audio + video player in groupsync

cmd/mxl-player/main.go

@@ -25,19 +25,22 @@ const (
 )
 
 func main() {
-	videoFlowId := "5fbec3b1-1b0f-417d-9059-8b94a47197ed"
-	audioFlowId := "5fbec3b1-1b0f-417d-9059-8b94a47197ec"
+	// video: 5fbec3b1-1b0f-417d-9059-8b94a47197ed
+	// audio: 5fbec3b1-1b0f-417d-9059-8b94a47197ec
+	var videoFlowId, audioFlowId string
 	mxlDomain := flag.String("d", "/dev/shm/mxl", "MXL domain")
 	mxlVideoFlowID := flag.String("v", videoFlowId, "MXL video flow UUID")
 	mxlAudioFlowID := flag.String("a", audioFlowId, "MXL audio flow UUID")
-	fmt.Println(*mxlAudioFlowID)
 	flag.Parse()
+	if *mxlVideoFlowID == "" && *mxlAudioFlowID == "" {
+		log.Fatal("need at least -v <video-flow> or -a <audio-flow>")
+	}
 
 	runtime.LockOSThread()
 	if err := sdl.Load(); err != nil {
 		panic(err)
 	}
-	if !sdl.Init(sdl.InitVideo) {
+	if !sdl.Init(sdl.InitVideo | sdl.InitAudio) {
 		log.Fatalf("SDL_Init: %s", sdl.GetError())
 		return
 	}
@@ -108,29 +111,123 @@ func main() {
 	}
 	defer vkDevice.Destroy()
 
-	mxlSrc, err := source.Open(*mxlDomain, *mxlVideoFlowID)
-	if err != nil {
-		log.Fatalf("source: %v", err)
-	}
-	defer func() { _ = mxlSrc.Close() }()
-	fmt.Printf("source: %dx%d stride=%d\n", mxlSrc.Width(), mxlSrc.Height(), mxlSrc.Stride())
+	var (
+		syncSrc     *source.SyncSource
+		videoSrc    *source.Source
+		audioSrc    *source.AudioSource
+		audioStream uintptr
+		audioBatch  uint64
+		aChans      uint64
+	)
 
