utun2/src/call/call_jitter.cpp

// Буфер звонка: храним речь в порядке доставки, ускоряем накопление без изменения высоты голоса.
#include "call_jitter.h"
#include "call_jitter_window.h"
#include <chrono>
#include "../../lib/debug_config.h"
#include "../../lib/mem.h"
#include "soundtouch/SoundTouch.h"

#include <algorithm>
#include <cmath>
#include <cstring>
#include <mutex>
#include <new>
#include <vector>

static constexpr int kSampleRate = 48000;
static constexpr int kFrameSamples = 960;
static constexpr int kMaxEncoded = 512;
static constexpr int kTargetMs = 60;
static constexpr double kMaxTempo = 1.6;
static constexpr int kTempoRampMs = 500;

template<typename T> static T clamp_value(T value, T low, T high) {
    return std::max(low, std::min(value, high));
}

struct call_jitter {
    std::mutex mtx;
    uint8_t ring[CALL_JITTER_MAX_FRAMES][kMaxEncoded];
    int lens[CALL_JITTER_MAX_FRAMES];
    int head = 0, count = 0;
    uint32_t dropped = 0, underruns = 0;
    int processing_ms = 0, tempo_x100 = 100; /* снимок аудиопотока под mtx */
    call_jitter_window depth_window, arrival_window;
    call_jitter_range range = {0, 0, 60}; /* опубликованный снимок под mtx */
    call_jitter_decode_fn decode;
    void* arg;
    soundtouch::SoundTouch st;
    int16_t pcm[kFrameSamples];
    float input[kFrameSamples];
    std::vector<float> output;
    bool primed = false, catching_up = false;
    double tempo = 1.0, ema_ms = kTargetMs;
};

extern "C" {
struct call_jitter* call_jitter_create(call_jitter_decode_fn decode, void* arg, int max_reserve_ms) {
    if (max_reserve_ms < 500 || max_reserve_ms > 2000) {
        DEBUG_ERROR(DEBUG_CATEGORY_CALL, "call_jitter: invalid max reserve=%dms", max_reserve_ms);
        return nullptr;
    }
    if (!decode) {
        DEBUG_ERROR(DEBUG_CATEGORY_CALL, "call_jitter: missing decoder");
        return nullptr;
    }
    void* memory = u_malloc(sizeof(call_jitter));
    if (!memory) {
        DEBUG_ERROR(DEBUG_CATEGORY_CALL, "call_jitter: allocation failed");
        return nullptr;
    }
    auto* j = new (memory) call_jitter;
    j->arrival_window.max_reserve = max_reserve_ms;
    j->decode = decode; j->arg = arg;
    j->st.setSampleRate(kSampleRate); j->st.setChannels(1); j->st.setTempo(1.0);
    j->st.setSetting(SETTING_USE_QUICKSEEK, 1);
    j->st.setSetting(SETTING_SEQUENCE_MS, 20);
    j->st.setSetting(SETTING_SEEKWINDOW_MS, 10);
    j->st.setSetting(SETTING_OVERLAP_MS, 4);
    j->st.setSetting(SETTING_USE_AA_FILTER, 0);
    DEBUG_INFO(DEBUG_CATEGORY_CALL, "call_jitter: created capacity=%dms target=%dms max_reserve=%dms tempo=1.0..%.1f ramp=%dms",
               CALL_JITTER_MAX_FRAMES * CALL_JITTER_FRAME_MS, kTargetMs, max_reserve_ms, kMaxTempo, kTempoRampMs);
    return j;
}

void call_jitter_destroy(struct call_jitter* j) {
    if (!j) return;
    j->~call_jitter(); u_free(j);
}

void call_jitter_push(struct call_jitter* j, const uint8_t* enc, int len, uint32_t sent_ms, int64_t received_ms) {
    if (!j || !enc || len <= 0 || len > kMaxEncoded || received_ms < 0) {
        DEBUG_WARN(DEBUG_CATEGORY_CALL, "call_jitter: invalid frame len=%d received=%lldms", len, (long long)received_ms);
        return;
    }
    std::lock_guard<std::mutex> lock(j->mtx);
    if (received_ms < j->arrival_window.last_received_ms) {
        DEBUG_WARN(DEBUG_CATEGORY_CALL, "call_jitter: receive clock moved backwards previous=%lldms now=%lldms",
                   (long long)j->arrival_window.last_received_ms, (long long)received_ms);
        return;
    }
    bool log_arrival = j->arrival_window.observe_arrival(received_ms, sent_ms);
    j->range.reserve_ms = j->arrival_window.reserve;
    if (log_arrival)
        DEBUG_DEBUG(DEBUG_CATEGORY_CALL, "call_jitter: delivery variation=%lldms reserve=%dms queued=%dms",
                    (long long)(j->arrival_window.high - j->arrival_window.low), j->arrival_window.reserve,
                    j->count * CALL_JITTER_FRAME_MS);
    if (j->count == CALL_JITTER_MAX_FRAMES) {
        j->head = (j->head + 1) % CALL_JITTER_MAX_FRAMES; j->count--; j->dropped++;
        if (j->dropped == 1 || j->dropped % 50 == 0)
            DEBUG_WARN(DEBUG_CATEGORY_CALL, "call_jitter: overflow capacity=15000ms dropped=%u", j->dropped);
    }
    int idx = (j->head + j->count) % CALL_JITTER_MAX_FRAMES;
    std::memcpy(j->ring[idx], enc, (size_t)len); j->lens[idx] = len; j->count++;
}

