refactor(hwcodec): 精简FFmpeg编译配置并移除解码器

- 优化FFmpeg编译选项,禁用不需要的库(avformat/swscale/swresample/avfilter等)
- 禁用所有解码器和大部分编码器,只保留实际使用的H264/H265/VP8/VP9编码器
- 移除hwcodec解码器模块,MJPEG解码改用libyuv实现
- 移除MJPEG编码器支持
- x86_64添加libmfx支持QSV编码器
- 修复H265 RKMPP编码器支持YUYV直接输入
This commit is contained in:
mofeng-git
2026-01-02 12:31:11 +08:00
parent 04e62d1e3f
commit be4de59f3b
13 changed files with 258 additions and 794 deletions

View File

@@ -107,7 +107,7 @@ RUN mkdir -p /tmp/ffmpeg-build && cd /tmp/ffmpeg-build \
&& echo 'export PKG_CONFIG_SYSROOT_DIR=""' >> /tmp/aarch64-pkg-config \
&& echo 'exec pkg-config "$@"' >> /tmp/aarch64-pkg-config \
&& chmod +x /tmp/aarch64-pkg-config \
# Build FFmpeg with RKMPP
# Build FFmpeg with RKMPP (minimal build for encoding only)
&& cd ffmpeg-rockchip \
&& ./configure \
--prefix=/usr/aarch64-linux-gnu \
@@ -120,26 +120,69 @@ RUN mkdir -p /tmp/ffmpeg-build && cd /tmp/ffmpeg-build \
--enable-version3 \
--enable-shared \
--disable-static \
# Hardware acceleration
--enable-libdrm \
--enable-rkmpp \
--enable-rkrga \
--enable-libv4l2 \
--enable-vaapi \
--enable-v4l2-m2m \
# Software encoding libraries
--enable-libx264 \
--enable-libx265 \
--enable-libvpx \
--enable-vaapi \
--enable-v4l2-m2m \
# Disable programs and docs
--disable-programs \
--disable-doc \
--disable-htmlpages \
--disable-manpages \
--disable-podpages \
--disable-txtpages \
# Disable network
--disable-network \
--disable-protocols \
# Disable unused libraries
--disable-avformat \
--disable-swscale \
--disable-swresample \
--disable-avfilter \
--disable-avdevice \
--disable-postproc \
# Disable all decoders
--disable-decoders \
# Disable all encoders, enable only needed ones
--disable-encoders \
--enable-encoder=h264_vaapi \
--enable-encoder=hevc_vaapi \
--enable-encoder=vp8_vaapi \
--enable-encoder=vp9_vaapi \
--enable-encoder=h264_rkmpp \
--enable-encoder=hevc_rkmpp \
--enable-encoder=h264_v4l2m2m \
--enable-encoder=hevc_v4l2m2m \
--enable-encoder=libx264 \
--enable-encoder=libx265 \
--enable-encoder=libvpx_vp8 \
--enable-encoder=libvpx_vp9 \
# Disable muxers/demuxers
--disable-muxers \
--disable-demuxers \
# Disable parsers except needed ones
--disable-parsers \
--enable-parser=h264 \
--enable-parser=hevc \
--enable-parser=vp8 \
--enable-parser=vp9 \
# Disable BSFs except needed ones
--disable-bsfs \
--enable-bsf=h264_mp4toannexb \
--enable-bsf=hevc_mp4toannexb \
# Disable hardware decoding
--disable-hwaccels \
# Disable other unused features
--disable-indevs \
--disable-outdevs \
--disable-filters \
--disable-debug \
--disable-decoder=mjpeg \
--disable-decoder=mjpegb \
&& make -j$(nproc) \
&& make install \
&& cd / \

View File

@@ -107,7 +107,7 @@ RUN mkdir -p /tmp/ffmpeg-build && cd /tmp/ffmpeg-build \
&& echo 'export PKG_CONFIG_SYSROOT_DIR=""' >> /tmp/armhf-pkg-config \
&& echo 'exec pkg-config "$@"' >> /tmp/armhf-pkg-config \
&& chmod +x /tmp/armhf-pkg-config \
# Build FFmpeg with RKMPP
# Build FFmpeg with RKMPP (minimal build for encoding only)
&& cd ffmpeg-rockchip \
&& ./configure \
--prefix=/usr/arm-linux-gnueabihf \
@@ -120,26 +120,69 @@ RUN mkdir -p /tmp/ffmpeg-build && cd /tmp/ffmpeg-build \
--enable-version3 \
--enable-shared \
--disable-static \
# Hardware acceleration
--enable-libdrm \
--enable-rkmpp \
--enable-rkrga \
--enable-libv4l2 \
--enable-vaapi \
--enable-v4l2-m2m \
# Software encoding libraries
--enable-libx264 \
--enable-libx265 \
--enable-libvpx \
--enable-vaapi \
--enable-v4l2-m2m \
# Disable programs and docs
--disable-programs \
--disable-doc \
--disable-htmlpages \
--disable-manpages \
--disable-podpages \
--disable-txtpages \
# Disable network
--disable-network \
--disable-protocols \
# Disable unused libraries
--disable-avformat \
--disable-swscale \
--disable-swresample \
--disable-avfilter \
--disable-avdevice \
--disable-postproc \
# Disable all decoders
--disable-decoders \
# Disable all encoders, enable only needed ones
--disable-encoders \
--enable-encoder=h264_vaapi \
--enable-encoder=hevc_vaapi \
--enable-encoder=vp8_vaapi \
--enable-encoder=vp9_vaapi \
--enable-encoder=h264_rkmpp \
--enable-encoder=hevc_rkmpp \
--enable-encoder=h264_v4l2m2m \
--enable-encoder=hevc_v4l2m2m \
--enable-encoder=libx264 \
--enable-encoder=libx265 \
--enable-encoder=libvpx_vp8 \
--enable-encoder=libvpx_vp9 \
# Disable muxers/demuxers
--disable-muxers \
--disable-demuxers \
# Disable parsers except needed ones
--disable-parsers \
--enable-parser=h264 \
--enable-parser=hevc \
--enable-parser=vp8 \
--enable-parser=vp9 \
# Disable BSFs except needed ones
--disable-bsfs \
--enable-bsf=h264_mp4toannexb \
--enable-bsf=hevc_mp4toannexb \
# Disable hardware decoding
--disable-hwaccels \
# Disable other unused features
--disable-indevs \
--disable-outdevs \
--disable-filters \
--disable-debug \
--disable-decoder=mjpeg \
--disable-decoder=mjpegb \
&& make -j$(nproc) \
&& make install \
&& cd / \

