Files
zap-stream-core/src/encode/audio.rs
2024-08-29 11:47:19 +01:00

329 lines
10 KiB
Rust

use std::ffi::CStr;
use std::mem::transmute;
use std::ptr;
use anyhow::Error;
use ffmpeg_sys_next::{
av_audio_fifo_alloc, av_audio_fifo_free, av_audio_fifo_read,
av_audio_fifo_size, av_audio_fifo_write,
av_channel_layout_copy, av_frame_alloc, av_frame_free, av_get_sample_fmt_name, av_packet_alloc, av_packet_free, av_samples_alloc_array_and_samples,
avcodec_alloc_context3, avcodec_free_context, avcodec_open2,
avcodec_receive_packet, avcodec_send_frame, swr_alloc_set_opts2,
swr_convert_frame, swr_free, swr_init, AVAudioFifo, AVCodec,
AVCodecContext, AVFrame, SwrContext, AVERROR,
};
use libc::EAGAIN;
use log::info;
use tokio::sync::mpsc::UnboundedSender;
use crate::encode::set_encoded_pkt_timing;
use crate::ipc::Rx;
use crate::pipeline::{AVFrameSource, AVPacketSource, PipelinePayload, PipelineProcessor};
use crate::utils::get_ffmpeg_error_msg;
use crate::variant::{AudioVariant, VariantStream, VariantStreamType};
pub struct AudioEncoder<T> {
variant: AudioVariant,
ctx: *mut AVCodecContext,
codec: *const AVCodec,
swr_ctx: *mut SwrContext,
fifo: *mut AVAudioFifo,
chan_in: T,
chan_out: UnboundedSender<PipelinePayload>,
pts: i64,
frame_pts: i64,
}
unsafe impl<T> Send for AudioEncoder<T> {}
unsafe impl<T> Sync for AudioEncoder<T> {}
impl<T> Drop for AudioEncoder<T> {
fn drop(&mut self) {
unsafe {
swr_free(&mut self.swr_ctx);
av_audio_fifo_free(self.fifo);
avcodec_free_context(&mut self.ctx);
}
}
}
impl<TRecv> AudioEncoder<TRecv>
where
TRecv: Rx<PipelinePayload>,
{
pub fn new(
chan_in: TRecv,
chan_out: UnboundedSender<PipelinePayload>,
variant: AudioVariant,
) -> Self {
Self {
ctx: ptr::null_mut(),
codec: ptr::null(),
swr_ctx: ptr::null_mut(),
fifo: ptr::null_mut(),
variant,
chan_in,
chan_out,
pts: 0,
frame_pts: 0,
}
}
unsafe fn setup_encoder(&mut self, frame: *mut AVFrame) -> Result<(), Error> {
if self.ctx.is_null() {
let encoder = self.variant.get_codec();
if encoder.is_null() {
return Err(Error::msg("Encoder not found"));
}
let ctx = avcodec_alloc_context3(encoder);
if ctx.is_null() {
return Err(Error::msg("Failed to allocate encoder context"));
}
self.variant.to_codec_context(ctx);
// setup re-sampler if output format does not match input format
if (*ctx).sample_fmt != transmute((*frame).format)
|| (*ctx).sample_rate != (*frame).sample_rate
|| (*ctx).ch_layout.nb_channels != (*frame).ch_layout.nb_channels
{
info!(
"Setup audio resampler: {}.{}@{} -> {}.{}@{}",
(*frame).ch_layout.nb_channels,
CStr::from_ptr(av_get_sample_fmt_name(transmute((*frame).format)))
.to_str()
.unwrap(),
(*frame).sample_rate,
(*ctx).ch_layout.nb_channels,
CStr::from_ptr(av_get_sample_fmt_name((*ctx).sample_fmt))
.to_str()
.unwrap(),
(*ctx).sample_rate
);
let mut swr_ctx = ptr::null_mut();
let ret = swr_alloc_set_opts2(
&mut swr_ctx,
&(*ctx).ch_layout,
(*ctx).sample_fmt,
(*ctx).sample_rate,
&(*frame).ch_layout,
transmute((*frame).format),
(*frame).sample_rate,
0,
ptr::null_mut(),
);
if ret < 0 {
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
let ret = swr_init(swr_ctx);
if ret < 0 {
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
self.swr_ctx = swr_ctx;
let fifo = av_audio_fifo_alloc((*ctx).sample_fmt, (*ctx).ch_layout.nb_channels, 1);
if fifo.is_null() {
return Err(Error::msg("Failed to allocate audio FIFO"));
}
self.fifo = fifo;
}
let ret = avcodec_open2(ctx, encoder, ptr::null_mut());
if ret < 0 {
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
// copy channel layout from codec
let mut px = (*encoder).ch_layouts;
while !px.is_null() {
if (*px).nb_channels as u16 == self.variant.channels {
av_channel_layout_copy(&mut (*ctx).ch_layout, px);
break;
}
px = px.add(1);
}
// let downstream steps know about the encoder
self.chan_out.send(PipelinePayload::EncoderInfo(
VariantStream::Audio(self.variant.clone()),
ctx,
))?;
self.ctx = ctx;
self.codec = encoder;
}
Ok(())
}
