Scalable video coding (SVC).

This commit is contained in:
Juliusz Chroboczek
2021-05-11 15:28:30 +02:00
parent 7590588a54
commit 6f9d7fc306
13 changed files with 660 additions and 39 deletions
+99
View File
@@ -1,6 +1,7 @@
package rtpconn
import (
"errors"
"strings"
"github.com/pion/rtp"
@@ -108,3 +109,101 @@ func isKeyframe(codec string, packet *rtp.Packet) (bool, bool) {
return false, false
}
}
var errTruncated = errors.New("truncated packet")
var errUnsupportedCodec = errors.New("unsupported codec")
func packetFlags(codec string, buf []byte) (seqno uint16, start bool, pid uint16, tid uint8, sid uint8, layersync bool, discardable bool, err error) {
if len(buf) < 12 {
err = errTruncated
return
}
seqno = (uint16(buf[2]) << 8) | uint16(buf[3])
if strings.EqualFold(codec, "video/vp8") {
var packet rtp.Packet
err = packet.Unmarshal(buf)
if err != nil {
return
}
var vp8 codecs.VP8Packet
_, err = vp8.Unmarshal(packet.Payload)
if err != nil {
return
}
start = vp8.S == 1 && vp8.PID == 0
pid = vp8.PictureID
tid = vp8.TID
layersync = vp8.Y == 1
discardable = vp8.N == 1
return
} else if strings.EqualFold(codec, "video/vp9") {
var packet rtp.Packet
err = packet.Unmarshal(buf)
if err != nil {
return
}
var vp9 codecs.VP9Packet
_, err = vp9.Unmarshal(packet.Payload)
if err != nil {
return
}
start = vp9.B
tid = vp9.TID
sid = vp9.SID
layersync = vp9.U
return
}
return
}
func rewritePacket(codec string, data []byte, seqno uint16, delta uint16) error {
if len(data) < 12 {
return errTruncated
}
data[2] = uint8(seqno >> 8)
data[3] = uint8(seqno)
if delta == 0 {
return nil
}
offset := 12
offset += int(data[0]&0x0F) * 4
if len(data) < offset+4 {
return errTruncated
}
if (data[0] & 0x10) != 0 {
length := uint16(data[offset+2])<<8 | uint16(data[offset+3])
offset += 4 + int(length)*4
if len(data) < offset+4 {
return errTruncated
}
}
if strings.EqualFold(codec, "video/vp8") {
x := (data[offset] & 0x80) != 0
if !x {
return nil
}
i := (data[offset+1] & 0x80) != 0
if !i {
return nil
}
m := (data[offset+2] & 0x80) != 0
if m {
pid := (uint16(data[offset+2]&0x7F) << 8) |
uint16(data[offset+3])
pid = (pid + delta) & 0x7FFF
data[offset+2] = 0x80 | byte((pid>>8)&0x7F)
data[offset+3] = byte(pid & 0xFF)
} else {
data[offset+2] = (data[offset+2] + uint8(delta)) & 0x7F
}
return nil
}
return errUnsupportedCodec
}
+148 -24
View File
@@ -20,6 +20,7 @@ import (
"github.com/jech/galene/ice"
"github.com/jech/galene/jitter"
"github.com/jech/galene/packetcache"
"github.com/jech/galene/packetmap"
"github.com/jech/galene/rtptime"
)
@@ -74,6 +75,7 @@ type downTrackAtomics struct {
srNTP uint64
remoteNTP uint64
remoteRTP uint32
layerInfo uint32
}
type rtpDownTrack struct {
@@ -81,6 +83,7 @@ type rtpDownTrack struct {
sender *webrtc.RTPSender
remote conn.UpTrack
ssrc webrtc.SSRC
packetmap packetmap.Map
maxBitrate *bitrate
maxREMBBitrate *bitrate
rate *estimator.Estimator
@@ -89,14 +92,6 @@ type rtpDownTrack struct {
cname atomic.Value
}
func (down *rtpDownTrack) Write(buf []byte) (int, error) {
n, err := down.track.Write(buf)
if err == nil {
down.rate.Accumulate(uint32(n))
}
return n, err
}