View File

@@ -16,19 +16,21 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
ENV PATH="/root/.cargo/bin:${PATH}"
# Install build dependencies (same as runtime Debian 12)
# Install build dependencies
RUN apt-get update && apt-get install -y --no-install-recommends \
# Build tools
build-essential \
pkg-config \
cmake \
nasm \
yasm \
git \
libclang-dev \
llvm \
protobuf-compiler \
libssl-dev \
mold \
wget \
# Core system libraries
libasound2-dev \
libv4l-dev \
@@ -37,13 +39,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
# Video/image processing
libjpeg62-turbo-dev \
libyuv-dev \
# FFmpeg and codecs
libavcodec-dev \
libavformat-dev \
libavutil-dev \
libswscale-dev \
libswresample-dev \
# Video codec libraries
# Video codec libraries (for FFmpeg build)
libx264-dev \
libx265-dev \
libvpx-dev \
@@ -60,8 +56,85 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
libxdmcp-dev \
&& rm -rf /var/lib/apt/lists/*
# Download and build FFmpeg with minimal configuration for encoding only
RUN mkdir -p /tmp/ffmpeg-build && cd /tmp/ffmpeg-build \
&& wget -q https://files.mofeng.run/src/image/other/ffmpeg.tar.gz \
&& tar -xzf ffmpeg.tar.gz \
&& cd ffmpeg/ffmpeg-rockchip \
&& ./configure \
--prefix=/usr/local \
--enable-gpl \
--enable-version3 \
--enable-shared \
--disable-static \
# Hardware acceleration
--enable-libdrm \
--enable-vaapi \
--enable-libmfx \
# Software encoding libraries
--enable-libx264 \
--enable-libx265 \
--enable-libvpx \
# Disable programs and docs
--disable-programs \
--disable-doc \
--disable-htmlpages \
--disable-manpages \
--disable-podpages \
--disable-txtpages \
# Disable network
--disable-network \
--disable-protocols \
# Disable unused libraries
--disable-avformat \
--disable-swscale \
--disable-swresample \
--disable-avfilter \
--disable-avdevice \
--disable-postproc \
# Disable all decoders
--disable-decoders \
# Disable all encoders, enable only needed ones
--disable-encoders \
--enable-encoder=h264_vaapi \
--enable-encoder=hevc_vaapi \
--enable-encoder=vp8_vaapi \
--enable-encoder=vp9_vaapi \
--enable-encoder=h264_qsv \
--enable-encoder=hevc_qsv \
--enable-encoder=libx264 \
--enable-encoder=libx265 \
--enable-encoder=libvpx_vp8 \
--enable-encoder=libvpx_vp9 \
# Disable muxers/demuxers
--disable-muxers \
--disable-demuxers \
# Disable parsers except needed ones
--disable-parsers \
--enable-parser=h264 \
--enable-parser=hevc \
--enable-parser=vp8 \
--enable-parser=vp9 \
# Disable BSFs except needed ones
--disable-bsfs \
--enable-bsf=h264_mp4toannexb \
--enable-bsf=hevc_mp4toannexb \
# Disable hardware decoding
--disable-hwaccels \
# Disable other unused features
--disable-indevs \
--disable-outdevs \
--disable-filters \
--disable-debug \
&& make -j$(nproc) \
&& make install \
&& ldconfig \
&& cd / \
&& rm -rf /tmp/ffmpeg-build
# Add Rust target
RUN rustup target add x86_64-unknown-linux-gnu
# Configure mold as the linker
ENV RUSTFLAGS="-C link-arg=-fuse-ld=mold"
# Configure mold as the linker and use custom FFmpeg
ENV RUSTFLAGS="-C link-arg=-fuse-ld=mold" \
PKG_CONFIG_PATH="/usr/local/lib/pkgconfig:${PKG_CONFIG_PATH}"

View File

@@ -101,7 +101,8 @@ mod ffmpeg {
fn link_system_ffmpeg(builder: &mut Build) {
use std::process::Command;
let libs = ["libavcodec", "libavutil", "libavformat", "libswscale"];
// Only need libavcodec and libavutil for encoding
let libs = ["libavcodec", "libavutil"];
for lib in &libs {
// Get cflags
@@ -134,7 +135,7 @@ mod ffmpeg {
}
}
} else {
panic!("pkg-config failed for {}. Install FFmpeg development libraries: sudo apt install libavcodec-dev libavformat-dev libavutil-dev libswscale-dev", lib);
panic!("pkg-config failed for {}. Install FFmpeg development libraries: sudo apt install libavcodec-dev libavutil-dev", lib);
}
} else {
panic!("pkg-config not found. Install pkg-config and FFmpeg development libraries.");
@@ -178,7 +179,8 @@ mod ffmpeg {
)
);
{
let mut static_libs = vec!["avcodec", "avutil", "avformat"];
// Only need avcodec and avutil for encoding
let mut static_libs = vec!["avcodec", "avutil"];
if target_os == "windows" {
static_libs.push("libmfx");
}
@@ -251,7 +253,7 @@ mod ffmpeg {
.unwrap();
builder.files(
["ffmpeg_ram_encode.cpp", "ffmpeg_ram_decode.cpp"].map(|f| ffmpeg_ram_dir.join(f)),
["ffmpeg_ram_encode.cpp"].map(|f| ffmpeg_ram_dir.join(f)),
);
}
}