-	r, err := renderer.New(renderer.Config{
-		PhysDevice:     vkPhysDevice,
-		Device:         vkDevice,
-		Queue:          vkQueue,
-		Surface:        vkSurf,
-		Window:         windowHandler,
-		GraphicsFamily: gfx,
-		VideoWidth:     mxlSrc.Width(),
-		VideoHeight:    mxlSrc.Height(),
-		VideoStride:    mxlSrc.Stride(),
-	})
-	if err != nil {
-		panic(err)
+	interleaveAudio := func(samples [][]byte) []byte {
+		frameBytes := int(audioBatch) * int(aChans) * 4
+		out := make([]byte, frameBytes)
+		for ch := uint64(0); ch < aChans; ch++ {
+			srcBytes := samples[ch]
+			for i := uint64(0); i < audioBatch; i++ {
+				srcOff := i * 4
+				dstOff := (i*aChans + ch) * 4
+				if srcOff+4 <= uint64(len(srcBytes)) {
+					copy(out[dstOff:dstOff+4], srcBytes[srcOff:srcOff+4])
+				}
+			}
+		}
+		return out
+	}
+
+	switch {
+	case *mxlVideoFlowID != "" && *mxlAudioFlowID != "":
+		syncSrc, err = source.OpenSync(*mxlDomain, *mxlVideoFlowID, *mxlAudioFlowID)
+		if err != nil {
+			log.Fatalf("sync source: %v", err)
+		}
+		aChans = syncSrc.Channels()
+		audioBatch = uint64(syncSrc.AudioRate().Num) / uint64(syncSrc.Rate().Num)
+		if audioBatch == 0 {
+			audioBatch = 1
+		}
+		audioStream = sdl.OpenAudioDeviceStream(sdl.AudioDeviceDefaultPlayback, sdl.AudioSpec{
+			Format:   sdl.AudioF32,
+			Channels: int32(aChans),
+			Freq:     int32(syncSrc.AudioRate().Num / syncSrc.AudioRate().Den),
+		})
+		if audioStream == 0 {
+			log.Fatalf("audio: %s", sdl.GetError())
+		}
+		sdl.ResumeAudioStreamDevice(audioStream)
+		fmt.Printf("sync: video %dx%d audio %dch batch=%d\n",
+			syncSrc.Width(), syncSrc.Height(), aChans, audioBatch)
+
+	case *mxlVideoFlowID != "":
+		videoSrc, err = source.Open(*mxlDomain, *mxlVideoFlowID)
+		if err != nil {
+			log.Fatalf("source: %v", err)
+		}
+		fmt.Printf("video: %dx%d stride=%d\n", videoSrc.Width(), videoSrc.Height(), videoSrc.Stride())
+
+	default:
+		audioSrc, err = source.OpenAudio(*mxlDomain, *mxlAudioFlowID)
+		if err != nil {
+			log.Fatalf("audio source: %v", err)
+		}
+		aChans = audioSrc.Channels()
+		audioBatch = uint64(audioSrc.Rate().Num) / (100 * uint64(audioSrc.Rate().Den))
+		if audioBatch == 0 {
+			audioBatch = 1
+		}
+		audioStream = sdl.OpenAudioDeviceStream(sdl.AudioDeviceDefaultPlayback, sdl.AudioSpec{
+			Format:   sdl.AudioF32,
+			Channels: int32(aChans),
+			Freq:     int32(audioSrc.Rate().Num / audioSrc.Rate().Den),
+		})
+		if audioStream == 0 {
+			log.Fatalf("audio: %s", sdl.GetError())
+		}
+		sdl.ResumeAudioStreamDevice(audioStream)
+		fmt.Printf("audio: %dch %d/%d Hz\n", aChans, audioSrc.Rate().Num, audioSrc.Rate().Den)
+	}
+
+	defer func() {
+		if syncSrc != nil {
+			_ = syncSrc.Close()
+		}
+		if videoSrc != nil {
+			_ = videoSrc.Close()
+		}
+		if audioSrc != nil {
+			_ = audioSrc.Close()
+		}
+	}()
+	if audioStream != 0 {
+		defer sdl.DestroyAudioStream(audioStream)
+	}
+
+	var r *renderer.Renderer
+	if *mxlVideoFlowID != "" {
+		var w, h, stride uint32
+		if syncSrc != nil {
+			w, h, stride = syncSrc.Width(), syncSrc.Height(), syncSrc.Stride()
+		} else {
+			w, h, stride = videoSrc.Width(), videoSrc.Height(), videoSrc.Stride()
+		}
+		r, err = renderer.New(renderer.Config{
+			PhysDevice:     vkPhysDevice,
+			Device:         vkDevice,
+			Queue:          vkQueue,
+			Surface:        vkSurf,
+			Window:         windowHandler,
+			GraphicsFamily: gfx,
+			VideoWidth:     w,
+			VideoHeight:    h,
+			VideoStride:    stride,
+		})
+		if err != nil {
+			panic(err)
+		}
+		defer r.Destroy()
+		defer vkDevice.WaitIdle()
 	}
-	defer r.Destroy()
-	defer vkDevice.WaitIdle()
 
 	ctx, cancel := context.WithCancel(context.Background())
 	defer cancel()
@@ -139,8 +236,14 @@ func main() {
 	failed := make(chan struct{})
 
