update dockerfile, readme

Author: AkashiSN

.github/workflows/build.yml

@@ -85,15 +85,15 @@ jobs:
         id: cuda_ubuntu_pull
         run: |
           docker pull nvidia/cuda:11.1-devel-ubuntu20.04
-          docker pull nvidia/cuda:11.1-runtime-ubuntu20.04
-      - name: run if cuda_ubuntu_pull failed
+          docker pull ubuntu:20.04
+        - name: run if cuda_ubuntu_pull failed
         if: failure() && steps.cuda_ubuntu_pull.outcome == 'failure'
         run: |
           docker pull nvidia/cuda:11.1-devel-ubuntu20.04
-          docker pull nvidia/cuda:11.1-runtime-ubuntu20.04
-      - name: build ffmpeg
+          docker pull ubuntu:20.04
+        - name: build ffmpeg
         run: |
-          docker build -t ffmpeg:cuda -f cuda-ubuntu.dockerfile .
+          docker build -t ffmpeg:cuda -f cuda.dockerfile .
       - name: test run ffmepg
         run: |
           docker run --rm ffmpeg:cuda -buildconf

README.md

@@ -32,6 +32,8 @@ because I don't have the resources and the time to maintain other systems.
 * `theora`: Free lossy video compression format
 * `opus`: Lossy audio coding format
 * `srt`: Secure Reliable Transport
+
+### HardwareAccel
 * `nv-codec`: [NVIDIA's GPU accelerated video codecs](https://devblogs.nvidia.com/nvidia-ffmpeg-transcoding-guide/). Installation is triggered only if CUDA installation is detected, follow [these](#Cuda-installation) instructions for installation. Supported codecs in nvcodec:
     * Decoders
         * H264 `h264_cuvid`
@@ -46,30 +48,43 @@ because I don't have the resources and the time to maintain other systems.
     * Encoders
         * H264 `nvenc_h264`
         * H265 `nvenc_hevc`
+* `amf`: [Advanced Media Framework SDK](https://gpuopen.com/advanced-media-framework/). Optimal access to AMD GPUs for multimedia processing. Supported codecs in amf:
+    * Encoders
+        * H264 `h264_amf`
+        * H265 `hevc_amf`
+* `vaapi`: [Video Acceleration API](https://trac.ffmpeg.org/wiki/Hardware/VAAPI). Supported codecs in vaapi:
+    * Encoders
+        * H264 `h264_vaapi`
+        * H265 `hevc_vaapi`
+        * Motion JPEG `mjpeg_vaapi`
+        * MPEG2 video `mpeg2_vaapi`
+        * VP8 `vp8_vaapi`
+        * VP9 `vp9_vaapi`
+
 
 ## Continuos Integration
 ffmpeg-build-script is rockstable. Every commit runs against Linux and MacOS with https://github.com/markus-perl/ffmpeg-build-script/actions just to make sure everything works as expected.
 
-Requirements MacOS
-------------
+
+## Requirements
+### MacOS
 
 * XCode 10.x or greater
 
-Requirements Linux
-------------
+### Linux
+
 * Debian >= Buster, Ubuntu => Focal Fossa, other Distributions might work too
-* build-essentials, curl, and Python3 is required installed:
+* build-essentials, curl, and Python3 is required installed
 
-```
+```bash
 # Debian and Ubuntu
 $ sudo apt-get install build-essential curl python3
 