View File

@@ -28,7 +28,6 @@ enum DataFormat {
VP8,
VP9,
AV1,
MJPEG,
};
// same as Driver

View File

@@ -1,328 +0,0 @@
// https://github.com/FFmpeg/FFmpeg/blob/master/doc/examples/hw_decode.c
// https://github.com/FFmpeg/FFmpeg/blob/master/doc/examples/decode_video.c
extern "C" {
#include <libavcodec/avcodec.h>
#include <libavutil/log.h>
#include <libavutil/opt.h>
#include <libavutil/pixdesc.h>
}
#include <memory>
#include <stdbool.h>
#define LOG_MODULE "FFMPEG_RAM_DEC"
#include <log.h>
#include <util.h>
#ifdef _WIN32
#include <libavutil/hwcontext_d3d11va.h>
#endif
#include "common.h"
#include "system.h"
// #define CFG_PKG_TRACE
namespace {
typedef void (*RamDecodeCallback)(const void *obj, int width, int height,
enum AVPixelFormat pixfmt,
int linesize[AV_NUM_DATA_POINTERS],
uint8_t *data[AV_NUM_DATA_POINTERS], int key);
class FFmpegRamDecoder {
public:
AVCodecContext *c_ = NULL;
AVBufferRef *hw_device_ctx_ = NULL;
AVFrame *sw_frame_ = NULL;
AVFrame *frame_ = NULL;
AVPacket *pkt_ = NULL;
bool hwaccel_ = true;
std::string name_;
AVHWDeviceType device_type_ = AV_HWDEVICE_TYPE_NONE;
int thread_count_ = 1;
RamDecodeCallback callback_ = NULL;
DataFormat data_format_;
#ifdef CFG_PKG_TRACE
int in_ = 0;
int out_ = 0;
#endif
FFmpegRamDecoder(const char *name, int device_type, int thread_count,
RamDecodeCallback callback) {
this->name_ = name;
this->device_type_ = (AVHWDeviceType)device_type;
this->thread_count_ = thread_count;
this->callback_ = callback;
}
~FFmpegRamDecoder() {}
void free_decoder() {
if (frame_)
av_frame_free(&frame_);
if (pkt_)
av_packet_free(&pkt_);
if (sw_frame_)
av_frame_free(&sw_frame_);
if (c_)
avcodec_free_context(&c_);
if (hw_device_ctx_)
av_buffer_unref(&hw_device_ctx_);
frame_ = NULL;
pkt_ = NULL;
sw_frame_ = NULL;
c_ = NULL;
hw_device_ctx_ = NULL;
}
int reset() {
if (name_.find("h264") != std::string::npos) {
data_format_ = DataFormat::H264;
} else if (name_.find("hevc") != std::string::npos) {
data_format_ = DataFormat::H265;
} else if (name_.find("mjpeg") != std::string::npos) {
data_format_ = DataFormat::MJPEG;
} else {
LOG_ERROR(std::string("unsupported data format:") + name_);
return -1;
}
free_decoder();
const AVCodec *codec = NULL;
hwaccel_ = device_type_ != AV_HWDEVICE_TYPE_NONE;
int ret;
if (!(codec = avcodec_find_decoder_by_name(name_.c_str()))) {
LOG_ERROR(std::string("avcodec_find_decoder_by_name ") + name_ + " failed");
return -1;
}
if (!(c_ = avcodec_alloc_context3(codec))) {
LOG_ERROR(std::string("Could not allocate video codec context"));
return -1;
}
c_->flags |= AV_CODEC_FLAG_LOW_DELAY;
c_->thread_count =
device_type_ != AV_HWDEVICE_TYPE_NONE ? 1 : thread_count_;
c_->thread_type = FF_THREAD_SLICE;
if (name_.find("qsv") != std::string::npos) {
if ((ret = av_opt_set(c_->priv_data, "async_depth", "1", 0)) < 0) {
LOG_ERROR(std::string("qsv set opt async_depth 1 failed"));
return -1;
}
// https://github.com/FFmpeg/FFmpeg/blob/c6364b711bad1fe2fbd90e5b2798f87080ddf5ea/libavcodec/qsvdec.c#L932
// for disable warning
c_->pkt_timebase = av_make_q(1, 30);
}
if (hwaccel_) {
ret =
av_hwdevice_ctx_create(&hw_device_ctx_, device_type_, NULL, NULL, 0);
if (ret < 0) {
LOG_ERROR(std::string("av_hwdevice_ctx_create failed, ret = ") + av_err2str(ret));
return -1;
}
c_->hw_device_ctx = av_buffer_ref(hw_device_ctx_);
if (!check_support()) {
LOG_ERROR(std::string("check_support failed"));
return -1;
}
if (!(sw_frame_ = av_frame_alloc())) {
LOG_ERROR(std::string("av_frame_alloc failed"));
return -1;
}
}
if (!(pkt_ = av_packet_alloc())) {
LOG_ERROR(std::string("av_packet_alloc failed"));
return -1;
}
if (!(frame_ = av_frame_alloc())) {
LOG_ERROR(std::string("av_frame_alloc failed"));
return -1;
}
if ((ret = avcodec_open2(c_, codec, NULL)) != 0) {
LOG_ERROR(std::string("avcodec_open2 failed, ret = ") + av_err2str(ret));
return -1;
}
#ifdef CFG_PKG_TRACE
in_ = 0;
out_ = 0;
#endif
return 0;
}
int decode(const uint8_t *data, int length, const void *obj) {
int ret = -1;
#ifdef CFG_PKG_TRACE
in_++;
LOG_DEBUG(std::string("delay DI: in:") + in_ + " out:" + out_);
#endif
if (!data || !length) {
LOG_ERROR(std::string("illegal decode parameter"));
return -1;
}
pkt_->data = (uint8_t *)data;
pkt_->size = length;
ret = do_decode(obj);
return ret;
}
private:
int do_decode(const void *obj) {
int ret;
AVFrame *tmp_frame = NULL;
bool decoded = false;
ret = avcodec_send_packet(c_, pkt_);
if (ret < 0) {
LOG_ERROR(std::string("avcodec_send_packet failed, ret = ") + av_err2str(ret));
return ret;
}
auto start = util::now();
while (ret >= 0 && util::elapsed_ms(start) < ENCODE_TIMEOUT_MS) {
if ((ret = avcodec_receive_frame(c_, frame_)) != 0) {
if (ret != AVERROR(EAGAIN)) {
LOG_ERROR(std::string("avcodec_receive_frame failed, ret = ") + av_err2str(ret));
}
goto _exit;
}
if (hwaccel_) {
if (!frame_->hw_frames_ctx) {
LOG_ERROR(std::string("hw_frames_ctx is NULL"));
goto _exit;
}
if ((ret = av_hwframe_transfer_data(sw_frame_, frame_, 0)) < 0) {
LOG_ERROR(std::string("av_hwframe_transfer_data failed, ret = ") +
av_err2str(ret));
goto _exit;
}
tmp_frame = sw_frame_;
} else {
tmp_frame = frame_;
}
decoded = true;
#ifdef CFG_PKG_TRACE
out_++;
LOG_DEBUG(std::string("delay DO: in:") + in_ + " out:" + out_);
#endif
#if FF_API_FRAME_KEY
int key_frame = frame_->flags & AV_FRAME_FLAG_KEY;
#else
int key_frame = frame_->key_frame;
#endif
callback_(obj, tmp_frame->width, tmp_frame->height,
(AVPixelFormat)tmp_frame->format, tmp_frame->linesize,
tmp_frame->data, key_frame);
}
_exit:
av_packet_unref(pkt_);
return decoded ? 0 : -1;
}
bool check_support() {
#ifdef _WIN32
if (device_type_ == AV_HWDEVICE_TYPE_D3D11VA) {
if (!c_->hw_device_ctx) {
LOG_ERROR(std::string("hw_device_ctx is NULL"));
return false;
}
AVHWDeviceContext *deviceContext =
(AVHWDeviceContext *)hw_device_ctx_->data;
if (!deviceContext) {
LOG_ERROR(std::string("deviceContext is NULL"));
return false;
}
AVD3D11VADeviceContext *d3d11vaDeviceContext =
(AVD3D11VADeviceContext *)deviceContext->hwctx;
if (!d3d11vaDeviceContext) {
LOG_ERROR(std::string("d3d11vaDeviceContext is NULL"));
return false;
}
ID3D11Device *device = d3d11vaDeviceContext->device;
if (!device) {
LOG_ERROR(std::string("device is NULL"));
return false;
}
std::unique_ptr<NativeDevice> native_ = std::make_unique<NativeDevice>();
if (!native_) {
LOG_ERROR(std::string("Failed to create native device"));
return false;
}
if (!native_->Init(0, (ID3D11Device *)device, 0)) {
LOG_ERROR(std::string("Failed to init native device"));
return false;
}
if (!native_->support_decode(data_format_)) {
LOG_ERROR(std::string("Failed to check support ") + name_);
return false;
}
return true;
} else {
return true;
}
#else
return true;
#endif
}
};
} // namespace
extern "C" void ffmpeg_ram_free_decoder(FFmpegRamDecoder *decoder) {
try {
if (!decoder)
return;
decoder->free_decoder();
delete decoder;
decoder = NULL;
} catch (const std::exception &e) {
LOG_ERROR(std::string("ffmpeg_ram_free_decoder exception:") + e.what());
}
}
extern "C" FFmpegRamDecoder *
ffmpeg_ram_new_decoder(const char *name, int device_type, int thread_count,
RamDecodeCallback callback) {
FFmpegRamDecoder *decoder = NULL;
try {
decoder = new FFmpegRamDecoder(name, device_type, thread_count, callback);
if (decoder) {
if (decoder->reset() == 0) {
return decoder;
}
}
} catch (std::exception &e) {
LOG_ERROR(std::string("new decoder exception:") + e.what());
}
if (decoder) {
decoder->free_decoder();
delete decoder;
decoder = NULL;
}
return NULL;
}
extern "C" int ffmpeg_ram_decode(FFmpegRamDecoder *decoder, const uint8_t *data,
int length, const void *obj) {
try {
int ret = decoder->decode(data, length, obj);
if (DataFormat::H265 == decoder->data_format_ && util_decode::has_flag_could_not_find_ref_with_poc()) {
return HWCODEC_ERR_HEVC_COULD_NOT_FIND_POC;
} else {
return ret == 0 ? HWCODEC_SUCCESS : HWCODEC_ERR_COMMON;
}
} catch (const std::exception &e) {
LOG_ERROR(std::string("ffmpeg_ram_decode exception:") + e.what());
}
return HWCODEC_ERR_COMMON;
}