 	reopen := func() error {
-		if err := mxlSrc.Close(); err != nil {
-			return fmt.Errorf("close old source: %w", err)
+		if syncSrc != nil {
+			_ = syncSrc.Close()
+		}
+		if videoSrc != nil {
+			_ = videoSrc.Close()
+		}
+		if audioSrc != nil {
+			_ = audioSrc.Close()
 		}
 		for {
 			select {
@@ -148,56 +251,166 @@ func main() {
 				return ctx.Err()
 			default:
 			}
-			s, err := source.Open(*mxlDomain, *mxlVideoFlowID)
-			if err == nil {
-				newSize := vk.DeviceSize(s.Stride()) * vk.DeviceSize(s.Height())
-				if newSize != r.FrameSize() {
-					if err := r.RecreateBuffers(newSize); err != nil {
-						return err
+			if syncSrc != nil {
+				s, e := source.OpenSync(*mxlDomain, *mxlVideoFlowID, *mxlAudioFlowID)
+				if e == nil {
+					if r != nil {
+						newSize := vk.DeviceSize(s.Stride()) * vk.DeviceSize(s.Height())
+						if newSize != r.FrameSize() {
+							if e = r.RecreateBuffers(newSize); e != nil {
+								return e
+							}
+						}
 					}
-					fmt.Printf("source: resolution changed, frameSize=%d\n", newSize)
+					syncSrc = s
+					return nil
 				}
-				mxlSrc = s
-				return nil
+				log.Printf("source: reopen retry: %v", e)
+			} else if videoSrc != nil {
+				s, e := source.Open(*mxlDomain, *mxlVideoFlowID)
+				if e == nil {
+					if r != nil {
+						newSize := vk.DeviceSize(s.Stride()) * vk.DeviceSize(s.Height())
+						if newSize != r.FrameSize() {
+							if e = r.RecreateBuffers(newSize); e != nil {
+								return e
+							}
+						}
+					}
+					videoSrc = s
+					return nil
+				}
+				log.Printf("source: reopen retry: %v", e)
+			} else if audioSrc != nil {
+				s, e := source.OpenAudio(*mxlDomain, *mxlAudioFlowID)
+				if e == nil {
+					audioSrc = s
+					return nil
+				}
+				log.Printf("source: reopen retry: %v", e)
 			}
-			log.Printf("source: reopen retry: %v", err)
 			time.Sleep(500 * time.Millisecond)
 		}
 	}
 
