ubuntu 18.04 ffmpeg cuvid硬解环境搭建

由于最近要在视频流上Mask R-CNN模型，cuda硬解也要上，又要折腾一下环境搭建，在此记录一下。

这里的硬解说的是调用nvidia的video codec sdk来加速编解码，使用cuvid硬解的方式有两种，一种是调用ffmpeg已经嵌入的硬解方案，另外一种是先调用ffmpeg再调用codec sdk。
从开发难度来说，第一种要稍微简单一些，只需要指定ffmpeg的video codec就可以启用硬解。 下面是我搭建的一些记录

安装依赖

由于我的机器搭建了各种各样的开发环境，这次编译缺少的一个是yasm包，执行下面命令就可以安装。如果有缺少其它包，自行搜索一下应该可以解决。

 sudo apt-get install yasm

安装nv codec header

由于编译ffmpeg需要用到nv codec的一些header，在这里下载，由于我用的ffmpeg版本最新的master分支(commit: d6489d),所以nv_header也要用最新的9.0版本，下载地址是：
https://github.com/FFmpeg/nv-codec-headers/releases

下载之后，执行

sudo make install

就可以完成header的安装。如果要卸载， 可以执行sudo make uninstall命令

编译ffmpeg

从 https://git.ffmpeg.org/ffmpeg.git clone 一下ffmpeg的源码
然后配置一下

./configure --enable-cuda --enable-cuvid --enable-nvenc --enable-nonfree --enable-libnpp --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64
sudo make -j6 
sudo make install

到这里基本就安装成功了，下面验证一下硬解是否生效

验证一下

由于是硬解，所以要验证一下是否生效，我用本地的一个视频文件试了一下, 用time命令对比一下同一个转码任务，哪个时间更短
ffmpeg的命令参数比较多，可以自行搜索看一下：

ffmpeg -y -i jiaofu2.rmvb -t 200 -c:v h264_nvenc output2.mp4

• -i 输入文件
• -c:v 编解码器video codec， h264_nvenc表示使用nvidia的video codec
• -t 目标时长
  如果不指定c:v, 就是用默认CPU软解
  使用硬解：