View File

@@ -5,10 +5,6 @@
#define AV_NUM_DATA_POINTERS 8
typedef void (*RamDecodeCallback)(const void *obj, int width, int height,
int pixfmt,
int linesize[AV_NUM_DATA_POINTERS],
uint8_t *data[AV_NUM_DATA_POINTERS], int key);
typedef void (*RamEncodeCallback)(const uint8_t *data, int len, int64_t pts,
int key, const void *obj);
@@ -18,18 +14,13 @@ void *ffmpeg_ram_new_encoder(const char *name, const char *mc_name, int width,
int thread_count, int gpu, int *linesize,
int *offset, int *length,
RamEncodeCallback callback);
void *ffmpeg_ram_new_decoder(const char *name, int device_type,
int thread_count, RamDecodeCallback callback);
int ffmpeg_ram_encode(void *encoder, const uint8_t *data, int length,
const void *obj, int64_t ms);
int ffmpeg_ram_decode(void *decoder, const uint8_t *data, int length,
const void *obj);
void ffmpeg_ram_free_encoder(void *encoder);
void ffmpeg_ram_free_decoder(void *decoder);
int ffmpeg_ram_get_linesize_offset_length(int pix_fmt, int width, int height,
int align, int *linesize, int *offset,
int *length);
int ffmpeg_ram_set_bitrate(void *encoder, int kbs);
void ffmpeg_ram_request_keyframe(void *encoder);
#endif // FFMPEG_RAM_FFI_H
#endif // FFMPEG_RAM_FFI_H

