docs/ffmpeg.md

335 lines
10 KiB
Markdown
Raw Permalink Normal View History

2024-10-19 11:31:30 +08:00
# 基于Ubuntu源码编译ffmpeg
2024-10-28 17:24:18 +08:00
## 编译SDL2可选
2024-10-19 11:31:30 +08:00
```shell
https://github.com/libsdl-org/SDL/releases/tag/release-2.30.8 # 下载源码
cd sdl/source/folder
./configure --prefix=/usr/local
make
sudo make install
```
## 安装ffmpeg依赖
1. 安装包管理工具 aptitude
aptitude 与 apt-get一样是 Debian 系中功能极其强大的包管理工具。 与 apt-get 不同的是aptitude 在处理依赖问题上更佳一些,当 aptitude 在删除一个包时,会同时删除本身所依赖的包。
```shell
sudo apt install aptitude
```
2. 安装 yasm、nasm
yasm、nasm 是两个汇编器,编译 FFmpeg 需要用到。
安装命令:
```shell
sudo aptitude install yasm nasm
```
3. 其他依赖
FFmpeg 的安装依赖许多库(如,音频编码库,音频解码库,视频编解码库等),这里不介绍每个库的具体作用,按照命令安装即可。
```shell
sudo apt-get install libgmp3-dev
sudo apt install pkg-config
sudo apt install gnutls-bin
sudo aptitude install libaom-dev
sudo aptitude install libass-dev
sudo aptitude install libbluray-dev
sudo aptitude install libfdk-aac-dev
sudo aptitude install libmp3lame-dev
sudo aptitude install libopencore-amrnb-dev
sudo aptitude install libopencore-amrwb-dev
sudo aptitude install libopenmpt-dev
sudo aptitude install libopus-dev
sudo aptitude install libshine-dev
sudo aptitude install libsnappy-dev
sudo aptitude install libsoxr-dev
sudo aptitude install libspeex-dev
sudo aptitude install libtheora-dev
sudo aptitude install libtwolame-dev
sudo aptitude install libvo-amrwbenc-dev
sudo aptitude install llibvpx-dev
sudo aptitude install libwavpack-dev
sudo aptitude install libwebp-dev
sudo aptitude install libx264-dev
sudo aptitude install libx265-dev
sudo aptitude install libxvidcore-dev
sudo aptitude install liblzma-dev
```
## 编译ffmpeg
1. 从官网下载源码
2. 进入根目录编译
```shell
2024-10-28 17:24:18 +08:00
./configure --enable-shared --enable-gpl --enable-sdl2 --enable-libx264 --enable-libx265 --enable-nvdec --enable-nvenc --enable-cuda --enable-cuvid
# 其中 --enable-gpl --enable-libx264 --enable-libx265 必不可少 --enable-sdl2 是ffmpeg播放器需要的
# 由于需要硬件英伟达显卡加速编解码 --enable-nvdec --enable-nvenc --enable-cuda --enable-cuvid
# --enable-shared 使make install时将动态库也正确安装否则运行ffmpeg时会显示缺少动态库
2024-10-19 11:31:30 +08:00
make
sudo make install
2024-10-28 17:24:18 +08:00
# 安装完后,确保安装目录,确保已经配置好动态库搜索的路径,
```
2024-10-19 11:31:30 +08:00
如果需要硬件英伟达显卡加速编解码,需要安装相关依赖
```shell
sudo apt install libdssialsacompat-dev