time ffmpeg -y -i jiaofu2.rmvb -t 200 -c:v h264_nvenc output2.mp4
ffmpeg version N-93679-gd6489ddb7a Copyright (c) 2000-2019 the FFmpeg developers
  built with gcc 7 (Ubuntu 7.3.0-27ubuntu1~18.04)
  configuration: --enable-cuda --enable-cuvid --enable-nvenc --enable-nonfree --enable-libnpp --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64
  libavutil      56. 26.100 / 56. 26.100
  libavcodec     58. 52.100 / 58. 52.100
  libavformat    58. 27.103 / 58. 27.103
  libavdevice    58.  7.100 / 58.  7.100
  libavfilter     7. 50.100 /  7. 50.100
  libswscale      5.  4.100 /  5.  4.100
  libswresample   3.  4.100 /  3.  4.100
[rm @ 0x555690d38040] Invalid stream index 2 for index at pos 2438339235
Input #0, rm, from 'jiaofu2.rmvb':
  Metadata:
    title           : �̸�2.The.Godfather.Part.II.1974.720p.Bluray
    author          : ��Ӱ�ӵ۹�CNXP����С������ס�ഺ��
    copyright       : bbs.cnxp.com(C)2010
    ASMRuleBook     : #($Bandwidth >= 0),Stream0Bandwidth = 96468, Stream1Bandwidth = 1503532;
    Audiences       : Easy RealMedia Tool's Audience;
    audioMode       : music
    Creation Date   : 7/22/2010 17:52:32
    Description     : This File is Created by Easy RealMedia Tools@!
    Email           : [email protected]
    Generated By    : Easy RealMedia Tools V1.8x
    HomeWeb         : http://redcheek.net
    Keywords        : Ӱ�ӵ۹���Ʒ
    Modification Date: 7/22/2010 17:52:32
    videoMode       : sharp
  Duration: 03:22:08.25, start: 0.000000, bitrate: 1608 kb/s
    Stream #0:0: Audio: cook (cook / 0x6B6F6F63), 44100 Hz, stereo, fltp, 96 kb/s
    Stream #0:1: Video: rv40 (RV40 / 0x30345652), yuv420p, 1280x720, 1499 kb/s, 23.98 fps, 23.98 tbr, 1k tbn, 1k tbc
Stream mapping:
  Stream #0:1 -> #0:0 (rv40 (native) -> h264 (h264_nvenc))
  Stream #0:0 -> #0:1 (cook (native) -> aac (native))
Press [q] to stop, [?] for help
Output #0, mp4, to 'output2.mp4':
  Metadata:
    title           : �̸�2.The.Godfather.Part.II.1974.720p.Bluray
    author          : ��Ӱ�ӵ۹�CNXP����С������ס�ഺ��
    copyright       : bbs.cnxp.com(C)2010
    ASMRuleBook     : #($Bandwidth >= 0),Stream0Bandwidth = 96468, Stream1Bandwidth = 1503532;
    Audiences       : Easy RealMedia Tool's Audience;
    audioMode       : music
    Creation Date   : 7/22/2010 17:52:32
    Description     : This File is Created by Easy RealMedia Tools@!
    Email           : [email protected]
    Generated By    : Easy RealMedia Tools V1.8x
    HomeWeb         : http://redcheek.net
    Keywords        : Ӱ�ӵ۹���Ʒ
    Modification Date: 7/22/2010 17:52:32
    videoMode       : sharp
    encoder         : Lavf58.27.103
    Stream #0:0: Video: h264 (h264_nvenc) (Main) (avc1 / 0x31637661), yuv420p(progressive), 1280x720, q=-1--1, 2000 kb/s, 23.98 fps, 785645.00 tbn, 23.98 tbc
    Metadata:
      encoder         : Lavc58.52.100 h264_nvenc
    Side data:
      cpb: bitrate max/min/avg: 0/0/2000000 buffer size: 4000000 vbv_delay: -1
    Stream #0:1: Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 128 kb/s
    Metadata:
      encoder         : Lavc58.52.100 aac
frame=  200 fps=0.0 q=15.0 size=    1280kB time=00:00:08.21 bitrate=1276.2kbits/frame=  426 fps=426 q=29.0 size=    5376kB time=00:00:17.64 bitrate=2496.3kbits/frame=  723 fps=481 q=15.0 size=    6912kB time=00:00:30.03 bitrate=1885.6kbits/frame= 1098 fps=548 q=16.0 size=    8192kB time=00:00:46.37 bitrate=1447.0kbits/frame= 1406 fps=562 q=15.0 size=   13312kB time=00:00:58.51 bitrate=1863.6kbits/frame= 1760 fps=586 q=17.0 size=   17152kB time=00:01:13.28 bitrate=1917.4kbits/frame= 2106 fps=601 q=15.0 size=   20736kB time=00:01:27.71 bitrate=1936.7kbits/frame= 2385 fps=596 q=18.0 size=   25856kB time=00:01:39.34 bitrate=2132.0kbits/frame= 2685 fps=596 q=22.0 size=   29696kB time=00:01:51.86 bitrate=2174.7kbits/frame= 2992 fps=598 q=20.0 size=   33536kB time=00:02:04.66 bitrate=2203.7kbits/frame= 3284 fps=597 q=19.0 size=   37120kB time=00:02:17.39 bitrate=2213.2kbits/frame= 3588 fps=598 q=23.0 size=   40704kB time=00:02:30.40 bitrate=2217.0kbits/frame= 3900 fps=600 q=22.0 size=   43520kB time=00:02:43.03 bitrate=2186.7kbits/frame= 4223 fps=603 q=19.0 size=   46848kB time=00:02:56.40 bitrate=2175.5kbits/frame= 4546 fps=606 q=22.0 size=   50688kB time=00:03:09.48 bitrate=2191.4kbits/frame= 4796 fps=606 q=21.0 Lsize=   53427kB time=00:03:20.00 bitrate=2188.3kbits/s dup=375 drop=0 speed=25.3x    
video:50217kB audio:3055kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.290995%
[aac @ 0x555690d6f4c0] Qavg: 10399.560