int call_jitter_pull(struct call_jitter* j, int16_t* out, int max_samples) {
    if (!j || !out || max_samples <= 0) return 0;
    int count, reserve;
    { std::lock_guard<std::mutex> lock(j->mtx); count = j->count; reserve = j->arrival_window.reserve; }
    // Готовый PCM учитываем в длительности исходной речи.
    double depth = count * CALL_JITTER_FRAME_MS
        + (j->st.numUnprocessedSamples() + j->st.numSamples() * j->tempo) * 1000.0 / kSampleRate;
    int64_t now_ms = std::chrono::duration_cast<std::chrono::milliseconds>(
        std::chrono::steady_clock::now().time_since_epoch()).count();
    bool log_stats = j->depth_window.observe(now_ms, (int)std::lround(depth));
    {
        std::lock_guard<std::mutex> lock(j->mtx);
        j->range.min_ms = (int)j->depth_window.low; j->range.max_ms = (int)j->depth_window.high;
    }
    if (log_stats)
        DEBUG_DEBUG(DEBUG_CATEGORY_CALL, "call_jitter: depth=%.0fms min=%lldms max=%lldms reserve=%dms tempo=%.2f",
                    depth, (long long)j->depth_window.low, (long long)j->depth_window.high, reserve, j->tempo);
    if (!j->primed) {
        if (depth < reserve) return 0;
        j->primed = true;
        DEBUG_INFO(DEBUG_CATEGORY_CALL, "call_jitter: playback resumed depth=%.0fms reserve=%dms", depth, reserve);
    }
    j->ema_ms += 0.2 * (depth - j->ema_ms);
    double target = 1.0 + (kMaxTempo - 1.0) * clamp_value((std::min(depth, j->ema_ms) - reserve - 15.0) / 925.0, 0.0, 1.0);
    // Линейный предел 1.0→1.6 за 500мс PCM. Снижение следует за глубиной буфера с тем же пределом.
    double step = (kMaxTempo - 1.0) * max_samples * 1000.0 / (kSampleRate * kTempoRampMs);
    j->tempo += clamp_value(target - j->tempo, -step, step);
    j->st.setTempo(j->tempo);
    bool catching = j->tempo > 1.05;
    if (catching != j->catching_up) {
        j->catching_up = catching;
        DEBUG_INFO(DEBUG_CATEGORY_CALL, "call_jitter: catch-up %s depth=%.0fms tempo=%.2f",
                   catching ? "started" : "finished", depth, j->tempo);
    }
    uint8_t enc[kMaxEncoded];
    while (j->st.numSamples() < (uint)max_samples) {
        int len;
        {
            std::lock_guard<std::mutex> lock(j->mtx);
            if (!j->count) break;
            len = j->lens[j->head]; std::memcpy(enc, j->ring[j->head], (size_t)len);
            j->head = (j->head + 1) % CALL_JITTER_MAX_FRAMES; j->count--;
        }
        int n = j->decode(j->arg, enc, len, j->pcm, kFrameSamples);
        if (n <= 0 || n > kFrameSamples) {
            DEBUG_WARN(DEBUG_CATEGORY_CALL, "call_jitter: decode failed len=%d result=%d", len, n);
            continue;
        }
        for (int i = 0; i < n; i++) j->input[i] = j->pcm[i] / 32768.0f;
        j->st.putSamples(j->input, (uint)n);
    }
    j->output.resize((size_t)max_samples);
    int n = (int)j->st.receiveSamples(j->output.data(), (uint)max_samples);
    for (int i = 0; i < n; i++) out[i] = (int16_t)std::lrintf(clamp_value(j->output[i] * 32767.0f, -32768.0f, 32767.0f));
    int processing_ms = (int)std::lround((j->st.numUnprocessedSamples() + j->st.numSamples() * j->tempo) * 1000.0 / kSampleRate);
    {
        std::lock_guard<std::mutex> lock(j->mtx);
        j->processing_ms = processing_ms; j->tempo_x100 = (int)std::lround(j->tempo * 100);
        if (n < max_samples) j->underruns++;
    }
    if (n < max_samples) {
        j->primed = false;
        DEBUG_DEBUG(DEBUG_CATEGORY_CALL, "call_jitter: waiting for refill pcm=%d/%d processing=%dms", n, max_samples, processing_ms);
    }
    return n;
}

void call_jitter_get_stats(struct call_jitter* j, int* depth_ms, int* tempo_x100,
                           uint32_t* dropped, uint32_t* underruns, struct call_jitter_range* range) {
    if (!j) return;
    std::lock_guard<std::mutex> lock(j->mtx);
    if (depth_ms) *depth_ms = j->count * CALL_JITTER_FRAME_MS + j->processing_ms;
    if (tempo_x100) *tempo_x100 = j->tempo_x100;
    if (range) *range = j->range;
    if (dropped) *dropped = j->dropped;
    if (underruns) *underruns = j->underruns;
}
}