 	go func() {
+		// Audio-only mode: independent loop, no grant/staged handshake.
+		if audioSrc != nil && syncSrc == nil && videoSrc == nil {
+			for {
+				select {
+				case <-ctx.Done():
+					return
+				default:
+				}
+				queued := sdl.GetAudioStreamQueued(audioStream)
+				maxQueued := int32(audioBatch) * int32(aChans) * 4 * 20
+				if queued > maxQueued {
+					select {
+					case <-time.After(10 * time.Millisecond):
+					case <-ctx.Done():
+						return
+					}
+					continue
+				}
+				f, err := audioSrc.NextAudio(ctx, audioBatch, 20*time.Millisecond)
+				if err != nil {
+					if errors.Is(err, context.Canceled) {
+						return
+					}
+					if errors.Is(err, mxl.ErrFlowInvalid) {
+						log.Printf("source: flow invalid, reopening")
+						if rerr := reopen(); rerr != nil {
+							log.Printf("source: reopen failed: %v", rerr)
+							cancel()
+							return
+						}
+						continue
+					}
+					log.Printf("source: %v", err)
+					cancel()
+					return
+				}
+				if f.Samples != nil && audioStream != 0 {
+					sdl.PutAudioStreamData(audioStream, interleaveAudio(f.Samples))
+				}
+			}
+		}
+
+		// Video (with or without sync) mode: grant/staged handshake.
 		for {
 			select {
 			case <-grant:
 			case <-ctx.Done():
 				return
 			}
-			f, err := mxlSrc.NextCtx(ctx, 200*time.Millisecond)
-			if err != nil {
-				if errors.Is(err, context.Canceled) {
+
+			var payload []byte
+			var grainIdx uint64
+
+			if syncSrc != nil {
+				vFrame, aFrame, err := syncSrc.NextSync(ctx, audioBatch, 200*time.Millisecond)
+				if err != nil {
+					if errors.Is(err, context.Canceled) {
+						return
+					}
+					if errors.Is(err, mxl.ErrFlowInvalid) {
+						log.Printf("source: flow invalid, reopening")
+						if rerr := reopen(); rerr != nil {
+							log.Printf("source: reopen failed: %v", rerr)
+							cancel()
+							return
+						}
+						select {
+						case failed <- struct{}{}:
+						case <-ctx.Done():
+							return
+						}
+						continue
+					}
+					log.Printf("source: %v", err)
+					cancel()
 					return
 				}
-				if errors.Is(err, mxl.ErrFlowInvalid) {
-					log.Printf("source: flow invalid, reopening")
-					if rerr := reopen(); rerr != nil {
-						log.Printf("source: reopen failed: %v", rerr)
-						cancel()
-						return
-					}
-					select {
-					case failed <- struct{}{}:
-					case <-ctx.Done():
-						return
-					}
-					continue
+				payload = vFrame.Payload
+				grainIdx = vFrame.Index
+				if aFrame.Samples != nil && audioStream != 0 {
+					sdl.PutAudioStreamData(audioStream, interleaveAudio(aFrame.Samples))
 				}
-				log.Printf("source: %v", err)
-				cancel()
-				return
+
+			} else if videoSrc != nil {
+				f, err := videoSrc.NextCtx(ctx, 200*time.Millisecond)
+				if err != nil {
+					if errors.Is(err, context.Canceled) {
+						return
+					}
+					if errors.Is(err, mxl.ErrFlowInvalid) {
+						log.Printf("source: flow invalid, reopening")
+						if rerr := reopen(); rerr != nil {
+							log.Printf("source: reopen failed: %v", rerr)
+							cancel()
+							return
+						}
+						select {
+						case failed <- struct{}{}:
+						case <-ctx.Done():
+							return
+						}
+						continue
+					}
+					log.Printf("source: %v", err)
+					cancel()
+					return
+				}
+				payload = f.Payload
+				grainIdx = f.Index
+
+			}
+
+			if r != nil {
+				vk.CopyToMapped(r.StagingMapped(), payload)
 			}
-			vk.CopyToMapped(r.StagingMapped(), f.Payload)
 			select {
-			case staged <- f.Index:
+			case staged <- grainIdx:
 			case <-ctx.Done():
 				return
 			}
@@ -239,7 +452,7 @@ func main() {
 		if !running {
 			break
 		}
-		if resized {
+		if resized && r != nil {
 			if err := r.RecreateSwapchain(); err != nil {
 				if errors.Is(err, renderer.ErrMinimized) {
 					resized = true
@@ -272,36 +485,46 @@ func main() {
 			continue
 		}
 
-		err := r.DrawFrame(mxlSrc.Width(), mxlSrc.Height(), mxlSrc.Stride())
-		if errors.Is(err, renderer.ErrOutOfDate) {
-			if rerr := r.RecreateSwapchain(); rerr != nil {
-				if errors.Is(rerr, renderer.ErrMinimized) {
-					resized = true
-					continue
+		if r != nil {
+			var w, h, stride uint32
+			if syncSrc != nil {
+				w, h, stride = syncSrc.Width(), syncSrc.Height(), syncSrc.Stride()
+			} else if videoSrc != nil {
+				w, h, stride = videoSrc.Width(), videoSrc.Height(), videoSrc.Stride()
+			}
+			err := r.DrawFrame(w, h, stride)
+			if errors.Is(err, renderer.ErrOutOfDate) {
+				if rerr := r.RecreateSwapchain(); rerr != nil {
+					if errors.Is(rerr, renderer.ErrMinimized) {
+						resized = true
+						continue
+					}
+					panic(rerr)
 				}
-				panic(rerr)
+				continue
+			}
+			if err != nil {
+				panic(err)
 			}
-			continue
-		}
-		if err != nil {
-			panic(err)
-		}
 
