	D / E	D / E

D : Decoding, E : Encoding

Reference

• NVIDIA NVDEC Video Format : https://en.wikipedia.org/wiki/Nvidia_NVDEC

• NVIDIA NVENV Video Format : https://en.wikipedia.org/wiki/Nvidia_NVENC

• CUDA Toolkit Installation Guide : https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#introduction

• NVIDIA Container Toolkit : https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/arch-overview.html#arch-overview

• Xilinx Video SDK : https://xilinx.github.io/video-sdk/v3.0/index.html