path: root/src/audio/synth.cc

// This file is part of the 64k demo project.
// It implements the multi-voice additive synthesis engine.
// Supports real-time spectrogram updates and peak detection.

#include "synth.h"
#include "audio/dct.h"
#include "audio/window.h"
#include "util/debug.h"
#include <atomic>
#include <math.h>
#include <stdio.h>  // For printf
#include <string.h> // For memset

#if defined(DEBUG_LOG_SYNTH)
#include "audio/audio.h"
#include "audio/audio_backend.h"
#endif /* defined(DEBUG_LOG_SYNTH) */

struct Voice {
  bool active;
  int spectrogram_id;
  float volume;
  float pan_left;
  float pan_right;

  int current_spectral_frame;
  int total_spectral_frames;

  float time_domain_buffer[DCT_SIZE];
  int buffer_pos;
  float fractional_pos;  // Fractional sample position for tempo scaling

  const volatile float* active_spectral_data;
};

static struct {
  Spectrogram spectrograms[MAX_SPECTROGRAMS];
  const volatile float* active_spectrogram_data[MAX_SPECTROGRAMS];
  bool spectrogram_registered[MAX_SPECTROGRAMS];
} g_synth_data;

static Voice g_voices[MAX_VOICES];
static volatile float g_current_output_peak =
    0.0f;                                   // Global peak for visualization
static float g_hamming_window[WINDOW_SIZE]; // Static window for optimization
static float g_tempo_scale = 1.0f;          // Playback speed multiplier

#if defined(DEBUG_LOG_SYNTH)
static float g_elapsed_time_sec = 0.0f; // Tracks elapsed time for event hooks
#endif /* defined(DEBUG_LOG_SYNTH) */

void synth_init() {
  memset(&g_synth_data, 0, sizeof(g_synth_data));
  memset(g_voices, 0, sizeof(g_voices));
  g_current_output_peak = 0.0f;
#if defined(DEBUG_LOG_SYNTH)
  g_elapsed_time_sec = 0.0f;
#endif /* defined(DEBUG_LOG_SYNTH) */
  // Initialize the Hamming window once
  hamming_window_512(g_hamming_window);
}

void synth_shutdown() {
  // Nothing to do here since we are not allocating memory
}

void synth_set_tempo_scale(float tempo_scale) {
  g_tempo_scale = tempo_scale;
}

int synth_register_spectrogram(const Spectrogram* spec) {
#if defined(DEBUG_LOG_SYNTH)
  // VALIDATION: Check spectrogram pointer and data
  if (spec == nullptr) {
    DEBUG_SYNTH( "[SYNTH ERROR] Null spectrogram pointer\n");
    return -1;
  }
  if (spec->spectral_data_a == nullptr || spec->spectral_data_b == nullptr) {
    DEBUG_SYNTH( "[SYNTH ERROR] Null spectral data pointers\n");
    return -1;
  }
  if (spec->num_frames <= 0 || spec->num_frames > 10000) {
    DEBUG_SYNTH( "[SYNTH ERROR] Invalid num_frames=%d (must be 1-10000)\n",
            spec->num_frames);
    return -1;
  }
  // VALIDATION: Check spectral data isn't all zeros (common corruption symptom)
  bool all_zero = true;
  const float* data = spec->spectral_data_a;
  const int samples_to_check = (spec->num_frames > 10) ? 10 * DCT_SIZE : spec->num_frames * DCT_SIZE;
  for (int j = 0; j < samples_to_check; ++j) {
    if (data[j] != 0.0f) {
      all_zero = false;
      break;
    }
  }
  if (all_zero) {
    DEBUG_SYNTH( "[SYNTH WARNING] Spectrogram appears to be all zeros (num_frames=%d)\n",
            spec->num_frames);
  }
#endif

  for (int i = 0; i < MAX_SPECTROGRAMS; ++i) {
    if (!g_synth_data.spectrogram_registered[i]) {
      g_synth_data.spectrograms[i] = *spec;
      g_synth_data.active_spectrogram_data[i] = spec->spectral_data_a;
      g_synth_data.spectrogram_registered[i] = true;
      return i;
    }
  }
  return -1; // No free slots
}

void synth_unregister_spectrogram(int spectrogram_id) {
  if (spectrogram_id >= 0 && spectrogram_id < MAX_SPECTROGRAMS) {
    g_synth_data.spectrogram_registered[spectrogram_id] = false;
  }
}

float* synth_begin_update(int spectrogram_id) {
  if (spectrogram_id < 0 || spectrogram_id >= MAX_SPECTROGRAMS ||
      !g_synth_data.spectrogram_registered[spectrogram_id]) {
    return nullptr;
  }

  const volatile float* active_ptr =
      g_synth_data.active_spectrogram_data[spectrogram_id];

  if (active_ptr == g_synth_data.spectrograms[spectrogram_id].spectral_data_a) {
    return (float*)(g_synth_data.spectrograms[spectrogram_id].spectral_data_b);
  } else {
    return (float*)(g_synth_data.spectrograms[spectrogram_id].spectral_data_a);
  }
}

void synth_commit_update(int spectrogram_id) {
  if (spectrogram_id < 0 || spectrogram_id >= MAX_SPECTROGRAMS ||
      !g_synth_data.spectrogram_registered[spectrogram_id]) {
    return;
  }

  const volatile float* old_active_ptr =
      g_synth_data.active_spectrogram_data[spectrogram_id];
  const float* new_active_ptr =
      (old_active_ptr ==
       g_synth_data.spectrograms[spectrogram_id].spectral_data_a)
          ? g_synth_data.spectrograms[spectrogram_id].spectral_data_b
          : g_synth_data.spectrograms[spectrogram_id].spectral_data_a;

