Skip to content

Stencil blendingmode and some speed optimisations #4889

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
wants to merge 4 commits into
base: main
Choose a base branch
from
Draft

Stencil blendingmode and some speed optimisations #4889

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
db1a81f
Optimisatins in blendSegment()
DedeHai Aug 30, 2025
9c6ab89
adding new blend mode: "Stencil" and some optimizations
DedeHai Aug 30, 2025
bfc9c5e
cleanup
DedeHai Aug 30, 2025
ae76ae8
cleaner code with a macro, added unused functions back
DedeHai Aug 31, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 1 addition & 2 deletions wled00/FX.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -455,9 +455,8 @@ class Segment {
bool check1 : 1; // checkmark 1
bool check2 : 1; // checkmark 2
bool check3 : 1; // checkmark 3
//uint8_t blendMode : 4; // segment blending modes: top, bottom, add, subtract, difference, multiply, divide, lighten, darken, screen, overlay, hardlight, softlight, dodge, burn
};
uint8_t blendMode; // segment blending modes: top, bottom, add, subtract, difference, multiply, divide, lighten, darken, screen, overlay, hardlight, softlight, dodge, burn
uint8_t blendMode; // segment blending modes: top, bottom, add, subtract, difference, multiply, divide, lighten, darken, screen, overlay, hardlight, softlight, dodge, burn, stencil
char *name; // segment name

// runtime data
Expand Down
94 changes: 57 additions & 37 deletions wled00/FX_fcn.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -369,7 +369,7 @@ void Segment::beginDraw(uint16_t prog) {
// minimum blend time is 100ms maximum is 65535ms
#ifndef WLED_SAVE_RAM
unsigned noOfBlends = ((255U * prog) / 0xFFFFU) - _t->_prevPaletteBlends;
if(noOfBlends > 255) noOfBlends = 255; // safety check
if (noOfBlends > 255) noOfBlends = 255; // safety check
for (unsigned i = 0; i < noOfBlends; i++, _t->_prevPaletteBlends++) nblendPaletteTowardPalette(_t->_palT, Segment::_currentPalette, 48);
Segment::_currentPalette = _t->_palT; // copy transitioning/temporary palette
#else
Expand Down Expand Up @@ -1292,9 +1292,6 @@ void WS2812FX::service() {
}

// https://en.wikipedia.org/wiki/Blend_modes but using a for top layer & b for bottom layer
static uint8_t _top (uint8_t a, uint8_t b) { return a; }
static uint8_t _bottom (uint8_t a, uint8_t b) { return b; }
static uint8_t _add (uint8_t a, uint8_t b) { unsigned t = a + b; return t > 255 ? 255 : t; }
static uint8_t _subtract (uint8_t a, uint8_t b) { return b > a ? (b - a) : 0; }
static uint8_t _difference(uint8_t a, uint8_t b) { return b > a ? (b - a) : (a - b); }
static uint8_t _average (uint8_t a, uint8_t b) { return (a + b) >> 1; }
Expand All @@ -1317,23 +1314,40 @@ static uint8_t _softlight (uint8_t a, uint8_t b) { return (b * b * (255 - 2 * a)
static uint8_t _dodge (uint8_t a, uint8_t b) { return _divide(~a,b); }
static uint8_t _burn (uint8_t a, uint8_t b) { return ~_divide(a,~b); }

void WS2812FX::blendSegment(const Segment &topSegment) const {

typedef uint8_t(*FuncType)(uint8_t, uint8_t);
FuncType funcs[] = {
_top, _bottom,
_add, _subtract, _difference, _average,
_multiply, _divide, _lighten, _darken, _screen, _overlay,
_hardlight, _softlight, _dodge, _burn
};

const size_t blendMode = topSegment.blendMode < (sizeof(funcs) / sizeof(FuncType)) ? topSegment.blendMode : 0;
const auto func = funcs[blendMode]; // blendMode % (sizeof(funcs) / sizeof(FuncType))
const auto blend = [&](uint32_t top, uint32_t bottom){ return RGBW32(func(R(top),R(bottom)), func(G(top),G(bottom)), func(B(top),B(bottom)), func(W(top),W(bottom))); };
static uint32_t segblend(CRGBW tcol, CRGBW bcol, CRGBW bgcol, uint8_t blendMode) {
CRGBW tC(tcol); CRGBW bC(bcol); CRGBW c; // note: using aliases shrinks code size for some weird compiler reasons, no speed difference in tests
// note2: using CRGBW instead of uint32_t improves speed as well as code size
switch (blendMode) {
case 0: return tcol.color32; // Top
case 1: return bcol.color32; // Bottom
case 2: return color_add(tcol, bcol, false); // Add
case 3: c.r = _subtract(tC.r, bC.r); c.g = _subtract(tC.g, bC.g); c.b = _subtract(tC.b, bC.b); c.w = _subtract(tC.w, bC.w); break; // Subtract
Copy link
Member

@willmmiles willmmiles Aug 31, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A bit of a nitpick, but I'd recommend using an inline metafunction or a macro to do the expansion for each channel.