/// Returns true if we should process audio frame from FIFO
/// false if nothing to process this frame
unsafe fn process_audio_frame(
&mut self,
frame: *mut AVFrame,
) -> Result<Option<*mut AVFrame>, Error> {
if self.swr_ctx.is_null() {
// no re-sampler, return input frame
return Ok(Some(frame));
}
let mut out_frame = self.new_frame();
let ret = swr_convert_frame(self.swr_ctx, out_frame, frame);
if ret < 0 {
av_frame_free(&mut out_frame);
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
// skip fifo
return Ok(Some(out_frame));
let ret = av_audio_fifo_write(
self.fifo,
(*out_frame).extended_data as *const *mut libc::c_void,
(*out_frame).nb_samples,
);
if ret < 0 {
av_frame_free(&mut out_frame);
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
if ret != (*out_frame).nb_samples {
av_frame_free(&mut out_frame);
return Err(Error::msg(format!(
"FIFO write {} != {}",
ret,
(*out_frame).nb_samples
)));
}
//info!("Resampled {}->{} (wrote={})", in_samples, (*out_frame).nb_samples, ret);
av_frame_free(&mut out_frame);
let buff = av_audio_fifo_size(self.fifo);
if buff < (*self.ctx).frame_size {
Ok(None)
} else {
let out_frame = self.read_fifo_frame()?;
Ok(Some(out_frame))
}
}
unsafe fn read_fifo_frame(&mut self) -> Result<*mut AVFrame, Error> {
let mut out_frame = self.new_frame();
let ret = av_samples_alloc_array_and_samples(
&mut (*out_frame).extended_data,
ptr::null_mut(),
(*out_frame).ch_layout.nb_channels,
(*out_frame).nb_samples,
transmute((*out_frame).format),
0,
);
if ret < 0 {
av_frame_free(&mut out_frame);
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
let ret = av_audio_fifo_read(
self.fifo,
(*out_frame).extended_data as *const *mut libc::c_void,
(*out_frame).nb_samples,
);
if ret < 0 {
av_frame_free(&mut out_frame);
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
assert_eq!(
ret,
(*out_frame).nb_samples,
"Read wrong number of samples from FIFO"
);
Ok(out_frame)
}
unsafe fn new_frame(&self) -> *mut AVFrame {
let out_frame = av_frame_alloc();
(*out_frame).nb_samples = (*self.ctx).frame_size;
av_channel_layout_copy(&mut (*out_frame).ch_layout, &(*self.ctx).ch_layout);
(*out_frame).format = (*self.ctx).sample_fmt as libc::c_int;
(*out_frame).sample_rate = (*self.ctx).sample_rate;
out_frame
}
unsafe fn process_frame(
&mut self,
frame: *mut AVFrame,
src: &AVFrameSource,
) -> Result<(), Error> {
let in_stream = match src {
AVFrameSource::Decoder(s) => *s,
AVFrameSource::Scaler(s) => *s,
};
self.setup_encoder(frame)?;
let frame = self.process_audio_frame(frame)?;
if frame.is_none() {
return Ok(());
}
let mut frame = frame.unwrap();
(*frame).pts = self.frame_pts;
self.frame_pts += (*frame).nb_samples as i64;
let mut ret = avcodec_send_frame(self.ctx, frame);
if ret < 0 && ret != AVERROR(EAGAIN) {
av_frame_free(&mut frame);
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
while ret > 0 || ret == AVERROR(EAGAIN) {
let mut pkt = av_packet_alloc();
ret = avcodec_receive_packet(self.ctx, pkt);
if ret < 0 {
av_frame_free(&mut frame);
av_packet_free(&mut pkt);
if ret == AVERROR(EAGAIN) {
return Ok(());
}
return Err(Error::msg(get_ffmpeg_error_msg(ret)));
}
set_encoded_pkt_timing(self.ctx, pkt, in_stream, &mut self.pts, &self.variant);
self.chan_out.send(PipelinePayload::AvPacket(
pkt,
AVPacketSource::Encoder(VariantStream::Audio(self.variant.clone())),
))?;
}
av_frame_free(&mut frame);
Ok(())
}
}
impl<TRecv> PipelineProcessor for AudioEncoder<TRecv>
where
TRecv: Rx<PipelinePayload>,
{
fn process(&mut self) -> Result<(), Error> {
while let Ok(pkg) = self.chan_in.try_recv_next() {
match pkg {
PipelinePayload::AvFrame(frm, ref src) => unsafe {
let idx = match src {
AVFrameSource::Decoder(s) => (**s).index,
_ => {
return Err(Error::msg(format!("Cannot process frame from: {:?}", src)))
}
};
if self.variant.src_index == idx as usize {
self.process_frame(frm, src)?;
}
},
_ => return Err(Error::msg("Payload not supported")),
}
}
Ok(())
}
}