WiP -Audio fastpath

platformio.ini

@@ -812,6 +812,7 @@ build_flags_S =
   -D USERMOD_AUDIOREACTIVE
   -D UM_AUDIOREACTIVE_USE_NEW_FFT ; use latest (upstream) FFTLib, instead of older library modified by blazoncek. Slightly faster, more accurate, needs 2KB RAM extra   
   -D USERMOD_ARTIFX ; WLEDMM usermod
+  -D WLEDMM_FASTPATH ; WLEDMM experimental option. Reduces audio lag (latency), and allows for faster LED framerates
   ; -D WLED_DISABLE_LOXONE
   ; -D WLED_DISABLE_ALEXA
   ; -D WLED_DISABLE_HUESYNC
@@ -914,7 +915,9 @@ lib_deps = ${esp32_4MB_M_base.lib_deps} ${common_mm.lib_deps_XL}
 board = esp32dev
 upload_speed = 460800 ; or 921600
 platform = ${esp32.platformV4}
-platform_packages = ${esp32.platformV4_packages}
+platform_packages = 
+  ${esp32.platformV4_packages}
+  toolchain-xtensa-esp32 @ 8.4.0+2021r2-patch5 ; align main tools with riscV tools
 build_unflags = ${common.build_unflags}
 build_flags = ${common.build_flags} ${esp32.build_flagsV4} ${common_mm.build_flags_S}
   -Wno-misleading-indentation -Wno-format-truncation 

usermods/audioreactive/audio_reactive.h

@@ -254,7 +254,11 @@ constexpr SRate_t SAMPLE_RATE = 22050;        // Base sample rate in Hz - 22Khz
 //constexpr SRate_t SAMPLE_RATE = 16000;        // 16kHz - use if FFTtask takes more than 20ms. Physical sample time -> 32ms
 //constexpr SRate_t SAMPLE_RATE = 20480;        // Base sample rate in Hz - 20Khz is experimental.    Physical sample time -> 25ms
 //constexpr SRate_t SAMPLE_RATE = 10240;        // Base sample rate in Hz - previous default.         Physical sample time -> 50ms
+#ifndef WLEDMM_FASTPATH
 #define FFT_MIN_CYCLE 21                      // minimum time before FFT task is repeated. Use with 22Khz sampling
+#else
+#define FFT_MIN_CYCLE 15                      // reduce min time, to allow faster catch-up when I2S is lagging 
+#endif
 //#define FFT_MIN_CYCLE 30                      // Use with 16Khz sampling
 //#define FFT_MIN_CYCLE 23                      // minimum time before FFT task is repeated. Use with 20Khz sampling
 //#define FFT_MIN_CYCLE 46                      // minimum time before FFT task is repeated. Use with 10Khz sampling
@@ -1129,8 +1133,13 @@ class AudioReactive : public Usermod {
     {
       float    sampleAdj;           // Gain adjusted sample value
       float    tmpSample;           // An interim sample variable used for calculatioins.
+#ifdef WLEDMM_FASTPATH
+      constexpr float weighting = 0.35f;  // slightly reduced filter strength, to reduce audio latency
+      constexpr float weighting2 = 0.25f;
+#else
       const float weighting = 0.2f; // Exponential filter weighting. Will be adjustable in a future release.
       const float weighting2 = 0.073f; // Exponential filter weighting, for rising signal (a bit more robust against spikes)
+#endif
       const int   AGC_preset = (soundAgc > 0)? (soundAgc-1): 0; // make sure the _compiler_ knows this value will not change while we are inside the function
       static bool isFrozen = false;
       static bool haveSilence = true;
@@ -1732,7 +1741,11 @@ class AudioReactive : public Usermod {
 
         // get AGC sensitivity and sound pressure
         static unsigned long lastEstimate = 0;
+#ifdef WLEDMM_FASTPATH
+        if (millis() - lastEstimate > 7) {
+#else
         if (millis() - lastEstimate > 12) {
+#endif
           lastEstimate = millis();
           agcSensitivity = getSensitivity();
           if (limiterOn)

usermods/audioreactive/audio_source.h

@@ -180,8 +180,13 @@ class I2SSource : public AudioSource {
 #if ESP_IDF_VERSION >= ESP_IDF_VERSION_VAL(4, 2, 0)
         .communication_format = i2s_comm_format_t(I2S_COMM_FORMAT_STAND_I2S),
         //.intr_alloc_flags = ESP_INTR_FLAG_LEVEL1,
+#ifdef WLEDMM_FASTPATH
+        .intr_alloc_flags = ESP_INTR_FLAG_LEVEL2|ESP_INTR_FLAG_LEVEL3,  // seems to reduce noise
+        .dma_buf_count = 28,                                            // 160ms buffer (128 * dma_buf_count / sampleRate)
+#else
         .intr_alloc_flags = ESP_INTR_FLAG_LEVEL2,
         .dma_buf_count = 8,
+#endif
         .dma_buf_len = _blockSize,
         .use_apll = 0,
         //.fixed_mclk = 0,