Copy link
Collaborator Author

@DedeHai DedeHai Aug 31, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

that would clean up the code significantly, I will check how to do that here.

Copy link
Collaborator Author

@DedeHai DedeHai Aug 31, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok done.

case 4: c.r = _difference(tC.r, bC.r); c.g = _difference(tC.g, bC.g); c.b = _difference(tC.b, bC.b); c.w = _difference(tC.w, bC.w); break; // Difference
case 5: c.r = _average(tC.r, bC.r); c.g = _average(tC.g, bC.g); c.b = _average(tC.b, bC.b); c.w = _average(tC.w, bC.w); break; // Average
case 6: c.r = _multiply(tC.r, bC.r); c.g = _multiply(tC.g, bC.g); c.b = _multiply(tC.b, bC.b); c.w = _multiply(tC.w, bC.w); break; // Multiply
case 7: c.r = _divide(tC.r, bC.r); c.g = _divide(tC.g, bC.g); c.b = _divide(tC.b, bC.b); c.w = _divide(tC.w, bC.w); break; // Divide
case 8: c.r = _lighten(tC.r, bC.r); c.g = _lighten(tC.g, bC.g); c.b = _lighten(tC.b, bC.b); c.w = _lighten(tC.w, bC.w); break; // Lighten
case 9: c.r = _darken(tC.r, bC.r); c.g = _darken(tC.g, bC.g); c.b = _darken(tC.b, bC.b); c.w = _darken(tC.w, bC.w); break; // Darken
case 10: c.r = _screen(tC.r, bC.r); c.g = _screen(tC.g, bC.g); c.b = _screen(tC.b, bC.b); c.w = _screen(tC.w, bC.w); break; // Screen
case 11: c.r = _overlay(tC.r, bC.r); c.g = _overlay(tC.g, bC.g); c.b = _overlay(tC.b, bC.b); c.w = _overlay(tC.w, bC.w); break; // Overlay
case 12: c.r = _hardlight(tC.r, bC.r); c.g = _hardlight(tC.g, bC.g); c.b = _hardlight(tC.b, bC.b); c.w = _hardlight(tC.w, bC.w); break; // Hardlight
case 13: c.r = _softlight(tC.r, bC.r); c.g = _softlight(tC.g, bC.g); c.b = _softlight(tC.b, bC.b); c.w = _softlight(tC.w, bC.w); break; // Softlight
case 14: c.r = _dodge(tC.r, bC.r); c.g = _dodge(tC.g, bC.g); c.b = _dodge(tC.b, bC.b); c.w = _dodge(tC.w, bC.w); break; // Dodge
case 15: c.r = _burn(tC.r, bC.r); c.g = _burn(tC.g, bC.g); c.b = _burn(tC.b, bC.b); c.w = _burn(tC.w, bC.w); break; // Burn
// note: stencil mode is a special case: it works only on full color comparison and wont work correctly on a colorchannel base
// enumerate to 32 to allow future additions above in case a function array will be used again
case 32: return tcol.color32 == bgcol.color32 ? bcol.color32 : tcol.color32; // Stencil: backgroundcolor -> transparent, use top color otherwise
default: return tcol.color32; // fallback to Top
}
return c.color32;
};

void WS2812FX::blendSegment(const Segment &topSegment) const {
const uint8_t blendMode = topSegment.blendMode;
const int length = topSegment.length(); // physical segment length (counts all pixels in 2D segment)
const int width = topSegment.width();
const int height = topSegment.height();
const uint32_t bgColor = topSegment.colors[1]; // background color for blend modes that need it (e.g. stencil)
const auto XY = [](int x, int y){ return x + y*Segment::maxWidth; };
const size_t matrixSize = Segment::maxWidth * Segment::maxHeight;
const size_t startIndx = XY(topSegment.start, topSegment.startY);
Expand Down Expand Up @@ -1408,23 +1422,24 @@ void WS2812FX::blendSegment(const Segment &topSegment) const {
const Segment *segO = topSegment.getOldSegment();
const int oCols = segO ? segO->virtualWidth() : nCols;
const int oRows = segO ? segO->virtualHeight() : nRows;
bool applyMirror = topSegment.mirror || topSegment.mirror_y;

