ffmpeg系列之linux平台下支持cuvid硬解码

前言

正文

1.编译安装需要的环境

参考　jianshu.com/p/032b47c48ada 注意这是ｗｉｎｄｏｗ的
linux可以参考官网说明
https://developer.download.nvidia.cn/designworks/ffmpeg/secure/Using_FFmpeg_with_NVIDIA_GPU_Hardware_Acceleration_v01.4.pdf?dp1Oy2tu6NTWK0yia_DyQgFH2-yb2N-o-jb6UXCstzU6Cj92bTb0F93KFN0WnhYbX4tf1Hvn-Tn4IgWlViwWrCtzb4qJxuyAyqD2ISAOqR6XAC8G52EQSjCMg1IYWF-0KmotmXVrMT8jM14MEsOnhjt6gi8eV6JnS_Xfh_LlCzpHd8MgC6SG_YwvwDtJeCABjGX5giY

2.2.2.1Compiling for LinuxFFmpegwith NVIDIA GPU acceleration is supported on all Linux platforms.To compileFFmpegon Linux, do the following:

1.Clone ffnvcodec
git clone https://git.videolan.org/git/ffmpeg/nv-codec-headers.git

2.Install ffnvcodec
cd nv-codec-headers && sudo make install && cd ..

3.Clone FFmpeg's public GIT repository.
git clone https://git.ffmpeg.org/ffmpeg.git ffmpeg/

4.Install necessary packages.
sudo apt-get install build-essential yasm cmake libtool libc6 libc6-dev unzip wget libnuma1 libnuma-dev

5.Configure
# ffmpeg 3.x
./configure --enable-nonfree --prefix=/usr/local/ffmpeg3.4.9 \
--enable-cuda-sdk --enable-cuvid  --enable-nvenc --enable-libnpp \
--extra-cflags=-I/usr/local/cuda/include \
--extra-ldflags=-L/usr/local/cuda/lib64 \
--enable-gpl --enable-shared --enable-pic --disable-libx264 --disable-libx265

#ffmpeg5.0
./configure --enable-nonfree --prefix=/usr/local/ffmpeg5.0 \
--enable-cuda-nvcc --enable-cuvid  --enable-nvenc --enable-libnpp \
--extra-cflags=-I/usr/local/cuda/include \
--extra-ldflags=-L/usr/local/cuda/lib64 \
--enable-libx264 --enable-libx265 --enable-gpl --enable-shared --enable-pic

6.Compile
make -j8

7.Install the libraries.
sudo make install

ps: ffmpeg-5.0直接编译成功，nvidia驱动版本470

【问题】编译ffmpeg-3.x时提示 nvcc fatal : Unsupported gpu architecture ‘compute_30’
【解决办法】 大致原因是ffmpeg的3.x版本编译时设置了显卡架构compute_30, 而本机的显卡对不上。设置位于ffmpeg/configure文件中。

参考http://arnon.dk/matching-sm-architectures-arch-and-gencode-for-various-nvidia-cards/ 查询显卡对应的架构应该是什么。
输入以下指令查询显卡信息，这里是GeForce GTX 1080 Ti

$ lspci | grep -i nvidia
01:00.0 VGA compatible controller: NVIDIA Corporation GP102 [GeForce GTX 1080 Ti] (rev a1)
01:00.1 Audio device: NVIDIA Corporation GP102 HDMI Audio Controller (rev a1)
03:00.0 VGA compatible controller: NVIDIA Corporation GP102 [GeForce GTX 1080 Ti] (rev a1)
03:00.1 Audio device: NVIDIA Corporation GP102 HDMI Audio Controller (rev a1)

对应的架构为compute_61

修改配置，重新编译即可成功

2.配置可执行文件和库到系统路径

sudo vim ~/.bashrc
# 在最末添加配置
export LD_LIBRARY_PATH=/usr/local/ffmpeg{version}/lib:$LD_LIBRARY_PATH
export PATH=/usr/local/ffmpeg{version}/bin:$PATH
#重新激活配置
source ~/.bashrc