real	0m8.017s
user	0m36.849s
sys	0m0.702s

使用软解：

 time ffmpeg -y -i jiaofu2.rmvb -t 200  output2.mp4
ffmpeg version N-93679-gd6489ddb7a Copyright (c) 2000-2019 the FFmpeg developers
  built with gcc 7 (Ubuntu 7.3.0-27ubuntu1~18.04)
  configuration: --enable-cuda --enable-cuvid --enable-nvenc --enable-nonfree --enable-libnpp --extra-cflags=-I/usr/local/cuda/include --extra-ldflags=-L/usr/local/cuda/lib64
  libavutil      56. 26.100 / 56. 26.100
  libavcodec     58. 52.100 / 58. 52.100
  libavformat    58. 27.103 / 58. 27.103
  libavdevice    58.  7.100 / 58.  7.100
  libavfilter     7. 50.100 /  7. 50.100
  libswscale      5.  4.100 /  5.  4.100
  libswresample   3.  4.100 /  3.  4.100
[rm @ 0x5556cfc4b000] Invalid stream index 2 for index at pos 2438339235
Input #0, rm, from 'jiaofu2.rmvb':
  Metadata:
    title           : �̸�2.The.Godfather.Part.II.1974.720p.Bluray
    author          : ��Ӱ�ӵ۹�CNXP����С������ס�ഺ��
    copyright       : bbs.cnxp.com(C)2010
    ASMRuleBook     : #($Bandwidth >= 0),Stream0Bandwidth = 96468, Stream1Bandwidth = 1503532;
    Audiences       : Easy RealMedia Tool's Audience;
    audioMode       : music
    Creation Date   : 7/22/2010 17:52:32
    Description     : This File is Created by Easy RealMedia Tools@!
    Email           : [email protected]
    Generated By    : Easy RealMedia Tools V1.8x
    HomeWeb         : http://redcheek.net
    Keywords        : Ӱ�ӵ۹���Ʒ
    Modification Date: 7/22/2010 17:52:32
    videoMode       : sharp
  Duration: 03:22:08.25, start: 0.000000, bitrate: 1608 kb/s
    Stream #0:0: Audio: cook (cook / 0x6B6F6F63), 44100 Hz, stereo, fltp, 96 kb/s
    Stream #0:1: Video: rv40 (RV40 / 0x30345652), yuv420p, 1280x720, 1499 kb/s, 23.98 fps, 23.98 tbr, 1k tbn, 1k tbc
Stream mapping:
  Stream #0:1 -> #0:0 (rv40 (native) -> mpeg4 (native))
  Stream #0:0 -> #0:1 (cook (native) -> aac (native))
Press [q] to stop, [?] for help
Output #0, mp4, to 'output2.mp4':
  Metadata:
    title           : �̸�2.The.Godfather.Part.II.1974.720p.Bluray
    author          : ��Ӱ�ӵ۹�CNXP����С������ס�ഺ��
    copyright       : bbs.cnxp.com(C)2010
    ASMRuleBook     : #($Bandwidth >= 0),Stream0Bandwidth = 96468, Stream1Bandwidth = 1503532;
    Audiences       : Easy RealMedia Tool's Audience;
    audioMode       : music
    Creation Date   : 7/22/2010 17:52:32
    Description     : This File is Created by Easy RealMedia Tools@!
    Email           : [email protected]
    Generated By    : Easy RealMedia Tools V1.8x
    HomeWeb         : http://redcheek.net
    Keywords        : Ӱ�ӵ۹���Ʒ
    Modification Date: 7/22/2010 17:52:32
    videoMode       : sharp