  // Atomic swap using GCC/Clang builtins for thread safety
  __atomic_store_n(
      (const float**)&g_synth_data.active_spectrogram_data[spectrogram_id],
      new_active_ptr, __ATOMIC_RELEASE);
}

void synth_trigger_voice(int spectrogram_id, float volume, float pan) {
  if (spectrogram_id < 0 || spectrogram_id >= MAX_SPECTROGRAMS ||
      !g_synth_data.spectrogram_registered[spectrogram_id]) {
#if defined(DEBUG_LOG_SYNTH)
    DEBUG_SYNTH( "[SYNTH ERROR] Invalid spectrogram_id=%d in trigger_voice\n",
            spectrogram_id);
#endif
    return;
  }

#if defined(DEBUG_LOG_SYNTH)
  // VALIDATION: Check volume and pan ranges
  if (volume < 0.0f || volume > 2.0f) {
    DEBUG_SYNTH( "[SYNTH WARNING] Unusual volume=%.2f for spectrogram_id=%d\n",
            volume, spectrogram_id);
  }
  if (pan < -1.0f || pan > 1.0f) {
    DEBUG_SYNTH( "[SYNTH WARNING] Invalid pan=%.2f (clamping) for spectrogram_id=%d\n",
            pan, spectrogram_id);
    pan = (pan < -1.0f) ? -1.0f : 1.0f;
  }
#endif

  for (int i = 0; i < MAX_VOICES; ++i) {
    if (!g_voices[i].active) {
      Voice& v = g_voices[i];
      v.active = true;
      v.spectrogram_id = spectrogram_id;
      v.volume = volume;

      // Simple linear panning
      v.pan_left = (pan > 0.0f) ? (1.0f - pan) : 1.0f;
      v.pan_right = (pan < 0.0f) ? (1.0f + pan) : 1.0f;

      v.current_spectral_frame = 0;
      v.total_spectral_frames =
          g_synth_data.spectrograms[spectrogram_id].num_frames;
      v.buffer_pos = DCT_SIZE; // Force IDCT on first render
      v.fractional_pos = 0.0f; // Initialize fractional position for tempo scaling
      v.active_spectral_data =
          g_synth_data.active_spectrogram_data[spectrogram_id];

#if defined(DEBUG_LOG_SYNTH)
      // Notify backend of voice trigger event (for testing/tracking)
      AudioBackend* backend = audio_get_backend();
      if (backend != nullptr) {
        backend->on_voice_triggered(g_elapsed_time_sec, spectrogram_id, volume,
                                     pan);
      }
#endif /* defined(DEBUG_LOG_SYNTH) */

      return; // Voice triggered
    }
  }
}

void synth_render(float* output_buffer, int num_frames) {
  // Use the pre-calculated window
  // float window[WINDOW_SIZE];
  // hamming_window_512(window);

  // Faster decay for more responsive visuals
  g_current_output_peak *= 0.90f;

  for (int i = 0; i < num_frames; ++i) {
    float left_sample = 0.0f;
    float right_sample = 0.0f;

    for (int v_idx = 0; v_idx < MAX_VOICES; ++v_idx) {
      Voice& v = g_voices[v_idx];
      if (!v.active)
        continue;

      if (v.buffer_pos >= DCT_SIZE) {
        if (v.current_spectral_frame >= v.total_spectral_frames) {
          v.active = false;
          continue;
        }

        // Fetch the latest active spectrogram pointer for this voice
        v.active_spectral_data =
            g_synth_data.active_spectrogram_data[v.spectrogram_id];

        const float* spectral_frame = (const float*)v.active_spectral_data +
                                      (v.current_spectral_frame * DCT_SIZE);

        float windowed_frame[DCT_SIZE];
        for (int j = 0; j < DCT_SIZE; ++j) {
          windowed_frame[j] =
              spectral_frame[j] * g_hamming_window[j]; // Use static window
        }

        idct_512(windowed_frame, v.time_domain_buffer);

        v.buffer_pos = 0;
        ++v.current_spectral_frame;
      }

      float voice_sample = v.time_domain_buffer[v.buffer_pos] * v.volume;
      left_sample += voice_sample * v.pan_left;
      right_sample += voice_sample * v.pan_right;

      // Advance voice position
      ++v.buffer_pos;
    }

    output_buffer[i * 2] = left_sample;
    output_buffer[i * 2 + 1] = right_sample;

    // Update the peak with the new max (attack)
    g_current_output_peak = fmaxf(
        g_current_output_peak, fmaxf(fabsf(left_sample), fabsf(right_sample)));
  }

#if defined(DEBUG_LOG_SYNTH)
  // Update elapsed time for event tracking (32000 Hz sample rate)
  const float sample_rate = 32000.0f;
  g_elapsed_time_sec += (float)num_frames / sample_rate;
#endif /* defined(DEBUG_LOG_SYNTH) */
}

int synth_get_active_voice_count() {
  int count = 0;
  for (int i = 0; i < MAX_VOICES; ++i) {
    if (g_voices[i].active) {
      ++count;
    }
  }
  return count;
}

float synth_get_output_peak() {
  return g_current_output_peak;
}