WiP -Audio fastpath

platformio.ini

@@ -199,7 +199,8 @@ upload_speed = 115200
 # ------------------------------------------------------------------------------
 lib_compat_mode = strict
 lib_deps =
-    fastled/FastLED @ 3.5.0
+    ;fastled/FastLED @ 3.5.0
+    https://github.com/FastLED/FastLED.git#master ;; up to 50% faster
     IRremoteESP8266 @ 2.8.2
     ;;https://github.com/Aircoookie/ESPAsyncWebServer.git @ ~2.0.7
     https://github.com/lost-hope/ESPAsyncWebServer.git#master  ;; WLEDMM to display .log and .wled files in /edit
@@ -244,8 +245,8 @@ lib_deps =
   #https://github.com/lorol/LITTLEFS.git
   ESPAsyncTCP @ 1.2.2
   ESPAsyncUDP
-  ${env.lib_deps}
   makuna/NeoPixelBus @ 2.6.9
+  ${env.lib_deps}
 
 [esp32]
 #platform = https://github.com/tasmota/platform-espressif32/releases/download/v2.0.2.3/platform-espressif32-2.0.2.3.zip
@@ -269,8 +270,9 @@ lib_deps =
   ; https://github.com/lorol/LITTLEFS.git
   ; WLEDMM specific: use patched version of lorol LittleFS
   https://github.com/softhack007/LITTLEFS-threadsafe.git#master
+  ;;makuna/NeoPixelBus @ 2.6.9 ;; WLEDMM default
+  makuna/NeoPixelBus @ 2.7.1
   ${env.lib_deps}
-  makuna/NeoPixelBus @ 2.6.9
 
 ;; ** For compiling with latest Frameworks (IDF4.4.x and arduino-esp32 v2.0.x) **
 ;;; standard V4 platform
@@ -296,9 +298,8 @@ build_flagsV4 = -g
 ;;; V4.4.x libraries (without LOROL_LITTLEFS; with newer NeoPixelBus)
 lib_depsV4 =
   https://github.com/pbolduc/AsyncTCP.git @ 1.2.0 ;; WLEDMM this must be first in the list, otherwise Aircoookie/ESPAsyncWebServer pulls in an older version of AsyncTCP !!
+  makuna/NeoPixelBus @ 2.7.3
   ${env.lib_deps}
-  ;https://github.com/Makuna/NeoPixelBus.git#master ;; NPB 2.6.9 tends to crash whith IDF V4.4.3 -> use latest NeoPixelBus dev instead
-  makuna/NeoPixelBus @ 2.7.1
 
 [esp32s2]
 build_flags = -g
@@ -314,8 +315,8 @@ build_flags = -g
 
 lib_deps =
   https://github.com/pbolduc/AsyncTCP.git @ 1.2.0 ;; WLEDMM this must be first in the list, otherwise Aircoookie/ESPAsyncWebServer pulls in an older version of AsyncTCP !!
+  makuna/NeoPixelBus @ 2.7.3 ;; WLEDMM - new version is more stable on -S2
   ${env.lib_deps}
-  makuna/NeoPixelBus @ 2.7.1 ;; WLEDMM - new version is more stable on -S2
 
 [esp32c3]
 build_flags = -g
@@ -352,8 +353,8 @@ build_flags = -g
 
 lib_deps =
   https://github.com/pbolduc/AsyncTCP.git @ 1.2.0
-  ${env.lib_deps}
   makuna/NeoPixelBus @ 2.7.3
+  ${env.lib_deps}
 
