Jan/WLED_MM_Infinity
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

soap effect speedup (MM specific)

Browse Source 
* postpone LEDs pixel drawing to the end of effect -> up to 30% faster
* some minor optimizations: reduce heap fragmentation, use "fast" types
This commit is contained in:
Frank
2025-12-17 16:55:09 +01:00
parent 4ec939cb4c
commit 49e6de33c7
1 changed files with 27 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

39
wled00/FX.cpp
View File

@@ -8738,15 +8738,17 @@ static const char _data_FX_MODE_2DDISTORTIONWAVES[] PROGMEM = "Distortion Waves@
//Idea from https://www.youtube.com/watch?v=DiHBgITrZck&ab_channel=StefanPetrick
// adapted for WLED by @blazoncek, optimization by @dedehai
static void soapPixels(bool isRow, uint8_t *noise3d, CRGB *pixels) {
const uint16_t cols = SEGMENT.virtualWidth();
const uint16_t rows = SEGMENT.virtualHeight();
const uint_fast16_t cols = SEGMENT.virtualWidth();
const uint_fast16_t rows = SEGMENT.virtualHeight();
const auto myXY = [&](int x, int y) { return x + y * cols; };
//const auto abs = [](int x) { return x<0 ? -x : x; }; // WLEDMM not any faster than abs()
const int tRC = isRow ? rows : cols; // transpose if isRow
const int tCR = isRow ? cols : rows; // transpose if isRow
const int amplitude = max(1, (tCR - 8) >> 3) * (1 + (SEGMENT.custom1 >> 5));
const int shift = 0; //(128 - SEGMENT.custom2)*2;
CRGB ledsbuff[tCR];
const size_t maxtCR = max(cols, rows); // WLEDMM always allocate max(cols,rows) buffer, to reduce heap fragmentation
CRGB ledsbuff[/*tCR*/ maxtCR]; // ledsbuff is a VLA in heap (not stack), allocated / removed for each function run - size = max(cols,rows) * sizeof(CRGB)
for (int i = 0; i < tRC; i++) {
int amount = ((int)noise3d[isRow ? i*cols : i] - 128) * amplitude + shift; // use first row/column: XY(0,i)/XY(i,0)
@@ -8779,20 +8781,27 @@ static void soapPixels(bool isRow, uint8_t *noise3d, CRGB *pixels) {
else PixelB = ColorFromPalette(SEGPALETTE, ~noise3d[indxB]*3);
ledsbuff[j] = (PixelA.nscale8(ease8InOutApprox(255 - fraction))) + (PixelB.nscale8(ease8InOutApprox(fraction)));
}
for (int j = 0; j < tCR; j++) {
CRGB c = ledsbuff[j];
if (isRow) std::swap(j,i);
SEGMENT.setPixelColorXY(i, j, pixels[myXY(i,j)] = c);
if (isRow) std::swap(j,i);
}
// upstream code
//for (int j = 0; j < tCR; j++) {
// CRGB c = ledsbuff[j];
// if (isRow) std::swap(j,i);
// SEGMENT.setPixelColorXY(i, j, pixels[myXY(i,j)] = c);
// if (isRow) std::swap(j,i);
//}
// WLEDMM optimization: only copy back pixels, dump to LEDs later -> up to 30% faster
if (isRow)
for (int j = 0; j < tCR; j++) pixels[myXY(j,i)] = ledsbuff[j]; //swap x and y
else
for (int j = 0; j < tCR; j++) pixels[myXY(i,j)] = ledsbuff[j];
}
}
uint16_t mode_2Dsoap() {
if (!strip.isMatrix) return mode_oops(); // not a 2D set-up
const uint16_t cols = SEGMENT.virtualWidth();
const uint16_t rows = SEGMENT.virtualHeight();
const uint_fast16_t cols = SEGMENT.virtualWidth();
const uint_fast16_t rows = SEGMENT.virtualHeight();
const auto myXY = [&](int x, int y) { return x + y * cols; };
const size_t segSize = SEGMENT.width() * SEGMENT.height(); // prevent reallocation if mirrored or grouped
@@ -8810,7 +8819,7 @@ uint16_t mode_2Dsoap() {
// init
if (SEGENV.call == 0) {
random16_set_seed(535); //WLEDMM SuperSync
SEGENV.setUpLeds();
//SEGENV.setUpLeds(); // leds buffer not needed, effect has its own pixel buffer
SEGMENT.fill(BLACK);
}
@@ -8839,6 +8848,12 @@ uint16_t mode_2Dsoap() {
soapPixels(true, noise3d, pixels); // rows
soapPixels(false, noise3d, pixels); // cols
// WLEDMM optimization: dumping out all pixels at once is ~20% faster
for (int i = 0; i < cols; i++) {
for (int j = 0; j < rows; j++) {
SEGMENT.setPixelColorXY(i, j, uint32_t(pixels[myXY(i,j)]) & 0x00FFFFFF);
} }
return FRAMETIME;
}
static const char _data_FX_MODE_2DSOAP[] PROGMEM = "Soap@!,Smoothness,Density;;!;2;pal=11";