Measure backend playback latency outside audio callbacks

tools/chatgui/transport/audio_diagnostics.cpp

@@ -35,6 +35,7 @@ struct Event {
     long tid;
     char text[192];
     audio_capture_timing capture{};
+    audio_playback_timing playback{};
 };
 struct Slot { std::atomic<uint64_t> sequence{0}; Event event{}; };
 std::array<Slot, kCapacity> queue;
@@ -51,7 +52,7 @@ uint64_t now(clockid_t clock) {
 
 /* Ограниченная MPSC-очередь: producer не ждёт, не выделяет память и не пишет лог. */
 void publish(const void* device, int kind, int stage, int64_t value, int64_t detail, const char* text,
-             const audio_capture_timing* capture = nullptr) {
+             const audio_capture_timing* capture = nullptr, const audio_playback_timing* playback = nullptr) {
     if (!running.load(std::memory_order_relaxed)) return;
     const auto wall = now(CLOCK_MONOTONIC), cpu = now(CLOCK_THREAD_CPUTIME_ID);
     uint64_t pos = writePosition.load(std::memory_order_relaxed);
@@ -64,6 +65,7 @@ void publish(const void* device, int kind, int stage, int64_t value, int64_t det
             e.kind = kind; e.stage = stage; e.value = value; e.detail = detail;
             e.tid = syscall(SYS_gettid);
             if (capture) e.capture = *capture;
+            if (playback) e.playback = *playback;
             e.text[0] = 0;
             if (text) {
                 std::strncpy(e.text, text, sizeof(e.text) - 1);
@@ -101,6 +103,8 @@ struct Device {
     audio_capture_timing capture{};
     uint64_t captureBlocks = 0, captureFrames = 0, captureSequence = 0, captureJumps = 0, captureMissing = 0, captureEstimated = 0;
     int64_t captureMinDelta = 0, captureMaxDelta = 0, captureMinLatency = 0, captureMaxLatency = 0;
+    uint64_t playbackBlocks = 0, playbackFrames = 0, playbackMeasured = 0;
+    int64_t playbackMinLatency = 0, playbackMaxLatency = 0, playbackMaxConverter = 0, playbackMaxBlock = 0;
 };
 
 void consume(const Event& e, std::map<uintptr_t, Device>& devices) {
@@ -146,6 +150,17 @@ void consume(const Event& e, std::map<uintptr_t, Device>& devices) {
         if (c.missing) d.captureMissing += c.frames;
         if (!c.measured) ++d.captureEstimated;
         d.capture = c;
+    } else if (e.kind == AUDIO_DIAG_PLAYBACK_CLOCK) {
+        const auto& p = e.playback;
+        ++d.playbackBlocks; d.playbackFrames += p.frames;
+        if (p.rate) d.playbackMaxBlock = std::max(d.playbackMaxBlock, int64_t(p.frames) * 1000000 / p.rate);
+        d.playbackMaxConverter = std::max(d.playbackMaxConverter, p.converter_us);
+        if (p.measured) {
+            if (!d.playbackMeasured) d.playbackMinLatency = d.playbackMaxLatency = p.latency_us;
+            ++d.playbackMeasured;
+            d.playbackMinLatency = std::min(d.playbackMinLatency, p.latency_us);
+            d.playbackMaxLatency = std::max(d.playbackMaxLatency, p.latency_us);
+        }
     } else if (e.kind == AUDIO_DIAG_LABEL) {
         d.label = e.text; d.id = uint64_t(e.value);
         DEBUG_INFO(DEBUG_CATEGORY_CALL, "AudioDiag: label dev=%p role=%s id=%016llx", (void*)e.device, e.text, (unsigned long long)d.id);
@@ -191,6 +206,15 @@ void report(std::map<uintptr_t, Device>& devices, uint64_t wall, double elapsed)
                             (unsigned long long)dropped);
     std::set<long> stalledThreads;
     for (auto& [key, d] : devices) {
+        if (d.playbackBlocks) {
+            DEBUG_DEBUG(DEBUG_CATEGORY_CALL, "AudioDiag: playback clock dev=%p role=%s blocks=%llu frames=%llu measured=%llu "
+                        "backend_us=%lld..%lld converter_max_us=%lld block_max_us=%lld",
+                        (void*)key, d.label.c_str(), (unsigned long long)d.playbackBlocks, (unsigned long long)d.playbackFrames,
+                        (unsigned long long)d.playbackMeasured, (long long)d.playbackMinLatency, (long long)d.playbackMaxLatency,
+                        (long long)d.playbackMaxConverter, (long long)d.playbackMaxBlock);
+            d.playbackBlocks = d.playbackFrames = d.playbackMeasured = 0;
+            d.playbackMinLatency = d.playbackMaxLatency = d.playbackMaxConverter = d.playbackMaxBlock = 0;
+        }
         if (d.captureBlocks) {
             DEBUG_DEBUG(DEBUG_CATEGORY_CALL, "AudioDiag: capture clock dev=%p role=%s id=%016llx blocks=%llu frames=%llu "
                         "sequence=%llu delta_us=%lld..%lld latency_us=%lld..%lld jumps=%llu missing_frames=%llu estimated_blocks=%llu",
@@ -295,6 +319,9 @@ void audio_diag_log(const void* log, unsigned int level, const char* message) {
 void audio_diag_capture_clock(const void* device, const audio_capture_timing* timing) {
     publish(device, AUDIO_DIAG_CAPTURE_CLOCK, AUDIO_DIAG_READ, 0, 0, nullptr, timing);
 }
+void audio_diag_playback_clock(const void* device, const audio_playback_timing* timing) {
+    publish(device, AUDIO_DIAG_PLAYBACK_CLOCK, AUDIO_DIAG_WRITE, 0, 0, nullptr, nullptr, timing);
+}
 #else
 void audio_diag_start(void) {}
 void audio_diag_stop(void) {}
@@ -302,4 +329,5 @@ void audio_diag_event(const void*, int, int, int64_t, int64_t) {}
 void audio_diag_label(const void*, uint64_t, const char*) {}
 void audio_diag_log(const void*, unsigned int, const char*) {}
 void audio_diag_capture_clock(const void*, const audio_capture_timing*) {}
+void audio_diag_playback_clock(const void*, const audio_playback_timing*) {}
 #endif