wled00/FX.cpp

@@ -2243,7 +2243,7 @@ uint16_t mode_colortwinkle() {
       }
     }
   }
-  return FRAMETIME_FIXED;
+  return FRAMETIME_FIXED_SLOW;
 }
 static const char _data_FX_MODE_COLORTWINKLE[] PROGMEM = "Colortwinkles@Fade speed,Spawn speed;;!;;m12=0"; //pixels
 
@@ -3096,7 +3096,7 @@ uint16_t candle(bool multi)
     }
   }
 
-  return FRAMETIME_FIXED;
+  return FRAMETIME_FIXED_SLOW;
 }
 
 
@@ -6042,6 +6042,7 @@ uint16_t mode_ripplepeak(void) {                // * Ripple peak. By Andrew Tuli
   // printUmData();
 
   if (SEGENV.call == 0) {
+    SEGMENT.setUpLeds();
     SEGENV.aux0 = 255;
     SEGMENT.custom1 = *binNum;
     SEGMENT.custom2 = *maxVol * 2;
@@ -6052,8 +6053,9 @@ uint16_t mode_ripplepeak(void) {                // * Ripple peak. By Andrew Tuli
   *binNum = SEGMENT.custom1;                              // Select a bin.
   *maxVol = SEGMENT.custom2 / 2;                          // Our volume comparator.
 
-  SEGMENT.fade_out(240);                                  // Lower frame rate means less effective fading than FastLED
-  SEGMENT.fade_out(240);
+  //SEGMENT.fade_out(240);                                  // Lower frame rate means less effective fading than FastLED
+  //SEGMENT.fade_out(240);
+  SEGMENT.fade_out(224);  // should be the same as 240 applied twice
 
   for (int i = 0; i < SEGMENT.intensity/16; i++) {   // Limit the number of ripples.
     if (samplePeak) ripples[i].state = 255;
@@ -6066,7 +6068,8 @@ uint16_t mode_ripplepeak(void) {                // * Ripple peak. By Andrew Tuli
         ripples[i].pos = random16(SEGLEN);
         #ifdef ESP32
           if (FFT_MajorPeak > 1)                          // log10(0) is "forbidden" (throws exception)
-          ripples[i].color = (int)(log10f(FFT_MajorPeak)*128);
+          //ripples[i].color = (int)(log10f(FFT_MajorPeak)*128);  // not to self: buggy !!
+          ripples[i].color = (int)(logf(FFT_MajorPeak)*32.0f);  // works up to 10025 hz
           else ripples[i].color = 0;
         #else
           ripples[i].color = random8();
@@ -6476,8 +6479,12 @@ uint16_t mode_midnoise(void) {                  // Midnoise. By Andrew Tuline.
   }
   float   volumeSmth   = *(float*)  um_data->u_data[0];
 
-  SEGMENT.fade_out(SEGMENT.speed);
-  SEGMENT.fade_out(SEGMENT.speed);
+  if (SEGENV.call == 0) {
+    SEGMENT.setUpLeds();
+    SEGMENT.fill(BLACK);
+  }
+  SEGMENT.fadeToBlackBy(SEGMENT.speed/2);
+  //SEGMENT.fade_out(SEGMENT.speed);
 
   float tmpSound2 = volumeSmth * (float)SEGMENT.intensity / 256.0;  // Too sensitive.
   tmpSound2 *= (float)SEGMENT.intensity / 128.0;              // Reduce sensitity/length.
@@ -6622,8 +6629,13 @@ uint16_t mode_plasmoid(void) {                  // Plasmoid. By Andrew Tuline.
   }
   float   volumeSmth   = *(float*)  um_data->u_data[0];
 
-  SEGMENT.fadeToBlackBy(32);
-
+  if (SEGENV.call == 0) {
+    SEGMENT.setUpLeds();
+    SEGMENT.fill(BLACK);
+  }
+  //SEGMENT.fadeToBlackBy(32);
+  SEGMENT.fadeToBlackBy(48);
+
   plasmoip->thisphase += beatsin8(6,-4,4);                          // You can change direction and speed individually.
   plasmoip->thatphase += beatsin8(7,-4,4);                          // Two phase values to make a complex pattern. By Andrew Tuline.
 