 # Fedora
 $ sudo dnf install @development-tools curl python3
 ```
 
-Installation
-------------
+## Installation
 
 ### Quick install and run
 
@@ -78,35 +93,90 @@ Open your command line and run (needs curl to be installed):
 ```bash
 $ bash <(curl -s "https://raw.githubusercontent.com/markus-perl/ffmpeg-build-script/master/web-install.sh?v1")
 ```
+
 This command downloads the build script and automatically starts the build process.
 
-### Run with Docker
+### Common installation
 
-#### Default - Without CUDA
 ```bash
 $ git clone https://github.com/markus-perl/ffmpeg-build-script.git
 $ cd ffmpeg-build-script
+$ ./build-ffmpeg --help
+```
+
+### Build in Docker
+
+The main advantage of using Docker is the ability to reliably build without polluting the host environment. And you don't even have to install the CUDA SDK on your host!
+
+If you are running below kind of operating system and having Docker version 19.03 or higher, this is your best option.
+* Ubuntu >= 16.04 (16.04, 18.04, 20.04)
+* Centos >= 7 (7, 8)
+
+1. Enable Docker BuildKit
+```bash
+$ export DOCKER_BUILDKIT=1
+```
+
+2. Set the following DIST (`ubuntu` or `centos`) and VER (ubuntu: `16.04` , `18.04`, `20.04` or centos: `7`, `8`) environment variables in conjunction with your operating system.
+```bash
+$ export DIST=centos
+$ export VER=8
+```
+
+3. Start the docker build as follows.
+```bash
+$ sudo -E docker build --tag=ffmpeg:cuda -f cuda.dockerfile --build-arg DIST=$DIST VER=$VER .
+```
+
+4. Build an `export.dockerfile` that copies only what you need from the image you just built as follows. When running, move the library in the lib to a location where the linker can find it or set the `LD_LIBRARY_PATH`.
+Since we have matched the operating system and version, it should work well with dynamic links. If it doesn't work, edit the `export.dockerfile` and copy the necessary libraries and try again.
+
+```bash
+$ sudo -E docker build --output type=local,dest=build -f export.dockerfile .
+$ ls build
+bin lib
+$ ls build/bin
+ffmpeg ffprobe
+$ ls build/lib
+libnppc.so.11 libnppicc.so.11 libnppidei.so.11 libnppig.so.11
+```
+
+### Build in Docker (full static ver.)
+If you're running an operating system other than the one above, a completely static build may work.
+It's easy to do, just run the following command.
+```bash
+$ sudo -E docker build --tag=ffmpeg:cuda-static --output type=local,dest=build -f full-static.dockerfile .
+```
+
+### Run with Docker
+You can also run the entire Docker if the above two fail.
+
+#### Default - Without CUDA
+If you don't use CUDA, it's simple and runs as follows.
+
+```bash
 $ sudo docker build --tag=ffmpeg .
-$ docker run ffmpeg -i https://files.coconut.co.s3.amazonaws.com/test.mp4 -f webm -c:v libvpx -c:a libvorbis - > /tmp/test.mp4
+$ sudo docker run ffmpeg -i https://files.coconut.co.s3.amazonaws.com/test.mp4 -f webm -c:v libvpx -c:a libvorbis - > test.mp4
 ```
 
 #### With CUDA
+If you use CUDA, Docker must be higher than 19.03.
+Install the driver and `nvidia-docker2` from [here](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#installing-docker-ce).
+You can perform hardware acceleration by GPU by running the following.
 ```bash
-$ git clone https://github.com/markus-perl/ffmpeg-build-script.git
-$ cd ffmpeg-build-script
-$ sudo docker build --tag=ffmpeg-cuda -f cuda-ubuntu.dockerfile .
-$ sudo docker run --gpus all ffmpeg-cuda -hwaccel cuvid -c:v h264_cuvid -i https://files.coconut.co.s3.amazonaws.com/test.mp4 -c:v hevc_nvenc -vf scale_npp=-1:1080 - > /tmp/test.mp4
+$ sudo docker build --tag=ffmpeg:cuda -f cuda-ubuntu.dockerfile .
+$ sudo docker run --gpus all ffmpeg-cuda -hwaccel cuvid -c:v h264_cuvid -i https://files.coconut.co.s3.amazonaws.com/test.mp4 -c:v hevc_nvenc -vf scale_npp=-1:1080 - > test.mp4
 ```
 
-### Common installation
+### Common build
+
+If you want to enable CUDA, please refer to [these](#Cuda-installation) and install the SDK.
 
