Skip to content

Nvidia and other fixes #200

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 6 commits into from
Jun 27, 2023
Merged

Nvidia and other fixes #200

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
7d0268c
Fixes to build nvidia filters and decoders
jtbr Jun 13, 2023
0d114a0
Make small (rather than fast) build an option
jtbr Jun 13, 2023
c700436
Consolidate ffmpeg install code, fix install prompt,
jtbr Jun 13, 2023
be86dc9
Update manpage db after adding
jtbr Jun 13, 2023
e929e88
Consolidates install further
jtbr Jun 15, 2023
e1a73d9
Reorganize readme for readability and add detail for clarity.
jtbr Jun 21, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
214 changes: 123 additions & 91 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,14 +22,33 @@ By downloading and using this script, you are fully aware of this.
Use this script at your own risk. I maintain this script in my spare time. Please do not file bug reports for systems
other than Debian and macOS, because I don't have the resources or time to maintain different systems.

## Installation
## Requirements to build

### Quick install and run (macOS, Linux)
### macOS

Open your command line and run (curl needs to be installed):
* XCode 10.x or greater

### Linux

* Debian >= Buster, Ubuntu => Focal Fossa; other distributions might work too
* Rocky Linux 8

```bash
# Debian and Ubuntu
$ sudo apt install build-essential curl

# Fedora
$ sudo dnf install @development-tools curl
```

## Quick Script Installation

Note: to enable hardware acceleration, see details below.
### Quick install and build (macOS, Linux)

Open your command line and run (curl needs to be installed):

```bash
# Without GPL and non-free codes, see https://ffmpeg.org/legal.html
$ bash <(curl -s "https://raw.githubusercontent.com/markus-perl/ffmpeg-build-script/master/web-install.sh?v1")

Expand All @@ -39,12 +58,15 @@ $ bash <(curl -s "https://raw.githubusercontent.com/markus-perl/ffmpeg-build-scr

This command downloads the build script and automatically starts the build process.

### Common installation (macOS, Linux)
### Common install and build (macOS, Linux)