## 测试
$ ffmpeg -version
ffmpeg version 3.4.9 Copyright (c) 2000-2021 the FFmpeg developers
built with gcc 7 (Ubuntu 7.5.0-3ubuntu1~18.04)
configuration: --enable-nonfree --prefix=/usr/local/ffmpeg3.4.9 --enable-cuda-sdk --enable-libnpp --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64 --enable-libx265 --enable-gpl --enable-shared --enable-pic --enable-libx264
libavutil      55. 78.100 / 55. 78.100
libavcodec     57.107.100 / 57.107.100
libavformat    57. 83.100 / 57. 83.100
libavdevice    57. 10.100 / 57. 10.100
libavfilter     6.107.100 /  6.107.100
libswscale      4.  8.100 /  4.  8.100
libswresample   2.  9.100 /  2.  9.100
libpostproc    54.  7.100 / 54.  7.100

3.测试使用

ffmpeg -c:v h264_cuvid -i rtsp://xxxx output.mp4

-c:v 命令也就是-codec:v　相当于-vcodec , 代表视频编解码器
测试结果发现：ｃｐｕ依然居高不下,gpu没动静，姿势不对

翻了下书：
１）查看解码器参数
ffmpeg -h decoder=h264_cuvid

ffmpeg version 4.1.4 Copyright (c) 2000-2019 the FFmpeg developers
  built with gcc 6.5.0 (Ubuntu 6.5.0-2ubuntu1~18.04) 20181026
  configuration: --enable-shared --prefix=/usr/local/ffmpeg --enable-gpl --enable-libx264 --enable-cuda --enable-nvenc --enable-nonfree --enable-libnpp --extra-cflags=-I/usr/local/cuda/include --extra-cflags=-fPIC --extra-ldflags=-L/usr/local/cuda/lib64
  libavutil      56. 22.100 / 56. 22.100
  libavcodec     58. 35.100 / 58. 35.100
  libavformat    58. 20.100 / 58. 20.100
  libavdevice    58.  5.100 / 58.  5.100
  libavfilter     7. 40.101 /  7. 40.101
  libswscale      5.  3.100 /  5.  3.100
  libswresample   3.  3.100 /  3.  3.100
  libpostproc    55.  3.100 / 55.  3.100
Decoder h264_cuvid [Nvidia CUVID H264 decoder]:
    General capabilities: delay avoidprobe hardware 
    Threading capabilities: none
    Supported hardware devices: cuda 
    Supported pixel formats: cuda nv12 p010le p016le
h264_cuvid AVOptions:
  -deint             <int>        .D.V..... Set deinterlacing mode (from 0 to 2) (default weave)
     weave                        .D.V..... Weave deinterlacing (do nothing)
     bob                          .D.V..... Bob deinterlacing
     adaptive                     .D.V..... Adaptive deinterlacing
  -gpu               <string>     .D.V..... GPU to be used for decoding
  -surfaces          <int>        .D.V..... Maximum surfaces to be used for decoding (from 0 to INT_MAX) (default 25)
  -drop_second_field <boolean>    .D.V..... Drop second field when deinterlacing (default false)
  -crop              <string>     .D.V..... Crop (top)x(bottom)x(left)x(right)
  -resize            <string>     .D.V..... Resize (width)x(height)

可以看到解码支持的像素格式有： cuda nv12 p010le p016le

２）换个姿势测试
突然发现上面ｃｐｕ居高不下，有可能是编码用的，所以给换下

ffmpeg -hwaccel cuvid -vcodec h264_cuvid -i rtsp://xxxx -c:v h264_nvenc output.mp4

［问题１］：这里指定了硬编码和硬解码，结果报错了

 Stream #0:0 -> #0:0 (h264 (h264_cuvid) -> h264 (h264_nvenc))
Press [q] to stop, [?] for help
[h264_nvenc @ 0x561de9b93780] Driver does not support the required nvenc API version. Required: 9.1 Found: 9.0
[h264_nvenc @ 0x561de9b93780] The minimum required Nvidia driver for nvenc is 390.25 or newer

看到第一句提示，我想起来nv-codec-headers.git下来是9.1版本　而不是9.0
看到第二句，我查了下nvidia驱动版本,430.26符合条件

因此去github上下载9.0版本
https://github.com/FFmpeg/nv-codec-headers/releases/tag/n9.0.18.3
（话说有人说直接checkout到9.0版本，怎么操作的，告诉我一下）

安装完后重新编译安装ffmpeg

./configure --enable-shared --prefix=/usr/local/ffmpeg --enable
-gpl --enable-libx264 --enable-cuda-sdk --enable-cuvid --enable-nvenc --enable-nonfree --enable-libnpp --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64

结束之后，重新使用硬编解码，果然不吃cpu了，成功