 
 # ------------------------------------------------------------------------------
@@ -812,6 +813,7 @@ 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 +916,9 @@ 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 
@@ -1239,16 +1243,14 @@ build_flags = ${esp32_4MB_V4_M_base.build_flags}
   -D WLED_RELEASE_NAME=esp32_4MB_V4_M
   -D WLED_WATCHDOG_TIMEOUT=0 #-D WLED_DISABLE_BROWNOUT_DET
   -D ARDUINO_USB_CDC_ON_BOOT=0 ; needed for arduino-esp32 >=2.0.4; avoids errors on startup
-  ; RAM:   [==        ]  24.8% (used 81316 bytes from 327680 bytes)
-  ; Flash: [==========]  97.2% (used 1528253 bytes from 1572864 bytes)
   -D WLED_DISABLE_LOXONE ; FLASH 1272 bytes
   -D WLED_DISABLE_ALEXA ;  RAM 116 bytes; FLASH 13524 bytes
   -D WLED_DISABLE_HUESYNC ;RAM 122 bytes; FLASH  6308 bytes
-  ; RAM:   [==        ]  24.7% (used 81076 bytes from 327680 bytes)
-  ; Flash: [==========]  95.8% (used 1506893 bytes from 1572864 bytes)
   ;-D WLED_DISABLE_MQTT ;   RAM 216 bytes; FLASH 16496 bytes
   -D WLED_DISABLE_INFRARED ;RAM 136 bytes; FLASH 24492 bytes  ;; softhack007 disabled to stay below 100% flash size
 lib_ignore = IRremoteESP8266 ; use with WLED_DISABLE_INFRARED for faster compilation
+;; RAM:   [==        ]  24.7% (used 80884 bytes from 327680 bytes)
+;; Flash: [==========]  99.0% (used 1556429 bytes from 1572864 bytes)
 
 ; compiled with ESP-IDF 4.4.1
 [env:esp32_16MB_V4_M]
@@ -1303,6 +1305,43 @@ lib_ignore = IRremoteESP8266 ; use with WLED_DISABLE_INFRARED for faster compila
 ;; RAM:   [==        ]  24.3% (used 79524 bytes from 327680 bytes)
 ;; Flash: [========= ]  93.2% (used 1466389 bytes from 1572864 bytes)
 
+;; similar to 4MB_PSRAM_S, but optimized for WROVER-E (chip revision >= 3) that doesn't need any workarounds for PSRAM any more
+;; tl;dr: its faster on PSRAM. But it will not work on all boards.
+[env:esp32_4MB_PSRAM_REV3_S]
+extends = esp32_4MB_V4_S_base
+;board = esp32cam
+board = lolin_d32_pro
+;board_build.partitions = tools/WLED_ESP32_4MB_256KB_FS.csv   ;; Alternative for 4MB flash:   1.8MB firmware, 256KB filesystem (esptool erase_flash needed before changing)
+build_unflags = ${esp32_4MB_V4_S_base.build_unflags}
+  -DARDUINO_EVENT_RUNNING_CORE=1  ;; we want to run wifi on core0, so remove the standard flag
+  -mfix-esp32-psram-cache-issue    ;; this fix is not needed any more for revision 3
+  -mfix-esp32-psram-cache-strategy=memw ;; same as above
+  ;;-Os ;; if you want to be even faster - use together with "-O2" in build_flags. Enable alternative default_partitions (above) to have more program space.
+
+build_flags = ${esp32_4MB_V4_S_base.build_flags}
+  -DARDUINO_EVENT_RUNNING_CORE=0  ;; assign Wifi to core0, to have more CPU on core#1 (arduino loop)
+  -DARDUINO_RUNNING_CORE=1        ;; should be default, but does not hurt
+  ;-DCONFIG_MBEDTLS_DYNAMIC_BUFFER=1 ;; optional - seems to move more buffers into PSRAM
+  ;; enable the next line together with "-Os" in build_unflags (unfortunately --> 104.0%% Flash)
+  ;;-O2 -fno-jump-tables -fno-tree-switch-conversion -Wwrite-strings -fstrict-volatile-bitfields
+  -DBOARD_HAS_PSRAM -D WLED_USE_PSRAM_JSON ;; -D WLED_USE_PSRAM ;; WLED_USE_PSRAM causes major slow-down (slow LEDs) on some ESP32 boards
+  -D WLED_RELEASE_NAME=esp32_4MB_PSRAM_REV3_S
+  -D WLED_WATCHDOG_TIMEOUT=0 #-D WLED_DISABLE_BROWNOUT_DET
+  -D ARDUINO_USB_CDC_ON_BOOT=0 ; needed for arduino-esp32 >=2.0.4; avoids errors on startup
+  -D WLED_DISABLE_LOXONE ;  FLASH 1272 bytes
+  -D WLED_DISABLE_HUESYNC ; RAM 122 bytes; FLASH  6308 bytes
+  -D WLED_DISABLE_ALEXA ;   RAM 116 bytes; FLASH 13524 bytes
+  -D WLED_DISABLE_MQTT ;    RAM 216 bytes; FLASH 16496 bytes
+  -D WLED_DISABLE_INFRARED ;RAM 136 bytes; FLASH 24492 bytes ewowi: disabled to stay below 100%
+  ;-D WLED_DISABLE_ADALIGHT  ;; To disable serial protocols (see upstream #3128)
+  ; -D WLED_ENABLE_DMX
+  ; -D WLED_DEBUG
+  ; -D SR_DEBUG
+  ; -D MIC_LOGGER
+lib_ignore = IRremoteESP8266 ; use with WLED_DISABLE_INFRARED for faster compilation
+;; RAM:   [==        ]  16.8% (used 55124 bytes from 327680 bytes)
+;; Flash: [========= ]  92.7% (used 1458729 bytes from 1572864 bytes)
+
 ;; PSRAM build env that only leaves 300Kb for filesystem (instead of 1MB), but adds 300kB for program space
 [env:esp32_4MB_PSRAM_M]
 extends = esp32_4MB_V4_M_base