git clone https://git.videolan.org/git/ffmpeg/nv-codec-headers.git
cd nv-codec-headers
make
sudo make install
2025-01-07 15:09:50 +08:00
# ./ffmpeg -hwaccels 查看支持哪些硬件编码
2024-11-21 16:41:04 +08:00
```
2025-01-15 10:24:25 +08:00
### 编译ffmeg arm64安卓版本
2024-11-21 16:41:04 +08:00
在源码目录下创建buildarm64.sh内容如下
```shell
#!/bin/bash
export NDK=~/android/android-ndk-r20b #这里配置先你的 NDK 路径
TOOLCHAIN=$NDK/toolchains/llvm/prebuilt/linux-x86_64
# 编译到安卓上,尽可能小一些(--enable-small),只是用来解码,也就取消编译封装编码相关的
function build_android
{
./configure \
--disable-programs \
--disable-avdevice \
--disable-encoders \
--disable-muxers \
--disable-filters \
--enable-neon \
--enable-gpl \
--disable-sdl2 \
--disable-postproc \
--disable-debug \
--enable-small \
--enable-static \
--enable-shared \
--disable-doc \
--enable-ffmpeg \
--disable-ffplay \
--disable-ffprobe \
--disable-avdevice \
--disable-doc \
--disable-symver \
--prefix=$PREFIX \
--cross-prefix=$CROSS_PREFIX \
--target-os=android \
--arch=$ARCH \
--cpu=$CPU \
--cc=$CC \
--cxx=$CXX \
--enable-cross-compile \
--sysroot=$SYSROOT \
--extra-cflags="-Os -fpic $OPTIMIZE_CFLAGS" \
--extra-ldflags="$ADDI_LDFLAGS"
make clean
make -j16
make install
echo "============================ build android arm64-v8a success =========================="
}
#arm64-v8a
ARCH=arm64
CPU=armv8-a
API=21
CC=$TOOLCHAIN/bin/aarch64-linux-android$API-clang
CXX=$TOOLCHAIN/bin/aarch64-linux-android$API-clang++
SYSROOT=$NDK/toolchains/llvm/prebuilt/linux-x86_64/sysroot
CROSS_PREFIX=$TOOLCHAIN/bin/aarch64-linux-android-
PREFIX=$(pwd)/android/$CPU
OPTIMIZE_CFLAGS="-march=$CPU"
build_android
```
```shell
chomd +x buildarm64.sh
./buildarm64.sh
# 编译好后,对应的源文件和库文件会安装到./android/armv-v8a中
```
2025-01-15 10:24:25 +08:00
### 在rk3568上交叉编译ffmpeg
使用瑞芯微上使用其硬件加速编码和解码
创建一个目录 ~/dev 之后所有的操作在这个目录下进行
1. 编译mpp
```shell
git clone -b jellyfin-mpp --depth=1 https://github.com/nyanmisaka/mpp.git rkmpp
mkdir rkmpp_build
pushd rkmpp_build
# 在cmake中指定交叉工具链gcc和g++ 改成自己的目录
# 注意在不要安装到本机的系统目录,最好新建一个目录,之后把所有的依赖安装到这个目录
cmake \
-DCMAKE_INSTALL_PREFIX=~/dev/rkffmpeg \
-DCMAKE_BUILD_TYPE=Release \
-DBUILD_SHARED_LIBS=ON \
-DBUILD_TEST=OFF \
-DCMAKE_C_COMPILER=/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/bin/aarch64-linux-gcc \
-DCMAKE_CXX_COMPILER=/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/bin/aarch64-linux-g++ \
..
