/* * Copyright (C) 2007 The Android Open Source Project * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ #define LOG_TAG "SurfaceFlinger" #include <stdlib.h> #include <stdint.h> #include <sys/types.h> #include <utils/Errors.h> #include <utils/Log.h> #include <GLES/gl.h> #include <GLES/glext.h> #include <hardware/hardware.h> #include "clz.h" #include "LayerBase.h" #include "LayerBlur.h" #include "SurfaceFlinger.h" #include "DisplayHardware/DisplayHardware.h" // We don't honor the premultiplied alpha flags, which means that // premultiplied surface may be composed using a non-premultiplied // equation. We do this because it may be a lot faster on some hardware // The correct value is HONOR_PREMULTIPLIED_ALPHA = 1 #define HONOR_PREMULTIPLIED_ALPHA 0 namespace android { // --------------------------------------------------------------------------- const uint32_t LayerBase::typeInfo = 1; const char* const LayerBase::typeID = "LayerBase"; const uint32_t LayerBaseClient::typeInfo = LayerBase::typeInfo | 2; const char* const LayerBaseClient::typeID = "LayerBaseClient"; // --------------------------------------------------------------------------- Vector<GLuint> LayerBase::deletedTextures; int32_t LayerBase::sIdentity = 0; LayerBase::LayerBase(SurfaceFlinger* flinger, DisplayID display) : dpy(display), contentDirty(false), mFlinger(flinger), mTransformed(false), mOrientation(0), mCanUseCopyBit(false), mTransactionFlags(0), mPremultipliedAlpha(true), mIdentity(uint32_t(android_atomic_inc(&sIdentity))), mInvalidate(0) { const DisplayHardware& hw(flinger->graphicPlane(0).displayHardware()); mFlags = hw.getFlags(); } LayerBase::~LayerBase() { } const GraphicPlane& LayerBase::graphicPlane(int dpy) const { return mFlinger->graphicPlane(dpy); } GraphicPlane& LayerBase::graphicPlane(int dpy) { return mFlinger->graphicPlane(dpy); } void LayerBase::initStates(uint32_t w, uint32_t h, uint32_t flags) { uint32_t layerFlags = 0; if (flags & ISurfaceComposer::eHidden) layerFlags = ISurfaceComposer::eLayerHidden; if (flags & ISurfaceComposer::eNonPremultiplied) mPremultipliedAlpha = false; mCurrentState.z = 0; mCurrentState.w = w; mCurrentState.h = h; mCurrentState.alpha = 0xFF; mCurrentState.flags = layerFlags; mCurrentState.sequence = 0; mCurrentState.transform.set(0, 0); // drawing state & current state are identical mDrawingState = mCurrentState; } void LayerBase::commitTransaction(bool skipSize) { const uint32_t w = mDrawingState.w; const uint32_t h = mDrawingState.h; mDrawingState = mCurrentState; if (skipSize) { mDrawingState.w = w; mDrawingState.h = h; } } void LayerBase::forceVisibilityTransaction() { // this can be called without SurfaceFlinger.mStateLock, but if we // can atomically increment the sequence number, it doesn't matter. android_atomic_inc(&mCurrentState.sequence); requestTransaction(); } bool LayerBase::requestTransaction() { int32_t old = setTransactionFlags(eTransactionNeeded); return ((old & eTransactionNeeded) == 0); } uint32_t LayerBase::getTransactionFlags(uint32_t flags) { return android_atomic_and(~flags, &mTransactionFlags) & flags; } uint32_t LayerBase::setTransactionFlags(uint32_t flags) { return android_atomic_or(flags, &mTransactionFlags); } void LayerBase::setSizeChanged(uint32_t w, uint32_t h) { } bool LayerBase::setPosition(int32_t x, int32_t y) { if (mCurrentState.transform.tx() == x && mCurrentState.transform.ty() == y) return false; mCurrentState.sequence++; mCurrentState.transform.set(x, y); requestTransaction(); return true; } bool LayerBase::setLayer(uint32_t z) { if (mCurrentState.z == z) return false; mCurrentState.sequence++; mCurrentState.z = z; requestTransaction(); return true; } bool LayerBase::setSize(uint32_t w, uint32_t h) { if (mCurrentState.w == w && mCurrentState.h == h) return false; setSizeChanged(w, h); mCurrentState.w = w; mCurrentState.h = h; requestTransaction(); return true; } bool LayerBase::setAlpha(uint8_t alpha) { if (mCurrentState.alpha == alpha) return false; mCurrentState.sequence++; mCurrentState.alpha = alpha; requestTransaction(); return true; } bool LayerBase::setMatrix(const layer_state_t::matrix22_t& matrix) { // TODO: check the matrix has changed mCurrentState.sequence++; mCurrentState.transform.set( matrix.dsdx, matrix.dsdy, matrix.dtdx, matrix.dtdy); requestTransaction(); return true; } bool LayerBase::setTransparentRegionHint(const Region& transparent) { // TODO: check the region has changed mCurrentState.sequence++; mCurrentState.transparentRegion = transparent; requestTransaction(); return true; } bool LayerBase::setFlags(uint8_t flags, uint8_t mask) { const uint32_t newFlags = (mCurrentState.flags & ~mask) | (flags & mask); if (mCurrentState.flags == newFlags) return false; mCurrentState.sequence++; mCurrentState.flags = newFlags; requestTransaction(); return true; } Rect LayerBase::visibleBounds() const { return mTransformedBounds; } void LayerBase::setVisibleRegion(const Region& visibleRegion) { // always called from main thread visibleRegionScreen = visibleRegion; } void LayerBase::setCoveredRegion(const Region& coveredRegion) { // always called from main thread coveredRegionScreen = coveredRegion; } uint32_t LayerBase::doTransaction(uint32_t flags) { const Layer::State& front(drawingState()); const Layer::State& temp(currentState()); if (temp.sequence != front.sequence) { // invalidate and recompute the visible regions if needed flags |= eVisibleRegion; this->contentDirty = true; } // Commit the transaction commitTransaction(flags & eRestartTransaction); return flags; } Point LayerBase::getPhysicalSize() const { const Layer::State& front(drawingState()); return Point(front.w, front.h); } void LayerBase::validateVisibility(const Transform& planeTransform) { const Layer::State& s(drawingState()); const Transform tr(planeTransform * s.transform); const bool transformed = tr.transformed(); const Point size(getPhysicalSize()); uint32_t w = size.x; uint32_t h = size.y; tr.transform(mVertices[0], 0, 0); tr.transform(mVertices[1], 0, h); tr.transform(mVertices[2], w, h); tr.transform(mVertices[3], w, 0); if (UNLIKELY(transformed)) { // NOTE: here we could also punt if we have too many rectangles // in the transparent region if (tr.preserveRects()) { // transform the transparent region transparentRegionScreen = tr.transform(s.transparentRegion); } else { // transformation too complex, can't do the transparent region // optimization. transparentRegionScreen.clear(); } } else { transparentRegionScreen = s.transparentRegion; } // cache a few things... mOrientation = tr.getOrientation(); mTransformedBounds = tr.makeBounds(w, h); mTransformed = transformed; mLeft = tr.tx(); mTop = tr.ty(); // see if we can/should use 2D h/w with the new configuration mCanUseCopyBit = false; copybit_device_t* copybit = mFlinger->getBlitEngine(); if (copybit) { const int step = copybit->get(copybit, COPYBIT_ROTATION_STEP_DEG); const int scaleBits = copybit->get(copybit, COPYBIT_SCALING_FRAC_BITS); mCanUseCopyBit = true; if ((mOrientation < 0) && (step > 1)) { // arbitrary orientations not supported mCanUseCopyBit = false; } else if ((mOrientation > 0) && (step > 90)) { // 90 deg rotations not supported mCanUseCopyBit = false; } else if ((tr.getType() & SkMatrix::kScale_Mask) && (scaleBits < 12)) { // arbitrary scaling not supported mCanUseCopyBit = false; } #if HONOR_PREMULTIPLIED_ALPHA else if (needsBlending() && mPremultipliedAlpha) { // pre-multiplied alpha not supported mCanUseCopyBit = false; } #endif else { // here, we determined we can use copybit if (tr.getType() & SkMatrix::kScale_Mask) { // and we have scaling if (!transparentRegionScreen.isRect()) { // we punt because blending is cheap (h/w) and the region is // complex, which may causes artifacts when copying // scaled content transparentRegionScreen.clear(); } } } } } void LayerBase::lockPageFlip(bool& recomputeVisibleRegions) { } void