 ```bash
-$ git clone https://github.com/markus-perl/ffmpeg-build-script.git
-$ cd ffmpeg-build-script
-$ ./build-ffmpeg --help
+$ ./build-ffmpeg --build
 ```
 
-### Cuda installation
+## Cuda installation
 
 CUDA is a parallel computing platform developed by NVIDIA.
 To be able to compile ffmpeg with CUDA support, you first need a compatible NVIDIA GPU.

cuda-ubuntu.dockerfile renamed to cuda.dockerfile

@@ -1,4 +1,7 @@
-FROM nvidia/cuda:11.1-devel-ubuntu20.04 AS build
+ARG DIST=ubuntu
+ARG VER=20.04
+
+FROM nvidia/cuda:11.1-devel-${DIST}${VER} AS build
 
 ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility,video
 
@@ -12,10 +15,16 @@ COPY ./build-ffmpeg /app/build-ffmpeg
 
 RUN SKIPINSTALL=yes /app/build-ffmpeg --build
 
-FROM nvidia/cuda:11.1-runtime-ubuntu20.04
 
-ENV NVIDIA_DRIVER_CAPABILITIES=compute,utility,video
+FROM ${DIST}:${VER}
+
+# Copy libnpp
+COPY --from=build /usr/local/cuda-11.1/targets/x86_64-linux/lib/libnppc.so.11 /lib/x86_64-linux-gnu/libnppc.so.11
+COPY --from=build /usr/local/cuda-11.1/targets/x86_64-linux/lib/libnppig.so.11 /lib/x86_64-linux-gnu/libnppig.so.11
+COPY --from=build /usr/local/cuda-11.1/targets/x86_64-linux/lib/libnppicc.so.11 /lib/x86_64-linux-gnu/libnppicc.so.11
+COPY --from=build /usr/local/cuda-11.1/targets/x86_64-linux/lib/libnppidei.so.11 /lib/x86_64-linux-gnu/libnppidei.so.11
 
+# Copy ffmpeg
 COPY --from=build /app/workspace/bin/ffmpeg /usr/bin/ffmpeg
 COPY --from=build /app/workspace/bin/ffprobe /usr/bin/ffprobe
 

export.dockerfile

@@ -0,0 +1,11 @@
+FROM scratch
+
+# Copy libnpp
+COPY --from=ffmpeg:cuda /lib/x86_64-linux-gnu/libnppc.so.11 /lib/libnppc.so.11
+COPY --from=ffmpeg:cuda /lib/x86_64-linux-gnu/libnppig.so.11 /lib/libnppig.so.11
+COPY --from=ffmpeg:cuda /lib/x86_64-linux-gnu/libnppicc.so.11 /lib/libnppicc.so.11
+COPY --from=ffmpeg:cuda /lib/x86_64-linux-gnu/libnppidei.so.11 /lib/libnppidei.so.11
+
+# Copy ffmpeg
+COPY --from=ffmpeg:cuda /usr/bin/ffmpeg /bin/ffmpeg
+COPY --from=ffmpeg:cuda /usr/bin/ffprobe /bin/ffprobe

full-static.dockerfile

@@ -11,13 +11,14 @@ WORKDIR /app
 COPY ./build-ffmpeg /app/build-ffmpeg
 
 RUN AUTOINSTALL=yes /app/build-ffmpeg --build --full-static
+
 RUN ldd /app/workspace/bin/ffmpeg ; exit 0
 RUN ldd /app/workspace/bin/ffprobe ; exit 0
 
 FROM scratch
 
-COPY --from=build /app/workspace/bin/ffmpeg /usr/bin/ffmpeg
-COPY --from=build /app/workspace/bin/ffprobe /usr/bin/ffprobe
+COPY --from=build /app/workspace/bin/ffmpeg /ffmpeg
+COPY --from=build /app/workspace/bin/ffprobe /ffprobe
 
 CMD         ["--help"]
-ENTRYPOINT  ["/usr/bin/ffmpeg"]
+ENTRYPOINT  ["/ffmpeg"]