    encoder         : Lavf58.27.103
    Stream #0:0: Video: mpeg4 (mp4v / 0x7634706D), yuv420p(progressive), 1280x720, q=2-31, 200 kb/s, 23.98 fps, 10981 tbn, 23.98 tbc
    Metadata:
      encoder         : Lavc58.52.100 mpeg4
    Side data:
      cpb: bitrate max/min/avg: 0/0/200000 buffer size: 0 vbv_delay: -1
    Stream #0:1: Audio: aac (LC) (mp4a / 0x6134706D), 44100 Hz, stereo, fltp, 128 kb/s
    Metadata:
      encoder         : Lavc58.52.100 aac
frame=  204 fps=0.0 q=31.0 size=     512kB time=00:00:08.46 bitrate= 495.4kbits/frame=  379 fps=379 q=31.0 size=    1536kB time=00:00:15.76 bitrate= 798.1kbits/frame=  554 fps=369 q=31.0 size=    2560kB time=00:00:23.55 bitrate= 890.3kbits/frame=  801 fps=400 q=31.0 size=    3072kB time=00:00:33.37 bitrate= 754.1kbits/frame= 1069 fps=427 q=24.8 size=    3584kB time=00:00:44.54 bitrate= 659.1kbits/frame= 1291 fps=430 q=31.0 size=    4352kB time=00:00:53.80 bitrate= 662.6kbits/frame= 1517 fps=433 q=31.0 size=    4864kB time=00:01:03.23 bitrate= 630.2kbits/frame= 1763 fps=440 q=31.0 size=    5376kB time=00:01:13.49 bitrate= 599.3kbits/frame= 1984 fps=440 q=31.0 size=    6144kB time=00:01:22.70 bitrate= 608.6kbits/frame= 2208 fps=441 q=31.0 size=    6912kB time=00:01:32.37 bitrate= 612.9kbits/frame= 2430 fps=441 q=31.0 size=    7680kB time=00:01:41.30 bitrate= 621.0kbits/frame= 2652 fps=441 q=31.0 size=    8448kB time=00:01:50.56 bitrate= 625.9kbits/frame= 2877 fps=442 q=31.0 size=    9216kB time=00:01:59.95 bitrate= 629.4kbits/frame= 3099 fps=442 q=24.8 size=   10240kB time=00:02:09.66 bitrate= 646.9kbits/frame= 3321 fps=442 q=24.8 size=   11264kB time=00:02:19.02 bitrate= 663.7kbits/frame= 3538 fps=442 q=31.0 size=   12032kB time=00:02:27.52 bitrate= 668.1kbits/frame= 3733 fps=439 q=31.0 size=   13312kB time=00:02:35.97 bitrate= 699.2kbits/frame= 3977 fps=441 q=31.0 size=   14080kB time=00:02:45.83 bitrate= 695.5kbits/frame= 4202 fps=442 q=31.0 size=   14848kB time=00:02:55.21 bitrate= 694.2kbits/frame= 4409 fps=440 q=31.0 size=   15616kB time=00:03:03.85 bitrate= 695.8kbits/frame= 4618 fps=439 q=31.0 size=   16384kB time=00:03:13.12 bitrate= 695.0kbits/frame= 4796 fps=439 q=31.0 Lsize=   17243kB time=00:03:20.00 bitrate= 706.3kbits/s dup=375 drop=0 speed=18.3x    
video:14031kB audio:3055kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.916266%
[aac @ 0x5556cfc82100] Qavg: 10399.560

real	0m10.985s
user	1m1.552s
sys	0m1.078s

小结

从两条命令的执行结果来看，硬解已经发挥作用了。
上面测试结果跟硬件相关，不同时间测到的结果可能不一样，只是作为参考。
硬解速度是25.3x，软解速度是18.3x
硬解耗时8.01 s，软解速度是10.98 s
从耗时和加速来看，硬解比软解稍微好一些
由于测试视频是720P的普通画质，而且只是一路视频，如果处理4K或者8K的视频，而且多路的话，硬解优势可能会更明显。