View File

@@ -1,202 +0,0 @@
use crate::ffmpeg::{init_av_log, AVHWDeviceType::*};
use crate::{
common::DataFormat::*,
ffmpeg::{AVHWDeviceType, AVPixelFormat},
ffmpeg_ram::{
ffmpeg_ram_decode, ffmpeg_ram_free_decoder, ffmpeg_ram_new_decoder, CodecInfo,
AV_NUM_DATA_POINTERS, Priority,
},
};
use log::error;
use std::{
ffi::{c_void, CString},
os::raw::c_int,
slice::from_raw_parts,
vec,
};
#[derive(Debug, Clone)]
pub struct DecodeContext {
pub name: String,
pub device_type: AVHWDeviceType,
pub thread_count: i32,
}
pub struct DecodeFrame {
pub pixfmt: AVPixelFormat,
pub width: i32,
pub height: i32,
pub data: Vec<Vec<u8>>,
pub linesize: Vec<i32>,
pub key: bool,
}
impl std::fmt::Display for DecodeFrame {
fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
let mut s = String::from("data:");
for data in self.data.iter() {
s.push_str(format!("{} ", data.len()).as_str());
}
s.push_str(", linesize:");
for linesize in self.linesize.iter() {
s.push_str(format!("{} ", linesize).as_str());
}
write!(
f,
"fixfmt:{}, width:{}, height:{},key:{}, {}",
self.pixfmt as i32, self.width, self.height, self.key, s,
)
}
}
pub struct Decoder {
codec: *mut c_void,
frames: *mut Vec<DecodeFrame>,
pub ctx: DecodeContext,
}
unsafe impl Send for Decoder {}
unsafe impl Sync for Decoder {}
impl Decoder {
pub fn new(ctx: DecodeContext) -> Result<Self, ()> {
init_av_log();
unsafe {
let codec = ffmpeg_ram_new_decoder(
CString::new(ctx.name.as_str()).map_err(|_| ())?.as_ptr(),
ctx.device_type as _,
ctx.thread_count,
Some(Decoder::callback),
);
if codec.is_null() {
return Err(());
}
Ok(Decoder {
codec,
frames: Box::into_raw(Box::new(Vec::<DecodeFrame>::new())),
ctx,
})
}
}
pub fn decode(&mut self, packet: &[u8]) -> Result<&mut Vec<DecodeFrame>, i32> {
unsafe {
(&mut *self.frames).clear();
let ret = ffmpeg_ram_decode(
self.codec,
packet.as_ptr(),
packet.len() as c_int,
self.frames as *const _ as *const c_void,
);
if ret < 0 {
Err(ret)
} else {
Ok(&mut *self.frames)
}
}
}
unsafe extern "C" fn callback(
obj: *const c_void,
width: c_int,
height: c_int,
pixfmt: c_int,
linesizes: *mut c_int,
datas: *mut *mut u8,
key: c_int,
) {
let frames = &mut *(obj as *mut Vec<DecodeFrame>);
let datas = from_raw_parts(datas, AV_NUM_DATA_POINTERS as _);
let linesizes = from_raw_parts(linesizes, AV_NUM_DATA_POINTERS as _);
let mut frame = DecodeFrame {
pixfmt: std::mem::transmute(pixfmt),
width,
height,
data: vec![],
linesize: vec![],
key: key != 0,
};
// Handle YUV420P and YUVJ420P (JPEG full-range) - same memory layout
if pixfmt == AVPixelFormat::AV_PIX_FMT_YUV420P as c_int
|| pixfmt == AVPixelFormat::AV_PIX_FMT_YUVJ420P as c_int
{
let y = from_raw_parts(datas[0], (linesizes[0] * height) as usize).to_vec();
let u = from_raw_parts(datas[1], (linesizes[1] * height / 2) as usize).to_vec();
let v = from_raw_parts(datas[2], (linesizes[2] * height / 2) as usize).to_vec();
frame.data.push(y);
frame.data.push(u);
frame.data.push(v);
frame.linesize.push(linesizes[0]);
frame.linesize.push(linesizes[1]);
frame.linesize.push(linesizes[2]);
frames.push(frame);
} else if pixfmt == AVPixelFormat::AV_PIX_FMT_YUV422P as c_int
|| pixfmt == AVPixelFormat::AV_PIX_FMT_YUVJ422P as c_int
{
// YUV422P: U and V planes have same height as Y (not half)
let y = from_raw_parts(datas[0], (linesizes[0] * height) as usize).to_vec();
let u = from_raw_parts(datas[1], (linesizes[1] * height) as usize).to_vec();
let v = from_raw_parts(datas[2], (linesizes[2] * height) as usize).to_vec();
frame.data.push(y);
frame.data.push(u);
frame.data.push(v);
frame.linesize.push(linesizes[0]);
frame.linesize.push(linesizes[1]);
frame.linesize.push(linesizes[2]);
frames.push(frame);
} else if pixfmt == AVPixelFormat::AV_PIX_FMT_NV12 as c_int
|| pixfmt == AVPixelFormat::AV_PIX_FMT_NV21 as c_int
{
let y = from_raw_parts(datas[0], (linesizes[0] * height) as usize).to_vec();
let uv = from_raw_parts(datas[1], (linesizes[1] * height / 2) as usize).to_vec();
frame.data.push(y);
frame.data.push(uv);
frame.linesize.push(linesizes[0]);
frame.linesize.push(linesizes[1]);
frames.push(frame);
} else {
error!("unsupported pixfmt {}", pixfmt as i32);
}
}
/// Returns available decoders for IP-KVM scenario.
/// Only MJPEG software decoder is supported as IP-KVM captures from video capture cards
/// that output MJPEG streams.
pub fn available_decoders() -> Vec<CodecInfo> {
// IP-KVM scenario only needs MJPEG decoding
// MJPEG comes from video capture cards, software decoding is sufficient
vec![CodecInfo {
name: "mjpeg".to_owned(),
format: MJPEG,
hwdevice: AV_HWDEVICE_TYPE_NONE,
priority: Priority::Best as _,
..Default::default()
}]
}
}
impl Drop for Decoder {
fn drop(&mut self) {
unsafe {
ffmpeg_ram_free_decoder(self.codec);
self.codec = std::ptr::null_mut();
let _ = Box::from_raw(self.frames);
}
}
}