LayerBase::unlockPageFlip( const Transform& planeTransform, Region& outDirtyRegion) { if ((android_atomic_and(~1, &mInvalidate)&1) == 1) { outDirtyRegion.orSelf(visibleRegionScreen); } } void LayerBase::finishPageFlip() { } void LayerBase::invalidate() { if ((android_atomic_or(1, &mInvalidate)&1) == 0) { mFlinger->signalEvent(); } } void LayerBase::drawRegion(const Region& reg) const { Region::iterator iterator(reg); if (iterator) { Rect r; const DisplayHardware& hw(graphicPlane(0).displayHardware()); const int32_t fbWidth = hw.getWidth(); const int32_t fbHeight = hw.getHeight(); const GLshort vertices[][2] = { { 0, 0 }, { fbWidth, 0 }, { fbWidth, fbHeight }, { 0, fbHeight } }; glVertexPointer(2, GL_SHORT, 0, vertices); while (iterator.iterate(&r)) { const GLint sy = fbHeight - (r.top + r.height()); glScissor(r.left, sy, r.width(), r.height()); glDrawArrays(GL_TRIANGLE_FAN, 0, 4); } } } void LayerBase::draw(const Region& inClip) const { // invalidate the region we'll update Region clip(inClip); // copy-on-write, so no-op most of the time // Remove the transparent area from the clipping region const State& s = drawingState(); if (LIKELY(!s.transparentRegion.isEmpty())) { clip.subtract(transparentRegionScreen); if (clip.isEmpty()) { // usually this won't happen because this should be taken care of // by SurfaceFlinger::computeVisibleRegions() return; } } // reset GL state glEnable(GL_SCISSOR_TEST); onDraw(clip); /* glDisable(GL_TEXTURE_2D); glDisable(GL_DITHER); glEnable(GL_BLEND); glBlendFunc(GL_ONE, GL_ONE_MINUS_SRC_ALPHA); glColor4x(0, 0x8000, 0, 0x10000); drawRegion(transparentRegionScreen); glDisable(GL_BLEND); */ } GLuint LayerBase::createTexture() const { GLuint textureName = -1; glGenTextures(1, &textureName); glBindTexture(GL_TEXTURE_2D, textureName); glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); if (mFlags & DisplayHardware::SLOW_CONFIG) { glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); } else { glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); } return textureName; } void LayerBase::clearWithOpenGL(const Region& clip) const { const DisplayHardware& hw(graphicPlane(0).displayHardware()); const uint32_t fbHeight = hw.getHeight(); glColor4x(0,0,0,0); glDisable(GL_TEXTURE_2D); glDisable(GL_BLEND); glDisable(GL_DITHER); Rect r; Region::iterator iterator(clip); if (iterator) { glEnable(GL_SCISSOR_TEST); glVertexPointer(2, GL_FIXED, 0, mVertices); while (iterator.iterate(&r)) { const GLint sy = fbHeight - (r.top + r.height()); glScissor(r.left, sy, r.width(), r.height()); glDrawArrays(GL_TRIANGLE_FAN, 0, 4); } } } void LayerBase::drawWithOpenGL(const Region& clip, GLint textureName, const GGLSurface& t, int transform) const { const DisplayHardware& hw(graphicPlane(0).displayHardware()); const uint32_t fbHeight = hw.getHeight(); const State& s(drawingState()); // bind our texture validateTexture(textureName); glEnable(GL_TEXTURE_2D); // Dithering... if (s.flags & ISurfaceComposer::eLayerDither) { glEnable(GL_DITHER); } else { glDisable(GL_DITHER); } if (UNLIKELY(s.alpha < 0xFF)) { // We have an alpha-modulation. We need to modulate all // texture components by alpha because we're always using // premultiplied alpha. // If the texture doesn't have an alpha channel we can // use REPLACE and switch to non premultiplied alpha // blending (SRCA/ONE_MINUS_SRCA). GLenum env, src; if (needsBlending()) { env = GL_MODULATE; src = mPremultipliedAlpha ? GL_ONE : GL_SRC_ALPHA; } else { env = GL_REPLACE; src = GL_SRC_ALPHA; } const GGLfixed alpha = (s.alpha << 16)/255; glColor4x(alpha, alpha, alpha, alpha); glEnable(GL_BLEND); glBlendFunc(src, GL_ONE_MINUS_SRC_ALPHA); glTexEnvx(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, env); } else { glTexEnvx(GL_TEXTURE_ENV, GL_TEXTURE_ENV_MODE, GL_REPLACE); glColor4x(0x10000, 0x10000, 0x10000, 0x10000); if (needsBlending()) { GLenum src = mPremultipliedAlpha ? GL_ONE : GL_SRC_ALPHA; glEnable(GL_BLEND); glBlendFunc(src, GL_ONE_MINUS_SRC_ALPHA); } else { glDisable(GL_BLEND); } } if (UNLIKELY(transformed() || !