```bash
$ git clone https://github.com/markus-perl/ffmpeg-build-script.git
$ cd ffmpeg-build-script
# Without GPL and non-free codecs
$ ./build-ffmpeg --build
# With GPL and non-free codecs
$ ./build-ffmpeg --enable-gpl-and-non-free --build
```

## Supported Codecs
Expand All @@ -54,7 +76,7 @@ $ ./build-ffmpeg --build
* `libsvtav1`: SVT-AV1 Encoder and Decoder
* `aom`: AV1 Video Codec (Experimental and very slow!)
* `librav1e`: rust based AV1 encoder (only available if [`cargo` is installed](https://doc.rust-lang.org/cargo/getting-started/installation.html))
* `libdav1d`: Fastest AV1 decoder developed by the VideoLAN and FFmpeg communities and sponsored by the AOMedia (only available if `meson` and `ninja` are installed)
* `libdav1d`: Fastest AV1 decoder developed by the VideoLAN and FFmpeg communities and sponsored by the AOMedia (only available if `meson` and `ninja` are available)
* `fdk_aac`: Fraunhofer FDK AAC Codec
* `xvidcore`: MPEG-4 video coding standard
* `VP8/VP9/webm`: VP8 / VP9 Video Codec for the WebM video file format
Expand Down Expand Up @@ -100,45 +122,113 @@ $ ./build-ffmpeg --build
* H264 `h264_amf`


### Apple M1 (Apple Silicon) Support
## Build Script Usage

The script also builds FFmpeg on a new MacBook with an Apple Silicon M1 processor.
```bash
Usage: build-ffmpeg [OPTIONS]
Options:
-h, --help Display usage information
--version Display version information
-b, --build Starts the build process
--enable-gpl-and-non-free Enable non-free codecs - https://ffmpeg.org/legal.html
--latest Build latest version of dependencies if newer available
-c, --cleanup Remove all working dirs
--small Prioritize small size over speed and usability; don't build manpages.
--full-static Complete static build of ffmpeg (eg. glibc, pthreads etc...) **only Linux**
Note: Because of the NSS (Name Service Switch), glibc does not recommend static links.
```

### LV2 Plugin Support
### Notes on static linking

If Python is available, the script will build a ffmpeg binary with lv2 plugin support.
- Because of the NSS (Name Service Switch), glibc does **not recommend** static links. See [more details here](https://sourceware.org/glibc/wiki/FAQ#Even_statically_linked_programs_need_some_shared_libraries_which_is_not_acceptable_for_me.__What_can_I_do.3F).

## Continuous Integration
- The libnpp in the CUDA SDK cannot be statically linked.
- Vaapi cannot be statically linked.

ffmpeg-build-script is very stable. Every commit runs against Linux and macOS
with https://github.com/markus-perl/ffmpeg-build-script/actions to make sure everything works as expected.

## Requirements
## Common full build (macOS, Linux)

### macOS
1) Install the [prerequisites](#requirements-to-build), above.
1) Install optional dependencies, as desired.
- If you have an NVIDIA GPU and want to enable CUDA acceleration, please refer to [these instructions](#Cuda-installation) to install the SDK.

* XCode 10.x or greater
- If you have an AMD GPU and want to enable AMF acceleration, please refer to [these instructions](#amf-installation) to install the drivers.

### Linux
- If you want to enable Vaapi acceleration (for most GPUs), please refer to [these instructions](#Vaapi-installation) to install the driver.

* Debian >= Buster, Ubuntu => Focal Fossa, other Distributions might work too
* Rocky Linux 8
* A development environment and curl is required
- If you want the `librav1e` AV1 encoder, please install [rust](https://doc.rust-lang.org/cargo/getting-started/installation.html) to get `cargo` for the build process. Be sure to start a new shell before building, so that `cargo` is in the new path. If desired, it can be removed with `rustup self uninstall` after the build is complete.

- If you want the `dav1d` AV1 decoder, please ensure that `python3` is installed. If using the system python installation, also ensure that `meson` and `ninja` are installed before running (otherwise the script will try to install them using `pip3`).

- If you want the `Lv2` filter plugin, please ensure that `python3` is installed.

1) Run the downoaded build script from the current directory, with your desired [options](#build-script-usage).
```bash
$ ./build-ffmpeg [your parameters here] --build
```
- Packages will be under the `packages/` subdirectory.
- Build results will be under the `workspace/` subdirectory.

Upon completion, build-ffmpeg will prompt you for whether to install.

1) Once installed, if you are satisfied with your ffmpeg build, the ffmpeg-build-script directory can be removed completely.

### 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 and the NVIDIA compiler nvcc from the CUDA toolkit.

- Ubuntu: To install the CUDA toolkit on Ubuntu, run
```bash
sudo apt install nvidia-cuda-toolkit
```
After compilation, you can run
```bash
sudo apt install nvidia-cuda-dev && sudo apt remove nvidia-cuda-toolkit
```
This removes the compilers but leaves the needed shared library `libnpp`.

- Other Linux distributions: Once you have the GPU and display driver installed, you can follow the
[official instructions](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html)
or [this blog](https://www.pugetsystems.com/labs/hpc/How-To-Install-CUDA-10-1-on-Ubuntu-19-04-1405/)
to setup the CUDA toolkit.

### Vaapi installation

You will need the libva driver, so please install it below.

```bash
# Debian and Ubuntu
$ sudo apt install build-essential curl
$ sudo apt install libva-dev vainfo

# Fedora
$ sudo dnf install @development-tools curl
# Fedora and CentOS
$ sudo dnf install libva-devel libva-intel-driver libva-utils
```

### Build in Docker (Linux)
### AMF installation

To use the AMF encoder, you will need to be using the AMD GPU Pro drivers with OpenCL support.
Download the drivers from https://www.amd.com/en/support and install the appropriate opencl versions.