-		if lastIndex != 0 && shownIndex > lastIndex {
-			if g := shownIndex - lastIndex - 1; g > 0 {
-				dropped += g
+			if lastIndex != 0 && shownIndex > lastIndex {
+				if g := shownIndex - lastIndex - 1; g > 0 {
+					dropped += g
+				}
 			}
-		}
-		lastIndex = shownIndex
-		frameCount++
-		if now := time.Now(); now.Sub(lastReport) >= time.Second {
-			dt := now.Sub(lastReport).Seconds()
-			fps := float64(frameCount) / dt
-			fmt.Printf("fps=%.1f dropped=%d idx=%d frameTime=%.2fms\n",
-				fps, dropped, shownIndex, float64(now.Sub(frameStart).Microseconds())/1000.0)
-			frameCount = 0
-			dropped = 0
-			lastReport = now
+			lastIndex = shownIndex
+			frameCount++
+			if now := time.Now(); now.Sub(lastReport) >= time.Second {
+				dt := now.Sub(lastReport).Seconds()
+				fps := float64(frameCount) / dt
+				fmt.Printf("fps=%.1f dropped=%d idx=%d frameTime=%.2fms\n",
+					fps, dropped, shownIndex, float64(now.Sub(frameStart).Microseconds())/1000.0)
+				frameCount = 0
+				dropped = 0
+				lastReport = now
+			}
+		} else {
+			time.Sleep(10 * time.Millisecond)
 		}
 	}
 }

internal/source/source.go

@@ -71,6 +71,11 @@ func Open(domain, flowID string) (*Source, error) {
 		inst.Close()
 		return nil, fmt.Errorf("parse flow def: %w", err)
 	}
+	if fd.FrameWidth == 0 || fd.FrameHeight == 0 {
+		r.Close()
+		inst.Close()
+		return nil, fmt.Errorf("flow has no video dimensions (not a video flow?)")
+	}
 	rate := info.Config.Common.GrainRate
 	idx := mxl.CurrentIndex(rate)
 	if idx == mxl.UndefinedIndex {
@@ -270,3 +275,199 @@ func (s *AudioSource) Close() error {
 