@@ -6832,7 +6844,7 @@ uint16_t mode_blurz(void) {                    // Blurz. By Andrew Tuline.
     SEGMENT.setPixelColor(segLoc, color_blend(SEGCOLOR(1), SEGMENT.color_from_palette((uint16_t)pixColor, false, PALETTE_SOLID_WRAP, 0),(uint8_t)pixIntensity)); // repaint center pixel after blur
   } else SEGMENT.blur(max(SEGMENT.intensity, (uint8_t)1));  // silence - just blur it again
 
-  return FRAMETIME;
+  return FRAMETIME_FIXED;
 } // mode_blurz()
 static const char _data_FX_MODE_BLURZ[] PROGMEM = "Blurz ☾@Fade rate,Blur;!,Color mix;!;1f;sx=48,ix=127,m12=0,si=0"; // Pixels, Beatsin
 #endif
@@ -6921,7 +6933,10 @@ uint16_t mode_freqmap(void) {                   // Map FFT_MajorPeak to SEGLEN.
   float my_magnitude  = *(float*)um_data->u_data[5] / 4.0f;
   if (FFT_MajorPeak < 1) FFT_MajorPeak = 1;                                         // log10(0) is "forbidden" (throws exception)
 
-  if (SEGENV.call == 0) SEGMENT.fill(BLACK);
+  if (SEGENV.call == 0) {
+    SEGMENT.setUpLeds();
+    SEGMENT.fill(BLACK);
+  }
   int fadeoutDelay = (256 - SEGMENT.speed) / 32;
   if ((fadeoutDelay <= 1 ) || ((SEGENV.call % fadeoutDelay) == 0)) SEGMENT.fade_out(SEGMENT.speed);
 
@@ -6944,7 +6959,7 @@ uint16_t mode_freqmap(void) {                   // Map FFT_MajorPeak to SEGLEN.
     SEGMENT.setPixelColor(locn, color_blend(SEGCOLOR(1), SEGMENT.color_from_palette(SEGMENT.intensity+pixCol, false, PALETTE_SOLID_WRAP, 0), bright));
   }
 
-  return FRAMETIME;
+  return FRAMETIME_FIXED;
 } // mode_freqmap()
 static const char _data_FX_MODE_FREQMAP[] PROGMEM = "Freqmap@Fade rate,Starting color;!,!;!;1f;m12=0,si=0"; // Pixels, Beatsin
 
@@ -7068,8 +7083,8 @@ uint16_t mode_freqwave(void) {                  // Freqwave. By Andreas Pleschun
     SEGMENT.fill(BLACK);
   }
 
-  uint8_t secondHand = micros()/(256-SEGMENT.speed)/500 % 16;
-  if((SEGMENT.speed > 254) || (SEGENV.aux0 != secondHand)) {   // WLEDMM allow run run at full speed
+  uint8_t secondHand = (SEGMENT.speed < 255) ? (micros()/(256-SEGMENT.speed)/500 % 16) : 0;
+  if((SEGMENT.speed > 254) || (SEGENV.aux0 != secondHand)) {   // WLEDMM allow to run at full speed
     SEGENV.aux0 = secondHand;
 
     float sensitivity = 0.5f * mapf(SEGMENT.custom3, 1, 31, 0.5, 10); // reduced resolution slider

wled00/FX.h

@@ -57,10 +57,21 @@
 #endif
 
 /* Not used in all effects yet */
+#if defined(ARDUINO_ARCH_ESP32) && defined(WLEDMM_FASTPATH)   // WLEDMM go faster on ESP32
+#define WLED_FPS         120
+#define FRAMETIME_FIXED  (strip.getFrameTime() < 10 ? 12 : 24)
+#define WLED_FPS_SLOW         60
+#define FRAMETIME_FIXED_SLOW  (15)    // = 66 FPS => 1000/66
+//#define FRAMETIME        _frametime
+#define FRAMETIME        strip.getFrameTime()
+#else
 #define WLED_FPS         42
 #define FRAMETIME_FIXED  (1000/WLED_FPS)
+#define WLED_FPS_SLOW         42
+#define FRAMETIME_FIXED_SLOW  (1000/WLED_FPS_SLOW)
 //#define FRAMETIME        _frametime
 #define FRAMETIME        strip.getFrameTime()
+#endif
 