```bash
./amdgpu-pro-install -y --opencl=rocr,legacy
```

### LV2 Plugin Support

If python 3 is available, the script will build a ffmpeg binary with [Lv2 filter](https://github.com/lv2/lv2/wiki) plugin support.

### Apple M1 (Apple Silicon) Support

The script also builds FFmpeg on a new MacBook with an Apple Silicon M1 processor.


## Build in Docker (Linux)

With Docker, FFmpeg can be built reliably without altering the host system. Also, there is no need to have the CUDA SDK
installed outside of the Docker image.

##### Default
#### Default

If you're running an operating system other than the one above, a completely static build may work. To build a full
statically linked binary inside Docker, just run the following command:
Expand All @@ -147,10 +237,10 @@ statically linked binary inside Docker, just run the following command:
$ docker build --tag=ffmpeg:default --output type=local,dest=build -f Dockerfile .
```

##### CUDA
#### CUDA
These builds are always built with the --enable-gpl-and-non-free switch, as CUDA is non-free. See https://ffmpeg.org/legal.html
```bash

```bash
## Start the build
$ docker build --tag=ffmpeg:cuda --output type=local,dest=build -f cuda-ubuntu.dockerfile .
```
Expand All @@ -170,7 +260,7 @@ $ ls build/lib
libnppc.so.11 libnppicc.so.11 libnppidei.so.11 libnppig.so.11
```

##### Full static version
#### Full static version

If you're running an operating system other than the one above, a completely static build may work. To build a full
statically linked binary inside Docker, just run the following command:
Expand Down Expand Up @@ -204,70 +294,6 @@ $ 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 build (macOS, Linux)

If you want to enable CUDA, please refer to [these](#Cuda-installation) and install the SDK.

If you want to enable Vaapi, please refer to [these](#Vaapi-installation) and install the driver.

```bash
$ ./build-ffmpeg --build
```

## 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.

- Ubuntu: To install the CUDA toolkit on Ubuntu, run "sudo apt install nvidia-cuda-toolkit"
- Other Linux distributions: Once you have the GPU and display driver installed, you can follow the
[official instructions](https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html)
or [this blog](https://www.pugetsystems.com/labs/hpc/How-To-Install-CUDA-10-1-on-Ubuntu-19-04-1405/)
to setup the CUDA toolkit.

## Vaapi installation

You will need the libva driver, so please install it below.

```bash
# Debian and Ubuntu
$ sudo apt install libva-dev vainfo

# Fedora and CentOS
$ sudo dnf install libva-devel libva-intel-driver libva-utils
```

## AMF installation

To use the AMF encoder, you will need to be using the AMD GPU Pro drivers with OpenCL support.
Download the drivers from https://www.amd.com/en/support and install the appropriate opencl versions.

```bash
./amdgpu-pro-install -y --opencl=rocr,legacy
```

## Usage

```bash
Usage: build-ffmpeg [OPTIONS]
Options:
-h, --help Display usage information
--version Display version information
-b, --build Starts the build process
--enable-gpl-and-non-free Enable non-free codecs - https://ffmpeg.org/legal.html
--latest Build latest version of dependencies if newer available
-c, --cleanup Remove all working dirs
--full-static Complete static build of ffmpeg (eg. glibc, pthreads etc...) **only Linux**
Note: Because of the NSS (Name Service Switch), glibc does not recommend static links.
```

## Notes of static link

- Because of the NSS (Name Service Switch), glibc does **not recommend** static links. See detail
below: https://sourceware.org/glibc/wiki/FAQ#Even_statically_linked_programs_need_some_shared_libraries_which_is_not_acceptable_for_me.__What_can_I_do.3F

- The libnpp in the CUDA SDK cannot be statically linked.
- Vaapi cannot be statically linked.

Contact
-------
Expand All @@ -281,6 +307,12 @@ Tested on
* Debian 10
* Ubuntu 20.04

## Continuous Integration

ffmpeg-build-script is very stable. Every commit runs against Linux and macOS
with https://github.com/markus-perl/ffmpeg-build-script/actions to make sure everything works as expected.


Example
-------

Expand Down
Loading