 func (s *AudioSource) Rate() mxl.Rational { return s.rate }
 func (s *AudioSource) Channels() uint64   { return s.chans }
+
+type SyncSource struct {
+	inst                  *mxl.Instance
+	vr                    *mxl.Reader
+	ar                    *mxl.Reader
+	group                 *mxl.SyncGroup
+	rate                  mxl.Rational // video rate
+	aRate                 mxl.Rational
+	chans                 uint64
+	idx                   uint64
+	width, height, stride uint32
+}
+
+func OpenSync(domain, videoFlow, audioFlow string) (*SyncSource, error) {
+	inst, err := mxl.NewInstance(domain, "")
+	if err != nil {
+		return nil, fmt.Errorf("NewInstance: %w", err)
+	}
+
+	vr, err := inst.NewReader(videoFlow)
+	if err != nil {
+		inst.Close()
+		return nil, fmt.Errorf("NewReader(video): %w", err)
+	}
+	ar, err := inst.NewReader(audioFlow)
+	if err != nil {
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("NewReader(audio): %w", err)
+	}
+
+	vInfo, err := vr.Info()
+	if err != nil {
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("Info(video): %w", err)
+	}
+	if !vInfo.Config.Common.Format.IsDiscrete() {
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("video flow is not discrete")
+	}
+	aInfo, err := ar.Info()
+	if err != nil {
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("Info(audio): %w", err)
+	}
+	if aInfo.Config.Common.Format.IsDiscrete() {
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("audio flow is not continuous")
+	}
+
+	def, err := inst.FlowDef(videoFlow)
+	if err != nil {
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("FlowDef: %w", err)
+	}
+	var fd flowDef
+	if err := json.Unmarshal([]byte(def), &fd); err != nil {
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("parse flow def: %w", err)
+	}
+
+	vRate := vInfo.Config.Common.GrainRate
+	idx := mxl.CurrentIndex(vRate)
+	if idx == mxl.UndefinedIndex {
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("invalid video rate %d/%d", vRate.Num, vRate.Den)
+	}
+
+	group, err := inst.NewSyncGroup()
+	if err != nil {
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("NewSyncGroup: %w", err)
+	}
+	if err := group.AddReader(vr); err != nil {
+		group.Close()
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("AddReader(video): %w", err)
+	}
+	if err := group.AddReader(ar); err != nil {
+		group.Close()
+		ar.Close()
+		vr.Close()
+		inst.Close()
+		return nil, fmt.Errorf("AddReader(audio): %w", err)
+	}
+
+	return &SyncSource{
+		inst:   inst,
+		vr:     vr,
+		ar:     ar,
+		group:  group,
+		rate:   vRate,
+		aRate:  aInfo.Config.Common.GrainRate,
+		chans:  uint64(aInfo.Config.Continuous.ChannelCount),
+		idx:    idx,
+		width:  uint32(fd.FrameWidth),
+		height: uint32(fd.FrameHeight),
+		stride: vInfo.Config.Discrete.SliceSizes[0],
+	}, nil
+}
+
+func (s *SyncSource) Close() error {
+	_ = s.group.Close()
+	_ = s.ar.Close()
+	_ = s.vr.Close()
+	return s.inst.Close()
+}
+
+// NextSync reads both at a synced timestamp. Returns video Frame + audio AudioFrame
+func (s *SyncSource) NextSync(ctx context.Context, audioBatch uint64, timeout time.Duration) (Frame, AudioFrame, error) {
+	for {
+		select {
+		case <-ctx.Done():
+			return Frame{}, AudioFrame{}, ctx.Err()
+		default:
+		}
+
+		ts := mxl.IndexToTimestamp(s.rate, s.idx)
+		err := s.group.WaitForDataAt(ts, timeout)
+		switch {
+		case err == nil:
+			// read video
+			g, gerr := s.vr.GetGrain(s.idx, 50*time.Millisecond)
+			if gerr != nil {
+				if errors.Is(gerr, mxl.ErrFlowInvalid) {
+					s.idx = mxl.CurrentIndex(s.rate)
+					continue
+				}
+				return Frame{}, AudioFrame{}, fmt.Errorf("GetGraing: %w", gerr)
+			}
+			// read audio at the same timestamp
+			aIdx := mxl.TimestampToIndex(s.aRate, ts)
+			av, aerr := s.ar.GetSamples(aIdx, int(audioBatch), 50*time.Millisecond)
+			vFrame := Frame{
+				Index: g.Index, Width: s.width, Height: s.height,
+				Stride: s.stride, Size: g.GrainSize,
+				Invalid: g.Invalid(), Payload: g.Payload,
+			}
+			s.idx++
+
+			var aFrame AudioFrame
+			if aerr == nil {
+				samples := make([][]byte, s.chans)
+				for ch := uint64(0); ch < s.chans; ch++ {
+					f1, f2, _ := av.ChannelFragments(ch)
+					if len(f2) > 0 {
+						samples[ch] = append(f1, f2...)
+					} else {
+						samples[ch] = f1
+					}
+				}
+				aFrame = AudioFrame{
+					Index: aIdx, SampleCount: audioBatch,
+					Channels: s.chans, Samples: samples,
+				}
+			}
+			// even if audio failed, video returns
+			return vFrame, aFrame, nil
+		case errors.Is(err, mxl.ErrTimeout), errors.Is(err, mxl.ErrOutOfRangeEarly):
+			select {
+			case <-time.After(5 * time.Millisecond):
+			case <-ctx.Done():
+				return Frame{}, AudioFrame{}, ctx.Err()
+			}
+		case errors.Is(err, mxl.ErrOutOfRangeLate):
+			s.idx = mxl.CurrentIndex(s.rate)
+		default:
+			return Frame{}, AudioFrame{}, fmt.Errorf("WaitForDataAt: %w", err)
+		}
+	}
+}
+
+func (s *SyncSource) Width() uint32           { return s.width }
+func (s *SyncSource) Height() uint32          { return s.height }
+func (s *SyncSource) Stride() uint32          { return s.stride }
+func (s *SyncSource) Rate() mxl.Rational      { return s.rate }
+func (s *SyncSource) AudioRate() mxl.Rational { return s.aRate }
+func (s *SyncSource) Channels() uint64        { return s.chans }