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,

usermods/usermod_v2_weather/usermod_v2_weather.h

@@ -35,7 +35,7 @@ uint16_t mode_2DWeather(void) {
       // currentTime = localTime;
     }
     else
-      color = ColorFromPalette(SEGPALETTE, map((uint8_t)weather_temps[x%100], 0, 40, 0, 255), 255, LINEARBLEND);
+      color = ColorFromPalette(SEGPALETTE, map((uint8_t)weather_temps[x%100], 0, 40, 1, 255), 255, LINEARBLEND);
 
     for (int y=0; y<SEGMENT.virtualHeight() * (weather_temps[x%100]-weather_minTemp)/(weather_maxTemp - weather_minTemp); y++) {
       SEGMENT.setPixelColorXY(x, SEGMENT.virtualHeight() - y, color);
@@ -58,7 +58,7 @@ uint16_t mode_2DWeather(void) {
   snprintf(tempString, 5, "%5.2f", currentTemp); // snprintf will prevent overflow
   // Serial.println();
 
-  CRGB color = ColorFromPalette(SEGPALETTE, map((uint8_t)currentTemp, 0, 40, 0, 255), 255, LINEARBLEND);
+  CRGB color = ColorFromPalette(SEGPALETTE, map((uint8_t)currentTemp, 0, 40, 1, 255), 255, LINEARBLEND);
   //really don't understand why this is not working if width < 16 (only works when Serial.println is uncommented ???)
   // uint16_t x = 0;
   // const uint16_t xSpace = (SEGMENT.virtualWidth()<16)?4:5;
@@ -78,17 +78,17 @@ uint16_t mode_2DWeather(void) {
   // // Serial.printf("  %d %d\n", x, xSpace);
 
  if (SEGMENT.virtualWidth() < 16) {
-    SEGMENT.drawCharacter(tempString[0], 0, -2, 5, 8, color);
-    SEGMENT.drawCharacter(tempString[1], 4, -2, 5, 8, color);
+    SEGMENT.drawCharacter(tempString[0], 0, -2, 5, 8, uint32_t(color), BLACK);
+    SEGMENT.drawCharacter(tempString[1], 4, -2, 5, 8, uint32_t(color), BLACK);
     SEGMENT.setPixelColorXY(8, 4, color);
-    SEGMENT.drawCharacter(tempString[3], 9, -2, 5, 8, color);
+    SEGMENT.drawCharacter(tempString[3], 9, -2, 5, 8, uint32_t(color), BLACK);
   }
   else {
-    SEGMENT.drawCharacter(tempString[0], 0, -2, 5, 8, color);
-    SEGMENT.drawCharacter(tempString[1], 5, -2, 5, 8, color);
+    SEGMENT.drawCharacter(tempString[0], 0, -2, 5, 8, uint32_t(color), BLACK);
+    SEGMENT.drawCharacter(tempString[1], 5, -2, 5, 8, uint32_t(color), BLACK);
     SEGMENT.setPixelColorXY(10, 4, color);
-    SEGMENT.drawCharacter(tempString[3], 12, -2, 5, 8, color);
-    SEGMENT.drawCharacter(tempString[4], 17, -2, 5, 8, color);
+    SEGMENT.drawCharacter(tempString[3], 12, -2, 5, 8, uint32_t(color), BLACK);
+    SEGMENT.drawCharacter(tempString[4], 17, -2, 5, 8, uint32_t(color), BLACK);
   }
 
   return FRAMETIME;

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
 
@@ -7067,8 +7082,8 @@ 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