const auto setMirroredPixel = [&](int x, int y, uint32_t c, uint8_t o) {
const int baseX = topSegment.start + x;
const int baseY = topSegment.startY + y;
size_t indx = XY(baseX, baseY); // absolute address on strip
_pixels[indx] = color_blend(_pixels[indx], blend(c, _pixels[indx]), o);
_pixels[indx] = color_blend(_pixels[indx], segblend(c, _pixels[indx], bgColor, blendMode), o);
if (_pixelCCT) _pixelCCT[indx] = cct;
// Apply mirroring
if (topSegment.mirror || topSegment.mirror_y) {
if (applyMirror) {
const int mirrorX = topSegment.start + width - x - 1;
const int mirrorY = topSegment.startY + height - y - 1;
const size_t idxMX = XY(topSegment.transpose ? baseX : mirrorX, topSegment.transpose ? mirrorY : baseY);
const size_t idxMY = XY(topSegment.transpose ? mirrorX : baseX, topSegment.transpose ? baseY : mirrorY);
const size_t idxMM = XY(mirrorX, mirrorY);
if (topSegment.mirror) _pixels[idxMX] = color_blend(_pixels[idxMX], blend(c, _pixels[idxMX]), o);
if (topSegment.mirror_y) _pixels[idxMY] = color_blend(_pixels[idxMY], blend(c, _pixels[idxMY]), o);
if (topSegment.mirror && topSegment.mirror_y) _pixels[idxMM] = color_blend(_pixels[idxMM], blend(c, _pixels[idxMM]), o);
if (topSegment.mirror) _pixels[idxMX] = color_blend(_pixels[idxMX], segblend(c, _pixels[idxMX], bgColor, blendMode), o);
if (topSegment.mirror_y) _pixels[idxMY] = color_blend(_pixels[idxMY], segblend(c, _pixels[idxMY], bgColor, blendMode), o);
if (topSegment.mirror && topSegment.mirror_y) _pixels[idxMM] = color_blend(_pixels[idxMM], segblend(c, _pixels[idxMM], bgColor, blendMode), o);
if (_pixelCCT) {
if (topSegment.mirror) _pixelCCT[idxMX] = cct;
if (topSegment.mirror_y) _pixelCCT[idxMY] = cct;
Expand All @@ -1436,7 +1451,16 @@ void WS2812FX::blendSegment(const Segment &topSegment) const {
// if we blend using "push" style we need to "shift" canvas to left/right/up/down
unsigned offsetX = (blendingStyle == BLEND_STYLE_PUSH_UP || blendingStyle == BLEND_STYLE_PUSH_DOWN) ? 0 : progInv * nCols / 0xFFFFU;
unsigned offsetY = (blendingStyle == BLEND_STYLE_PUSH_LEFT || blendingStyle == BLEND_STYLE_PUSH_RIGHT) ? 0 : progInv * nRows / 0xFFFFU;