func (down *rtpDownTrack) SetTimeOffset(ntp uint64, rtp uint32) {
atomic.StoreUint64(&down.atomics.remoteNTP, ntp)
atomic.StoreUint32(&down.atomics.remoteRTP, rtp)
@@ -131,6 +126,17 @@ func (down *rtpDownTrack) SetCname(cname string) {
down.cname.Store(cname)
}
func (down *rtpDownTrack) getLayerInfo() (uint8, uint8, uint8) {
info := atomic.LoadUint32(&down.atomics.layerInfo)
return uint8(info >> 16), uint8(info >> 8), uint8(info)
}
func (down *rtpDownTrack) setLayerInfo(layer, wanted, max uint8) {
atomic.StoreUint32(&down.atomics.layerInfo,
(uint32(layer)<<16)|(uint32(wanted)<<8)|uint32(max),
)
}
const (
negotiationUnneeded = iota
negotiationNeeded
@@ -179,17 +185,108 @@ func newDownConn(c group.Client, id string, remote conn.Up) (*rtpDownConnection,
return conn, nil
}
func (t *rtpDownTrack) GetMaxBitrate() uint64 {
var packetBufPool = sync.Pool{
New: func() interface{} {
return make([]byte, packetcache.BufSize)
},
}
func (down *rtpDownTrack) Write(buf []byte) (int, error) {
codec := down.remote.Codec().MimeType
seqno, start, pid, tid, _, u, _, err := packetFlags(codec, buf)
if err != nil {
return 0, err
}
layer, wantedLayer, maxLayer := down.getLayerInfo()
if tid > maxLayer {
if layer == maxLayer {
wantedLayer = tid
layer = tid
}
maxLayer = tid
if wantedLayer > maxLayer {
wantedLayer = maxLayer
}
down.setLayerInfo(layer, wantedLayer, maxLayer)
down.adjustLayer()
}
if start && layer != wantedLayer {
if u || wantedLayer < layer {
layer = wantedLayer
down.setLayerInfo(layer, wantedLayer, maxLayer)
}
}
if tid > layer {
ok := down.packetmap.Drop(seqno, pid)
if ok {
return 0, nil
}
}
ok, newseqno, piddelta := down.packetmap.Map(seqno, pid)
if !ok {
return 0, nil
}
if newseqno == seqno && piddelta == 0 {
return down.write(buf)
}
ibuf2 := packetBufPool.Get()
defer packetBufPool.Put(ibuf2)
buf2 := ibuf2.([]byte)
n := copy(buf2, buf)
err = rewritePacket(codec, buf2[:n], newseqno, piddelta)
if err != nil {
return 0, err
}
return down.write(buf2[:n])
}
func (down *rtpDownTrack) write(buf []byte) (int, error) {
n, err := down.track.Write(buf)
if err == nil {
down.rate.Accumulate(uint32(n))
}
return n, err
}
func (t *rtpDownTrack) GetMaxBitrate() (uint64, int) {
now := rtptime.Jiffies()
layer, _, _ := t.getLayerInfo()
r := t.maxBitrate.Get(now)
if r == ^uint64(0) {
r = 512 * 1024
}
rr := t.maxREMBBitrate.Get(now)
if rr == 0 || r < rr {
return r
return r, int(layer)
}
return rr, int(layer)
}
func (t *rtpDownTrack) adjustLayer() {
max, _ := t.GetMaxBitrate()
r, _ := t.rate.Estimate()
rate := uint64(r) * 8
if rate < max*7/8 {
layer, wanted, max := t.getLayerInfo()
if layer < max {
wanted = layer + 1
t.setLayerInfo(layer, wanted, max)
}
} else if rate > max*3/2 {
layer, wanted, max := t.getLayerInfo()
if layer > 0 {
wanted = layer - 1
t.setLayerInfo(layer, wanted, max)
}
}
return rr
}
func (down *rtpDownConnection) addICECandidate(candidate *webrtc.ICECandidateInit) error {
@@ -240,6 +337,7 @@ type rtpUpTrack struct {
srTime uint64
srNTPTime uint64
srRTPTime uint32
maxLayer uint8
local []conn.DownTrack
bufferedNACKs []uint16
}
@@ -598,7 +696,11 @@ func gotNACK(conn *rtpDownConnection, track *rtpDownTrack, p *rtcp.TransportLaye