(mFlags & DisplayHardware::DRAW_TEXTURE_EXTENSION) )) { //StopWatch watch("GL transformed"); Region::iterator iterator(clip); if (iterator) { // always use high-quality filtering with fast configurations bool fast = !(mFlags & DisplayHardware::SLOW_CONFIG); if (!fast && s.flags & ISurfaceComposer::eLayerFilter) { glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR); glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR); } const GLfixed texCoords[4][2] = { { 0, 0 }, { 0, 0x10000 }, { 0x10000, 0x10000 }, { 0x10000, 0 } }; glMatrixMode(GL_TEXTURE); glLoadIdentity(); if (transform == HAL_TRANSFORM_ROT_90) { glTranslatef(0, 1, 0); glRotatef(-90, 0, 0, 1); } if (!(mFlags & DisplayHardware::NPOT_EXTENSION)) { // find the smallest power-of-two that will accommodate our surface GLuint tw = 1 << (31 - clz(t.width)); GLuint th = 1 << (31 - clz(t.height)); if (tw < t.width) tw <<= 1; if (th < t.height) th <<= 1; // this divide should be relatively fast because it's // a power-of-two (optimized path in libgcc) GLfloat ws = GLfloat(t.width) /tw; GLfloat hs = GLfloat(t.height)/th; glScalef(ws, hs, 1.0f); } glEnableClientState(GL_TEXTURE_COORD_ARRAY); glVertexPointer(2, GL_FIXED, 0, mVertices); glTexCoordPointer(2, GL_FIXED, 0, texCoords); Rect r; while (iterator.iterate(&r)) { const GLint sy = fbHeight - (r.top + r.height()); glScissor(r.left, sy, r.width(), r.height()); glDrawArrays(GL_TRIANGLE_FAN, 0, 4); } if (!fast && s.flags & ISurfaceComposer::eLayerFilter) { glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); glTexParameterx(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); } glDisableClientState(GL_TEXTURE_COORD_ARRAY); } } else { Region::iterator iterator(clip); if (iterator) { Rect r; GLint crop[4] = { 0, t.height, t.width, -t.height }; glTexParameteriv(GL_TEXTURE_2D, GL_TEXTURE_CROP_RECT_OES, crop); int x = tx(); int y = ty(); y = fbHeight - (y + t.height); while (iterator.iterate(&r)) { const GLint sy = fbHeight - (r.top + r.height()); glScissor(r.left, sy, r.width(), r.height()); glDrawTexiOES(x, y, 0, t.width, t.height); } } } } void LayerBase::validateTexture(GLint textureName) const { glBindTexture(GL_TEXTURE_2D, textureName); // TODO: reload the texture if needed // this is currently done in loadTexture() below } void LayerBase::loadTexture(const Region& dirty, GLint textureName, const GGLSurface& t, GLuint& textureWidth, GLuint& textureHeight) const { // TODO: defer the actual texture reload until LayerBase::validateTexture // is called. uint32_t flags = mFlags; glBindTexture(GL_TEXTURE_2D, textureName); GLuint tw = t.width; GLuint th = t.height; /* * In OpenGL ES we can't specify a stride with glTexImage2D (however, * GL_UNPACK_ALIGNMENT is 4, which in essence allows a limited form of * stride). * So if the stride here isn't representable with GL_UNPACK_ALIGNMENT, we * need to do something reasonable (here creating a bigger texture). * * extra pixels = (((stride - width) * pixelsize) / GL_UNPACK_ALIGNMENT); * * This situation doesn't happen often, but some h/w have a limitation * for their framebuffer (eg: must be multiple of 8 pixels), and * we need to take that into account when using these buffers as * textures. * * This should never be a problem with POT textures */ tw += (((t.stride - tw) * bytesPerPixel(t.format)) / 4); /* * round to POT if needed */ GLuint texture_w = tw; GLuint texture_h = th; if (!