make -j $(nproc)
make install
```
2. build rga
创建一个在~/dev 目录下创建cross_file.txt, 将工具链换成自己的,以及安装目录
```txt
[host_machine]
system = 'linux'
cpu_family = 'aarch64'
cpu = 'armv8-a'
endian = 'little'
[binaries]
c = '/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/bin/aarch64-linux-gcc'
cpp = '/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/bin/aarch64-linux-g++'
ar = '/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/aarch64-rockchip-linux-gnu/bin/ar'
strip = '/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/aarch64-rockchip-linux-gnu/bin/strip'
[properties]
c_args = ['-fpermissive']
cpp_args = ['-fpermissive']
[paths]
prefix = '~/dev/rkffmpeg'
libdir = 'lib'
```
```shell
git clone -b jellyfin-rga --depth=1 https://github.com/nyanmisaka/rk-mirrors.git rkrga
# 不需要cd 进行
meson setup rkrga rkrga_build \
--prefix=/usr \
--libdir=lib \
--buildtype=release \
--default-library=shared \
-Dcpp_args=-fpermissive \
-Dlibdrm=false \
-cross-file=./cross_file.txt \
-Dlibrga_demo=false
meson configure rkrga_build
ninja -C rkrga_build install
```
3. 编译drm
首先去官网下载源码包https://dri.freedesktop.org/libdrm/
下载完成后开始解压缩解压缩完成后进入源码目录首先创建install和build目录然后创建cross_file.txt文件文件内容如下
```txt
[host_machine]
system = 'linux'
cpu_family = 'aarch64'
cpu = 'armv8-a'
endian = 'little'
[binaries]
c = '/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/bin/aarch64-linux-gcc'
cpp = '/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/bin/aarch64-linux-g++'
ar = '/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/aarch64-rockchip-linux-gnu/bin/ar'
strip = '/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/aarch64-rockchip-linux-gnu/bin/strip'
[properties]
c_args = ['-fpermissive']
cpp_args = ['-fpermissive']
[paths]
prefix = '~/dev/rkffmpeg'
libdir = 'lib'
```
然后直接执行指令ninja && ninja install即可使用
4. 编译ffmpeg
上面的依赖编译并安装好后正式开始编译ffmpeg添加rk芯片的mpp来加速编解码
在~/dev 下载ffmpeg 进入到源码目录 创建build.sh 内容如下
```shell
#!/bin/bash
export TOOLCHAIN=/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/bin/aarch64-linux
export GCC=$TOOLCHAIN-gcc
export GXX=$TOOLCHAIN-g++
# 编译到开发版,尽可能小一些(--enable-small),取消了一些其它硬件加速和封装
# 更换 cc cxx ar strip 以及--extra-cflags --extra-ldflags
function build_android
{
./configure \
--sysroot=./ \
--extra-cflags="-I~/dev/rkffmpeg/include" \
--extra-ldflags="-L~/dev/rkffmpeg/lib" \
--prefix=$PREFIX \
--enable-cross-compile \
--arch=arm64 \
--cpu=armv8-a \
ar=/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/aarch64-rockchip-linux-gnu/bin/ar \
strip=/home/user/baidu/buildroot-cross/gcc_for_rk356x_buildroot/aarch64-rockchip-linux-gnu/bin/strip \
--cc=$GCC \
--cxx=$GXX \
--enable-libdrm \
--disable-nvenc \
--disable-nvdec \
--disable-vdpau \
--disable-vaapi \
--enable-rkmpp \
--enable-rkrga \
--enable-version3 \
--disable-programs \
--disable-avdevice \
--disable-muxers \
--disable-filters \
--enable-neon \
--enable-gpl \
--disable-sdl2 \
--disable-postproc \
--disable-debug \
--enable-small \
--enable-static \
--enable-shared \
--disable-doc \
--enable-ffmpeg \
--disable-ffprobe \
--disable-avdevice \
--disable-doc \
--disable-symver
echo "============================ build android arm64-v8a success =========================="
}
#arm64-v8a
PREFIX=$(pwd)/rk-arm64
build_android
```
编译
```shell
chmod +x build.sh
make clean
make -j16
make install
```
最后把编译好的ffmpeg 放到~/dev/rkffmpeg 中 一并打包移植到开发版上
2024-11-21 16:41:04 +08:00
## 硬编码
```c++
// 与软编码相比,没有很多额外的配置
// 仅仅是在查找编码器时按名称查找 hevc_nvenc ,其余和软编码一样
auto pCodec = avcodec_find_encoder_by_name("hevc_nvenc");
// ...
// 其次就是注意软编码出来的每一帧数据 前面都会带有SEI增强帧关键帧后面才是I帧或者P帧
auto naluHeader = (pkt->data[4] & 0x7E) >> 1; // 前四个字节是0x00 00 00 01 分隔符
if (naluHeader == 39) { // SEI
auto header = getNextSplit(pkt->data, pkt->size); // 往后找跳过这个SEI
if (header == nullptr)
return 0;
naluHeader = ((*header) & 0x7E) >> 1;
// cout << "naluHeader: " << naluHeader << endl;
}
```