var packet rtp.Packet
buf := make([]byte, packetcache.BufSize)
for _, nack := range p.Nacks {
nack.Range(func(seqno uint16) bool {
nack.Range(func(s uint16) bool {
ok, seqno, _ := track.packetmap.Reverse(s)
if !ok {
return true
}
l := track.remote.GetRTP(seqno, buf)
if l == 0 {
unhandled = append(unhandled, seqno)
@@ -785,28 +887,44 @@ func sendUpRTCP(up *rtpUpConnection) error {
continue
}
ssrcs = append(ssrcs, uint32(t.track.SSRC()))
var r uint64
if t.Kind() == webrtc.RTPCodecTypeAudio {
r = 100 * 1024
rate += 100 * 1024
} else if t.Label() == "l" {
r = group.LowBitrate
rate += group.LowBitrate
} else {
minrate := ^uint64(0)
maxrate := uint64(group.MinBitrate)
maxlayer := 0
local := t.getLocal()
r = ^uint64(0)
for _, down := range local {
rr := down.GetMaxBitrate()
if rr < group.MinBitrate {
rr = group.MinBitrate
r, l := down.GetMaxBitrate()
if maxlayer < l {
maxlayer = l
}
if r > rr {
r = rr
if r < group.MinBitrate {
r = group.MinBitrate
}
if minrate > r {
minrate = r
}
if maxrate < r {
maxrate = r
}
}
if r == ^uint64(0) {
r = 512 * 1024
// assume that each layer takes two times less
// throughput than the higher one. Then we've
// got enough slack for a factor of 2^(layers-1).
for i := 0; i < maxlayer; i++ {
if minrate < ^uint64(0)/2 {
minrate *= 2
}
}
if minrate < maxrate {
rate += minrate
} else {
rate += maxrate
}
}
rate += r
}
if rate < ^uint64(0) && len(ssrcs) > 0 {
@@ -968,6 +1086,7 @@ func rtcpDownListener(conn *rtpDownConnection, track *rtpDownTrack, s *webrtc.RT
continue
}
adjust := false
jiffies := rtptime.Jiffies()
for _, p := range ps {
@@ -994,10 +1113,12 @@ func rtcpDownListener(conn *rtpDownConnection, track *rtpDownTrack, s *webrtc.RT
}
case *rtcp.ReceiverEstimatedMaximumBitrate:
track.maxREMBBitrate.Set(p.Bitrate, jiffies)
adjust = true
case *rtcp.ReceiverReport:
for _, r := range p.Reports {
if r.SSRC == uint32(track.ssrc) {
handleReport(track, r, jiffies)
adjust = true
}
}
case *rtcp.SenderReport:
@@ -1010,6 +1131,9 @@ func rtcpDownListener(conn *rtpDownConnection, track *rtpDownTrack, s *webrtc.RT
gotNACK(conn, track, p)
}
}
if adjust {
track.adjustLayer()
}
}
}
+4 -1
View File
@@ -46,9 +46,11 @@ func (c *webClient) GetStats() *stats.Client {
jiffies := rtptime.Jiffies()
for _, down := range c.down {
conns := stats.Conn{
Id: down.id,
Id: down.id,
}
for _, t := range down.tracks {
l, _, _ := t.getLayerInfo()
layer := int(l)
rate, _ := t.rate.Estimate()
rtt := rtptime.ToDuration(t.getRTT(),
rtptime.JiffiesPerSec)
@@ -56,6 +58,7 @@ func (c *webClient) GetStats() *stats.Client {
j := time.Duration(jitter) * time.Second /
time.Duration(t.track.Codec().ClockRate)
conns.Tracks = append(conns.Tracks, stats.Track{
Layer: &layer,
Bitrate: uint64(rate) * 8,
MaxBitrate: t.maxBitrate.Get(jiffies),
Loss: float64(loss) / 256.0,
+3
View File
@@ -1270,6 +1270,9 @@ func handleClientMessage(c *webClient, m clientMessage) error {
return closeDownConn(c, m.Id, message)
}
down := getDownConn(c, m.Id)
if down == nil {
return ErrUnknownId
}
if down.negotiationNeeded > negotiationUnneeded {
err := negotiate(
c, down,