(flags & DisplayHardware::NPOT_EXTENSION)) { // find the smallest power-of-two that will accommodate our surface texture_w = 1 << (31 - clz(t.width)); texture_h = 1 << (31 - clz(t.height)); if (texture_w < t.width) texture_w <<= 1; if (texture_h < t.height) texture_h <<= 1; if (texture_w != tw || texture_h != th) { // we can't use DIRECT_TEXTURE since we changed the size // of the texture flags &= ~DisplayHardware::DIRECT_TEXTURE; } } if (flags & DisplayHardware::DIRECT_TEXTURE) { // here we're guaranteed that texture_{w|h} == t{w|h} if (t.format == GGL_PIXEL_FORMAT_RGB_565) { glTexImage2D(GL_DIRECT_TEXTURE_2D_QUALCOMM, 0, GL_RGB, tw, th, 0, GL_RGB, GL_UNSIGNED_SHORT_5_6_5, t.data); } else if (t.format == GGL_PIXEL_FORMAT_RGBA_4444) { glTexImage2D(GL_DIRECT_TEXTURE_2D_QUALCOMM, 0, GL_RGBA, tw, th, 0, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4, t.data); } else if (t.format == GGL_PIXEL_FORMAT_RGBA_8888) { glTexImage2D(GL_DIRECT_TEXTURE_2D_QUALCOMM, 0, GL_RGBA, tw, th, 0, GL_RGBA, GL_UNSIGNED_BYTE, t.data); } else if (t.format == GGL_PIXEL_FORMAT_BGRA_8888) { // TODO: add GL_BGRA extension } else { // oops, we don't handle this format, try the regular path goto regular; } textureWidth = tw; textureHeight = th; } else { regular: Rect bounds(dirty.bounds()); GLvoid* data = 0; if (texture_w!=textureWidth || texture_h!=textureHeight) { // texture size changed, we need to create a new one if (!textureWidth || !textureHeight) { // this is the first time, load the whole texture if (texture_w==tw && texture_h==th) { // we can do it one pass data = t.data; } else { // we have to create the texture first because it // doesn't match the size of the buffer bounds.set(Rect(tw, th)); } } if (t.format == GGL_PIXEL_FORMAT_RGB_565) { glTexImage2D(GL_TEXTURE_2D, 0, GL_RGB, texture_w, texture_h, 0, GL_RGB, GL_UNSIGNED_SHORT_5_6_5, data); } else if (t.format == GGL_PIXEL_FORMAT_RGBA_4444) { glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, texture_w, texture_h, 0, GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4, data); } else if (t.format == GGL_PIXEL_FORMAT_RGBA_8888) { glTexImage2D(GL_TEXTURE_2D, 0, GL_RGBA, texture_w, texture_h, 0, GL_RGBA, GL_UNSIGNED_BYTE, data); } else if ( t.format == GGL_PIXEL_FORMAT_YCbCr_422_SP || t.format == GGL_PIXEL_FORMAT_YCbCr_420_SP) { // just show the Y plane of YUV buffers data = t.data; glTexImage2D(GL_TEXTURE_2D, 0, GL_LUMINANCE, texture_w, texture_h, 0, GL_LUMINANCE, GL_UNSIGNED_BYTE, data); } else { // oops, we don't handle this format! LOGE("layer %p, texture=%d, using format %d, which is not " "supported by the GL", this, textureName, t.format); textureName = -1; } textureWidth = texture_w; textureHeight = texture_h; } if (!data && textureName>=0) { if (t.format == GGL_PIXEL_FORMAT_RGB_565) { glTexSubImage2D(GL_TEXTURE_2D, 0, 0, bounds.top, t.width, bounds.height(), GL_RGB, GL_UNSIGNED_SHORT_5_6_5, t.data + bounds.top*t.width*2); } else if (t.format == GGL_PIXEL_FORMAT_RGBA_4444) { glTexSubImage2D(GL_TEXTURE_2D, 0, 0, bounds.top, t.width, bounds.height(), GL_RGBA, GL_UNSIGNED_SHORT_4_4_4_4, t.data + bounds.top*t.width*2); } else if (t.format == GGL_PIXEL_FORMAT_RGBA_8888) { glTexSubImage2D(GL_TEXTURE_2D, 0, 0, bounds.top, t.width, bounds.height(), GL_RGBA, GL_UNSIGNED_BYTE, t.data + bounds.top*t.width*4); } } } } bool LayerBase::canUseCopybit() const { return mCanUseCopyBit; } // --------------------------------------------------------------------------- LayerBaseClient::LayerBaseClient(SurfaceFlinger* flinger, DisplayID display, Client* c, int32_t i) : LayerBase(flinger, display), client(c), lcblk( c ? &(c->ctrlblk->layers[i]) : 0 ), mIndex(i) { if (client) { client->bindLayer(this, i); // Initialize this layer's control block memset(this->lcblk, 0, sizeof(layer_cblk_t)); this->lcblk->identity = mIdentity; Region::writeEmpty(&(this->lcblk->region[0]), sizeof(flat_region_t)); Region::writeEmpty(&(this->lcblk->region[1]), sizeof(flat_region_t)); } } LayerBaseClient::~LayerBaseClient() { if (client) { client->free(mIndex); } } int32_t LayerBaseClient::serverIndex() const { if (client) { return (client->cid<<16)|mIndex; } return 0xFFFF0000 | mIndex; } sp<LayerBaseClient::Surface> LayerBaseClient::getSurface() const { return new Surface(clientIndex(), mIdentity); } // --------------------------------------------------------------------------- }; // namespace android