View File

@@ -447,14 +447,6 @@ impl Encoder {
}
}
// Add MJPEG software encoder if not already present
if !res.iter().any(|c| c.format == MJPEG) {
if let Some(mjpeg_soft) = soft_codecs.mjpeg {
debug!("Adding software MJPEG encoder: {}", mjpeg_soft.name);
res.push(mjpeg_soft);
}
}
res
}

View File

@@ -12,7 +12,6 @@ use std::ffi::c_int;
include!(concat!(env!("OUT_DIR"), "/ffmpeg_ram_ffi.rs"));
pub mod decode;
pub mod encode;
pub enum Priority {
@@ -52,7 +51,6 @@ impl CodecInfo {
let mut vp8: Option<CodecInfo> = None;
let mut vp9: Option<CodecInfo> = None;
let mut av1: Option<CodecInfo> = None;
let mut mjpeg: Option<CodecInfo> = None;
for coder in coders {
match coder.format {
@@ -96,14 +94,6 @@ impl CodecInfo {
}
None => av1 = Some(coder),
},
DataFormat::MJPEG => match &mjpeg {
Some(old) => {
if old.priority > coder.priority {
mjpeg = Some(coder)
}
}
None => mjpeg = Some(coder),
},
}
}
CodecInfos {
@@ -112,7 +102,6 @@ impl CodecInfo {
vp8,
vp9,
av1,
mjpeg,
}
}
@@ -147,13 +136,6 @@ impl CodecInfo {
priority: Priority::Soft as _,
}),
av1: None,
mjpeg: Some(CodecInfo {
name: "mjpeg".to_owned(),
mc_name: Default::default(),
format: MJPEG,
hwdevice: AV_HWDEVICE_TYPE_NONE,
priority: Priority::Soft as _,
}),
}
}
}
@@ -165,7 +147,6 @@ pub struct CodecInfos {
pub vp8: Option<CodecInfo>,
pub vp9: Option<CodecInfo>,
pub av1: Option<CodecInfo>,
pub mjpeg: Option<CodecInfo>,
}
impl CodecInfos {