const unsigned groupLen = topSegment.groupLength();
bool applyReverse = topSegment.reverse || topSegment.reverse_y || topSegment.transpose;
int pushOffsetX = 0, pushOffsetY = 0;
// if we blend using "push" style we need to "shift" canvas to left/right/up/down
switch (blendingStyle) {
case BLEND_STYLE_PUSH_RIGHT: pushOffsetX = offsetX; break;
case BLEND_STYLE_PUSH_LEFT: pushOffsetX = -offsetX + nCols; break;
case BLEND_STYLE_PUSH_DOWN: pushOffsetY = offsetY; break;
case BLEND_STYLE_PUSH_UP: pushOffsetY = -offsetY + nRows; break;
}
// we only traverse new segment, not old one
for (int r = 0; r < nRows; r++) for (int c = 0; c < nCols; c++) {
const bool clipped = topSegment.isPixelXYClipped(c, r);
Expand All @@ -1446,13 +1470,8 @@ void WS2812FX::blendSegment(const Segment &topSegment) const {
int vRows = seg == segO ? oRows : nRows; // old segment may have different dimensions
int x = c;
int y = r;
// if we blend using "push" style we need to "shift" canvas to left/right/up/down
switch (blendingStyle) {
case BLEND_STYLE_PUSH_RIGHT: x = (x + offsetX) % nCols; break;
case BLEND_STYLE_PUSH_LEFT: x = (x - offsetX + nCols) % nCols; break;
case BLEND_STYLE_PUSH_DOWN: y = (y + offsetY) % nRows; break;
case BLEND_STYLE_PUSH_UP: y = (y - offsetY + nRows) % nRows; break;
}
if (pushOffsetX != 0) x = (x + pushOffsetX) % nCols;
if (pushOffsetY != 0) y = (y + pushOffsetY) % nRows;
uint32_t c_a = BLACK;
if (x < vCols && y < vRows) c_a = seg->getPixelColorRaw(x + y*vCols); // will get clipped pixel from old segment or unclipped pixel from new segment
if (segO && blendingStyle == BLEND_STYLE_FADE
Expand All @@ -1469,11 +1488,12 @@ void WS2812FX::blendSegment(const Segment &topSegment) const {
// map it into frame buffer
x = c; // restore coordiates if we were PUSHing
y = r;
if (topSegment.reverse ) x = nCols - x - 1;
if (topSegment.reverse_y) y = nRows - y - 1;
if (topSegment.transpose) std::swap(x,y); // swap X & Y if segment transposed
if (applyReverse) {
if (topSegment.reverse ) x = nCols - x - 1;
if (topSegment.reverse_y) y = nRows - y - 1;
if (topSegment.transpose) std::swap(x,y); // swap X & Y if segment transposed
}
// expand pixel
const unsigned groupLen = topSegment.groupLength();
if (groupLen == 1) {
setMirroredPixel(x, y, c_a, opacity);
} else {
Expand Down Expand Up @@ -1502,12 +1522,12 @@ void WS2812FX::blendSegment(const Segment &topSegment) const {
unsigned indxM = topSegment.stop - i - 1;
indxM += topSegment.offset; // offset/phase
if (indxM >= topSegment.stop) indxM -= length; // wrap
_pixels[indxM] = color_blend(_pixels[indxM], blend(c, _pixels[indxM]), o);
_pixels[indxM] = color_blend(_pixels[indxM], segblend(c, _pixels[indxM], bgColor, blendMode), o);
if (_pixelCCT) _pixelCCT[indxM] = cct;
}
indx += topSegment.offset; // offset/phase
if (indx >= topSegment.stop) indx -= length; // wrap
_pixels[indx] = color_blend(_pixels[indx], blend(c, _pixels[indx]), o);
_pixels[indx] = color_blend(_pixels[indx], segblend(c, _pixels[indx], bgColor, blendMode), o);
if (_pixelCCT) _pixelCCT[indx] = cct;
};

Expand Down Expand Up @@ -1590,7 +1610,7 @@ void WS2812FX::show() {
}

uint32_t c = _pixels[i]; // need a copy, do not modify _pixels directly (no byte access allowed on ESP32)
if(c > 0 && !(realtimeMode && arlsDisableGammaCorrection))
if (c > 0 && !(realtimeMode && arlsDisableGammaCorrection))
c = gamma32(c); // apply gamma correction if enabled note: applying gamma after brightness has too much color loss
BusManager::setPixelColor(getMappedPixelIndex(i), c);
}
Expand Down
3 changes: 2 additions & 1 deletion wled00/data/index.js
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -825,6 +825,7 @@ function populateSegments(s)
`<option value="13" ${inst.bm==13?' selected':''}>Soft Light</option>`+
`<option value="14" ${inst.bm==14?' selected':''}>Dodge</option>`+
`<option value="15" ${inst.bm==15?' selected':''}>Burn</option>`+
`<option value="32" ${inst.bm==32?' selected':''}>Stencil</option>`+
`</select></div>`+
`</div>`;
let sndSim = `<div data-snd="si" class="lbl-s hide">Sound sim<br>`+
Expand Down Expand Up @@ -2358,7 +2359,7 @@ function setSi(s)

function setBm(s)
{
var value = gId(`seg${s}bm`).selectedIndex;
var value = gId(`seg${s}bm`).value;
var obj = {"seg": {"id": s, "bm": value}};
requestJson(obj);
}
Expand Down
4 changes: 1 addition & 3 deletions wled00/json.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -302,9 +302,7 @@ static bool deserializeSegment(JsonObject elem, byte it, byte presetId = 0)
seg.check2 = getBoolVal(elem["o2"], seg.check2);
seg.check3 = getBoolVal(elem["o3"], seg.check3);

uint8_t blend = seg.blendMode;
getVal(elem["bm"], blend, 0, 15); // we can't pass reference to bitfield
seg.blendMode = constrain(blend, 0, 15);
getVal(elem["bm"], seg.blendMode);

JsonArray iarr = elem[F("i")]; //set individual LEDs
if (!iarr.isNull()) {
Expand Down