tools/chatgui/transport/audio_diagnostics.h

@@ -16,7 +16,7 @@ enum audio_diag_kind {
     AUDIO_DIAG_ENTER, AUDIO_DIAG_LEAVE, AUDIO_DIAG_OPEN, AUDIO_DIAG_CLOSE,
     AUDIO_DIAG_NOTIFY, AUDIO_DIAG_BACKEND_STARTED, AUDIO_DIAG_HOLE, AUDIO_DIAG_UNFILLED, AUDIO_DIAG_CAPTURE_DROP,
     AUDIO_DIAG_NO_PROGRESS, AUDIO_DIAG_BACKEND_UNDERRUN, AUDIO_DIAG_BACKEND_OVERFLOW, AUDIO_DIAG_ERROR, AUDIO_DIAG_LABEL, AUDIO_DIAG_LOG,
-    AUDIO_DIAG_CAPTURE_CLOCK
+    AUDIO_DIAG_CAPTURE_CLOCK, AUDIO_DIAG_PLAYBACK_CLOCK
 };
 
 struct audio_capture_timing {
@@ -25,6 +25,12 @@ struct audio_capture_timing {
     int measured, missing;
 };
 
+struct audio_playback_timing {
+    int64_t latency_us, converter_us;
+    unsigned frames, rate;
+    int measured;
+};
+
 /* start/stop — управляющий поток; перед stop все producers должны быть остановлены. */
 void audio_diag_start(void);
 void audio_diag_stop(void);
@@ -32,6 +38,7 @@ void audio_diag_event(const void* device, int kind, int stage, int64_t value, in
 void audio_diag_label(const void* device, uint64_t id, const char* name);
 void audio_diag_log(const void* log, unsigned int level, const char* message);
 void audio_diag_capture_clock(const void* device, const struct audio_capture_timing* timing);
+void audio_diag_playback_clock(const void* device, const struct audio_playback_timing* timing);
 
 #ifdef __cplusplus
 }

tools/chatgui/transport/miniaudio_impl.c

@@ -165,6 +165,11 @@ static void audio_timing_begin(void* opaque, unsigned int frames, int capture) {
         struct audio_capture_timing timing = {time, latency, (int64_t)converter_delay * 1000000 / device->sampleRate,
                                              frames, rate, audio_timing.valid, audio_timing.missing};
         audio_diag_capture_clock(device, &timing);
+    } else {
+        struct audio_playback_timing timing = {latency,
+            (int64_t)(converter_delay + device->playback.inputCacheRemaining) * 1000000 / device->sampleRate,
+            frames, rate, audio_timing.valid};
+        audio_diag_playback_clock(device, &timing);
     }
     audio_timing.hole_pending = 0;
 }