View File

@@ -1,36 +1,29 @@
//! MJPEG VAAPI hardware decoder
//! MJPEG decoder implementations
//!
//! Uses hwcodec's FFmpeg VAAPI backend to decode MJPEG to NV12.
//! This provides hardware-accelerated JPEG decoding with direct NV12 output,
//! which is the optimal format for VAAPI H264 encoding.
//! Provides MJPEG decoding using libyuv for SIMD-accelerated decoding.
//! All decoders output to standard YUV formats suitable for encoding.
use std::sync::Once;
use tracing::{debug, info, warn};
use hwcodec::ffmpeg::AVHWDeviceType;
use hwcodec::ffmpeg::AVPixelFormat;
use hwcodec::ffmpeg_ram::decode::{DecodeContext, DecodeFrame, Decoder};
use tracing::{debug, info};
use crate::error::{AppError, Result};
use crate::video::format::Resolution;
// libyuv for SIMD-accelerated YUV conversion
static INIT_LOGGING: Once = Once::new();
/// Initialize hwcodec logging (only once)
fn init_hwcodec_logging() {
/// Initialize decoder logging (only once)
fn init_decoder_logging() {
INIT_LOGGING.call_once(|| {
debug!("hwcodec MJPEG decoder logging initialized");
debug!("MJPEG decoder logging initialized");
});
}
/// MJPEG VAAPI decoder configuration
/// MJPEG decoder configuration
#[derive(Debug, Clone)]
pub struct MjpegVaapiDecoderConfig {
/// Expected resolution (can be updated from decoded frame)
pub resolution: Resolution,
/// Use hardware acceleration (VAAPI)
/// Use hardware acceleration (ignored, kept for API compatibility)
pub use_hwaccel: bool,
}
@@ -266,62 +259,34 @@ impl DecodedNv12Frame {
}
}
/// MJPEG VAAPI hardware decoder
/// MJPEG decoder with NV12 output
///
/// Decodes MJPEG frames to NV12 format using VAAPI hardware acceleration.
/// This is optimal for feeding into VAAPI H264 encoder.
/// Uses libyuv for SIMD-accelerated MJPEG decoding to YUV420P,
/// then converts to NV12 for hardware encoder compatibility.
/// Named "VaapiDecoder" for API compatibility with existing code.
pub struct MjpegVaapiDecoder {
/// hwcodec decoder instance
decoder: Decoder,
/// Configuration
config: MjpegVaapiDecoderConfig,
/// Frame counter
frame_count: u64,
/// Whether hardware acceleration is active
hwaccel_active: bool,
}
impl MjpegVaapiDecoder {
/// Create a new MJPEG decoder
/// Note: VAAPI does not support MJPEG decoding on most hardware,
/// so we use software decoding and convert to NV12 for VAAPI encoding.
pub fn new(config: MjpegVaapiDecoderConfig) -> Result<Self> {
init_hwcodec_logging();
// VAAPI doesn't support MJPEG decoding, always use software decoder
// The output will be converted to NV12 for VAAPI H264 encoding
let device_type = AVHWDeviceType::AV_HWDEVICE_TYPE_NONE;
init_decoder_logging();
info!(
"Creating MJPEG decoder with software decoding (VAAPI doesn't support MJPEG decode)"
);
let ctx = DecodeContext {
name: "mjpeg".to_string(),
device_type,
thread_count: 4, // Use multiple threads for software decoding
};
let decoder = Decoder::new(ctx).map_err(|_| {
AppError::VideoError("Failed to create MJPEG software decoder".to_string())
})?;
// hwaccel is not actually active for MJPEG decoding
let hwaccel_active = false;
info!(
"MJPEG decoder created successfully (software decode, will convert to NV12)"
"Creating MJPEG decoder with libyuv (SIMD-accelerated, NV12 output)"
);
Ok(Self {
decoder,
config,
frame_count: 0,
hwaccel_active,
})
}
/// Create with default config (VAAPI enabled)
/// Create with default config
pub fn with_vaapi(resolution: Resolution) -> Result<Self> {
Self::new(MjpegVaapiDecoderConfig {
resolution,
@@ -329,7 +294,7 @@ impl MjpegVaapiDecoder {
})
}
/// Create with software decoding (fallback)
/// Create with software decoding (same as with_vaapi, kept for API compatibility)
pub fn with_software(resolution: Resolution) -> Result<Self> {
Self::new(MjpegVaapiDecoderConfig {
resolution,
@@ -337,9 +302,9 @@ impl MjpegVaapiDecoder {
})
}
/// Check if hardware acceleration is active
/// Check if hardware acceleration is active (always false, using libyuv)
pub fn is_hwaccel_active(&self) -> bool {
self.hwaccel_active
false
}
/// Decode MJPEG frame to NV12
@@ -357,130 +322,27 @@ impl MjpegVaapiDecoder {
self.frame_count += 1;
let frames = self.decoder.decode(jpeg_data).map_err(|e| {
AppError::VideoError(format!("MJPEG decode failed: error code {}", e))
})?;
// Get JPEG dimensions
let (width, height) = libyuv::mjpeg_size(jpeg_data)
.map_err(|e| AppError::VideoError(format!("Failed to read MJPEG size: {}", e)))?;
if frames.is_empty() {
return Err(AppError::VideoError("Decoder returned no frames".to_string()));
}
let frame = &frames[0];
// Handle different output formats
// VAAPI MJPEG decoder may output NV12, YUV420P, or YUVJ420P (JPEG full-range)
if frame.pixfmt == AVPixelFormat::AV_PIX_FMT_NV12
|| frame.pixfmt == AVPixelFormat::AV_PIX_FMT_NV21
{
// NV12/NV21 format: Y plane + UV interleaved plane
if frame.data.len() < 2 {
return Err(AppError::VideoError("Invalid NV12 frame data".to_string()));
}
return Ok(DecodedNv12Frame {
y_plane: frame.data[0].clone(),
uv_plane: frame.data[1].clone(),
y_linesize: frame.linesize[0],
uv_linesize: frame.linesize[1],
width: frame.width,
height: frame.height,
});
}
// YUV420P or YUVJ420P (JPEG full-range) - need to convert to NV12
if frame.pixfmt == AVPixelFormat::AV_PIX_FMT_YUV420P
|| frame.pixfmt == AVPixelFormat::AV_PIX_FMT_YUVJ420P
{
return Self::convert_yuv420p_to_nv12_static(frame);
}
// YUV422P or YUVJ422P (JPEG full-range 4:2:2) - need to convert to NV12
if frame.pixfmt == AVPixelFormat::AV_PIX_FMT_YUV422P
|| frame.pixfmt == AVPixelFormat::AV_PIX_FMT_YUVJ422P
{
return Self::convert_yuv422p_to_nv12_static(frame);
}
Err(AppError::VideoError(format!(
"Unexpected output format: {:?} (expected NV12, YUV420P, YUV422P, or YUVJ variants)",
frame.pixfmt
)))
}
/// Convert YUV420P frame to NV12 format using libyuv (SIMD accelerated)
fn convert_yuv420p_to_nv12_static(frame: &DecodeFrame) -> Result<DecodedNv12Frame> {
if frame.data.len() < 3 {
return Err(AppError::VideoError("Invalid YUV420P frame data".to_string()));
}
let width = frame.width as i32;
let height = frame.height as i32;
let y_linesize = frame.linesize[0];
let u_linesize = frame.linesize[1];
let v_linesize = frame.linesize[2];
// Allocate packed NV12 output buffer
let nv12_size = (width * height * 3 / 2) as usize;
let mut nv12_data = vec![0u8; nv12_size];
// Use libyuv for SIMD-accelerated I420 → NV12 conversion
libyuv::i420_to_nv12_planar(
&frame.data[0], y_linesize,
&frame.data[1], u_linesize,
&frame.data[2], v_linesize,
&mut nv12_data,
width, height,
).map_err(|e| AppError::VideoError(format!("libyuv I420→NV12 failed: {}", e)))?;
// Split into Y and UV planes for DecodedNv12Frame
// Decode MJPEG to YUV420P first
let y_size = (width * height) as usize;
let y_plane = nv12_data[..y_size].to_vec();
let uv_plane = nv12_data[y_size..].to_vec();
let uv_size = y_size / 4;
let yuv420_size = y_size + uv_size * 2;
let mut yuv_data = vec![0u8; yuv420_size];
Ok(DecodedNv12Frame {
y_plane,
uv_plane,
y_linesize: width, // Output is packed, no padding
uv_linesize: width,
width: frame.width,
height: frame.height,
})
}
libyuv::mjpeg_to_i420(jpeg_data, &mut yuv_data, width, height)
.map_err(|e| AppError::VideoError(format!("libyuv MJPEG→I420 failed: {}", e)))?;
/// Convert YUV422P frame to NV12 format using libyuv (SIMD accelerated)
/// Pipeline: I422 (YUV422P) → I420 → NV12
fn convert_yuv422p_to_nv12_static(frame: &DecodeFrame) -> Result<DecodedNv12Frame> {
if frame.data.len() < 3 {
return Err(AppError::VideoError("Invalid YUV422P frame data".to_string()));
}
let width = frame.width as i32;
let height = frame.height as i32;
let y_linesize = frame.linesize[0];
let u_linesize = frame.linesize[1];
let v_linesize = frame.linesize[2];
// Step 1: I422 → I420 (vertical chroma downsampling via SIMD)
let i420_size = (width * height * 3 / 2) as usize;
let mut i420_data = vec![0u8; i420_size];
libyuv::i422_to_i420_planar(
&frame.data[0], y_linesize,
&frame.data[1], u_linesize,
&frame.data[2], v_linesize,
&mut i420_data,
width, height,
).map_err(|e| AppError::VideoError(format!("libyuv I422→I420 failed: {}", e)))?;
// Step 2: I420 → NV12 (UV interleaving via SIMD)
// Convert I420 to NV12
let nv12_size = (width * height * 3 / 2) as usize;
let mut nv12_data = vec![0u8; nv12_size];
libyuv::i420_to_nv12(&i420_data, &mut nv12_data, width, height)
libyuv::i420_to_nv12(&yuv_data, &mut nv12_data, width, height)
.map_err(|e| AppError::VideoError(format!("libyuv I420→NV12 failed: {}", e)))?;
// Split into Y and UV planes for DecodedNv12Frame
let y_size = (width * height) as usize;
// Split into Y and UV planes
let y_plane = nv12_data[..y_size].to_vec();
let uv_plane = nv12_data[y_size..].to_vec();
@@ -489,8 +351,8 @@ impl MjpegVaapiDecoder {
uv_plane,
y_linesize: width,
uv_linesize: width,
width: frame.width,
height: frame.height,
width,
height,
})
}
@@ -600,36 +462,10 @@ impl MjpegTurboDecoder {
}
}
/// Check if MJPEG VAAPI decoder is available
pub fn is_mjpeg_vaapi_available() -> bool {
let ctx = DecodeContext {
name: "mjpeg".to_string(),
device_type: AVHWDeviceType::AV_HWDEVICE_TYPE_VAAPI,
thread_count: 1,
};
match Decoder::new(ctx) {
Ok(_) => {
info!("MJPEG VAAPI decoder is available");
true
}
Err(_) => {
warn!("MJPEG VAAPI decoder is not available");
false
}
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn test_mjpeg_vaapi_availability() {
let available = is_mjpeg_vaapi_available();
println!("MJPEG VAAPI available: {}", available);
}
#[test]
fn test_decoder_creation() {
let config = MjpegVaapiDecoderConfig::default();