 /* each segment uses 52 bytes of SRAM memory, so if you're application fails because of
   insufficient memory, decreasing MAX_NUM_SEGMENTS may help */
@@ -560,7 +571,7 @@ typedef struct Segment {
     void setPixelColor(float i, uint32_t c, bool aa = true);
     void setPixelColor(float i, uint8_t r, uint8_t g, uint8_t b, uint8_t w = 0, bool aa = true) { setPixelColor(i, RGBW32(r,g,b,w), aa); }
     void setPixelColor(float i, CRGB c, bool aa = true)                                         { setPixelColor(i, RGBW32(c.r,c.g,c.b,0), aa); }
-    uint32_t getPixelColor(int i);
+    uint32_t __attribute__((pure)) getPixelColor(int i);  // WLEDMM attribute added
     // 1D support functions (some implement 2D as well)
     void blur(uint8_t);
     void fill(uint32_t c);
@@ -573,7 +584,7 @@ typedef struct Segment {
     void addPixelColor(int n, CRGB c)                             { addPixelColor(n, RGBW32(c.r,c.g,c.b,0)); } // automatically inline
     void fadePixelColor(uint16_t n, uint8_t fade);
     uint8_t get_random_wheel_index(uint8_t pos);
-    uint32_t color_from_palette(uint16_t, bool mapping, bool wrap, uint8_t mcol, uint8_t pbri = 255);
+	uint32_t __attribute__((pure)) color_from_palette(uint_fast16_t, bool mapping, bool wrap, uint8_t mcol, uint8_t pbri = 255);
     uint32_t color_wheel(uint8_t pos);
 
     // 2D matrix
@@ -583,14 +594,14 @@ typedef struct Segment {
     void createjMap(); //WLEDMM jMap
     void deletejMap(); //WLEDMM jMap
   #ifndef WLED_DISABLE_2D
-    uint16_t XY(uint16_t x, uint16_t y); // support function to get relative index within segment (for leds[])
+    uint16_t __attribute__((pure)) XY(uint16_t x, uint16_t y); // support function to get relative index within segment (for leds[]) // WLEDMM attribute pure
     void setPixelColorXY(int x, int y, uint32_t c); // set relative pixel within segment with color
     void setPixelColorXY(int x, int y, byte r, byte g, byte b, byte w = 0) { setPixelColorXY(x, y, RGBW32(r,g,b,w)); } // automatically inline
     void setPixelColorXY(int x, int y, CRGB c)                             { setPixelColorXY(x, y, RGBW32(c.r,c.g,c.b,0)); } // automatically inline
-    void setPixelColorXY(float x, float y, uint32_t c, bool aa = true);
+    void setPixelColorXY(float x, float y, uint32_t c, bool aa = true, bool fast = true);
     void setPixelColorXY(float x, float y, byte r, byte g, byte b, byte w = 0, bool aa = true) { setPixelColorXY(x, y, RGBW32(r,g,b,w), aa); }
     void setPixelColorXY(float x, float y, CRGB c, bool aa = true)                             { setPixelColorXY(x, y, RGBW32(c.r,c.g,c.b,0), aa); }
-    uint32_t getPixelColorXY(uint16_t x, uint16_t y);
+    uint32_t __attribute__((pure)) getPixelColorXY(uint16_t x, uint16_t y);   // WLEDMM attribute pure
     // 2D support functions
     void blendPixelColorXY(uint16_t x, uint16_t y, uint32_t color, uint8_t blend);
     void blendPixelColorXY(uint16_t x, uint16_t y, CRGB c, uint8_t blend)  { blendPixelColorXY(x, y, RGBW32(c.r,c.g,c.b,0), blend); }
@@ -686,11 +697,11 @@ class WS2812FX {  // 96 bytes
       _length(DEFAULT_LED_COUNT),
       _brightness(DEFAULT_BRIGHTNESS),
       _transitionDur(750),
-      _targetFps(WLED_FPS),
-      _frametime(FRAMETIME_FIXED),
+      _targetFps(WLED_FPS_SLOW),        // WLEDMM
+      _frametime(FRAMETIME_FIXED_SLOW), // WLEDMM
       _cumulativeFps(2),
 #ifdef ARDUINO_ARCH_ESP32
-      _cumulativeFps500(2*500),      // WLEDMM more accurate FPS measurement for ESP32
+      _cumulativeFps500(2*500),          // WLEDMM more accurate FPS measurement for ESP32
       _lastShow500(0),
 #endif
       _isServicing(false),
@@ -809,8 +820,8 @@ class WS2812FX {  // 96 bytes
 
     uint32_t
       now,
-      timebase,
-      getPixelColor(uint16_t);
+      timebase;
+    uint32_t __attribute__((pure)) getPixelColor(uint_fast16_t);   // WLEDMM attribute pure = does not have side-effects
 
     inline uint32_t getLastShow(void) { return _lastShow; }
     inline uint32_t segColor(uint8_t i) { return _colors_t[i]; }