View File

@@ -92,6 +92,8 @@ pub enum H265InputFormat {
Yuv420p,
/// NV12 - Y plane + interleaved UV plane (optimal for hardware encoders)
Nv12,
/// YUYV422 - packed YUV 4:2:2 format (optimal for RKMPP direct input)
Yuyv422,
}
impl Default for H265InputFormat {
@@ -145,6 +147,23 @@ impl H265Config {
}
}
/// Create config for low latency streaming with YUYV422 input (optimal for RKMPP direct input)
pub fn low_latency_yuyv422(resolution: Resolution, bitrate_kbps: u32) -> Self {
Self {
base: EncoderConfig {
resolution,
input_format: PixelFormat::Yuyv,
quality: bitrate_kbps,
fps: 30,
gop_size: 30,
},
bitrate_kbps,
gop_size: 30,
fps: 30,
input_format: H265InputFormat::Yuyv422,
}
}
/// Create config for quality streaming
pub fn quality(resolution: Resolution, bitrate_kbps: u32) -> Self {
Self {
@@ -311,13 +330,14 @@ impl H265Encoder {
let width = config.base.resolution.width;
let height = config.base.resolution.height;
// Software encoders (libx265) require YUV420P, hardware encoders use NV12
// Software encoders (libx265) require YUV420P, hardware encoders use NV12 or YUYV422
let (pixfmt, actual_input_format) = if is_software {
(AVPixelFormat::AV_PIX_FMT_YUV420P, H265InputFormat::Yuv420p)
} else {
match config.input_format {
H265InputFormat::Nv12 => (AVPixelFormat::AV_PIX_FMT_NV12, H265InputFormat::Nv12),
H265InputFormat::Yuv420p => (AVPixelFormat::AV_PIX_FMT_YUV420P, H265InputFormat::Yuv420p),
H265InputFormat::Yuyv422 => (AVPixelFormat::AV_PIX_FMT_YUYV422, H265InputFormat::Yuyv422),
}
};
@@ -548,6 +568,7 @@ impl Encoder for H265Encoder {
match self.config.input_format {
H265InputFormat::Nv12 => matches!(format, PixelFormat::Nv12),
H265InputFormat::Yuv420p => matches!(format, PixelFormat::Yuv420),
H265InputFormat::Yuyv422 => matches!(format, PixelFormat::Yuyv),
}
}
}

View File

@@ -411,9 +411,22 @@ impl SharedVideoPipeline {
Box::new(H264EncoderWrapper(encoder))
}
VideoEncoderType::H265 => {
let encoder_config = H265Config::low_latency(config.resolution, config.bitrate_kbps);
// Determine H265 input format based on backend and input format
let encoder_config = if use_yuyv_direct {
H265Config::low_latency_yuyv422(config.resolution, config.bitrate_kbps)
} else {
H265Config::low_latency(config.resolution, config.bitrate_kbps)
};
let encoder = if let Some(ref backend) = config.encoder_backend {
let encoder = if use_yuyv_direct {
// Force RKMPP backend for YUYV direct input
let codec_name = get_codec_name(VideoEncoderType::H265, Some(EncoderBackend::Rkmpp))
.ok_or_else(|| AppError::VideoError(
"RKMPP backend not available for H.265".to_string()
))?;
info!("Creating H265 encoder with RKMPP backend for YUYV direct input (codec: {})", codec_name);
H265Encoder::with_codec(encoder_config, &codec_name)?
} else if let Some(ref backend) = config.encoder_backend {
let codec_name = get_codec_name(VideoEncoderType::H265, Some(*backend))
.ok_or_else(|| AppError::VideoError(format!(
"Backend {:?} does not support H.265", backend