/* * Copyright 2017 Google Inc. * * Use of this source code is governed by a BSD-style license that can be * found in the LICENSE file. */ #include "GrCoverageCountingPathRenderer.h" #include "GrCaps.h" #include "GrClip.h" #include "GrGpu.h" #include "GrGpuCommandBuffer.h" #include "GrOpFlushState.h" #include "GrProxyProvider.h" #include "GrRenderTargetOpList.h" #include "GrStyle.h" #include "GrTexture.h" #include "SkMakeUnique.h" #include "SkMatrix.h" #include "SkPathOps.h" #include "ccpr/GrCCClipProcessor.h" // Shorthand for keeping line lengths under control with nested classes... using CCPR = GrCoverageCountingPathRenderer; // If a path spans more pixels than this, we need to crop it or else analytic AA can run out of fp32 // precision. static constexpr float kPathCropThreshold = 1 << 16; static void crop_path(const SkPath& path, const SkIRect& cropbox, SkPath* out) { SkPath cropPath; cropPath.addRect(SkRect::Make(cropbox)); if (!Op(cropPath, path, kIntersect_SkPathOp, out)) { // This can fail if the PathOps encounter NaN or infinities. out->reset(); } } bool GrCoverageCountingPathRenderer::IsSupported(const GrCaps& caps) { const GrShaderCaps& shaderCaps = *caps.shaderCaps(); return shaderCaps.integerSupport() && shaderCaps.flatInterpolationSupport() && caps.instanceAttribSupport() && GrCaps::kNone_MapFlags != caps.mapBufferFlags() && caps.isConfigTexturable(kAlpha_half_GrPixelConfig) && caps.isConfigRenderable(kAlpha_half_GrPixelConfig, /*withMSAA=*/false) && !caps.blacklistCoverageCounting(); } sk_sp<GrCoverageCountingPathRenderer> GrCoverageCountingPathRenderer::CreateIfSupported( const GrCaps& caps, bool drawCachablePaths) { auto ccpr = IsSupported(caps) ? new GrCoverageCountingPathRenderer(drawCachablePaths) : nullptr; return sk_sp<GrCoverageCountingPathRenderer>(ccpr); } GrPathRenderer::CanDrawPath GrCoverageCountingPathRenderer::onCanDrawPath( const CanDrawPathArgs& args) const { if (args.fShape->hasUnstyledKey() && !fDrawCachablePaths) { return CanDrawPath::kNo; } if (!args.fShape->style().isSimpleFill() || args.fShape->inverseFilled() || args.fViewMatrix->hasPerspective() || GrAAType::kCoverage != args.fAAType) { return CanDrawPath::kNo; } SkPath path; args.fShape->asPath(&path); if (SkPathPriv::ConicWeightCnt(path)) { return CanDrawPath::kNo; } SkRect devBounds; SkIRect devIBounds; args.fViewMatrix->mapRect(&devBounds, path.getBounds()); devBounds.roundOut(&devIBounds); if (!devIBounds.intersect(*args.fClipConservativeBounds)) { // Path is completely clipped away. Our code will eventually notice this before doing any // real work. return CanDrawPath::kYes; } if (devIBounds.height() * devIBounds.width() > 256 * 256) { // Large paths can blow up the atlas fast. And they are not ideal for a two-pass rendering // algorithm. Give the simpler direct renderers a chance before we commit to drawing it. return CanDrawPath::kAsBackup; } if (args.fShape->hasUnstyledKey() && path.countVerbs() > 50) { // Complex paths do better cached in an SDF, if the renderer will accept them. return CanDrawPath::kAsBackup; } return CanDrawPath::kYes; } bool GrCoverageCountingPathRenderer::onDrawPath(const DrawPathArgs& args) { SkASSERT(!fFlushing); auto op = skstd::make_unique<DrawPathsOp>(this, args, args.fPaint.getColor()); args.fRenderTargetContext->addDrawOp(*args.fClip, std::move(op)); return true; } CCPR::DrawPathsOp::DrawPathsOp(GrCoverageCountingPathRenderer* ccpr, const DrawPathArgs& args, GrColor color) : INHERITED(ClassID()) , fCCPR(ccpr) , fSRGBFlags(GrPipeline::SRGBFlagsFromPaint(args.fPaint)) , fProcessors(std::move(args.fPaint)) , fTailDraw(&fHeadDraw) , fOwningRTPendingPaths(nullptr) { SkDEBUGCODE(++fCCPR->fPendingDrawOpsCount); SkDEBUGCODE(fBaseInstance = -1); SkDEBUGCODE(fInstanceCount = 1); SkDEBUGCODE(fNumSkippedInstances = 0); GrRenderTargetContext* const rtc = args.fRenderTargetContext; SkRect devBounds; args.fViewMatrix->mapRect(&devBounds, args.fShape->bounds()); args.fClip->getConservativeBounds(rtc->width(), rtc->height(), &fHeadDraw.fClipIBounds, nullptr); if (SkTMax(devBounds.height(), devBounds.width()) > kPathCropThreshold) { // The path is too large. We need to crop it or analytic AA can run out of fp32 precision. SkPath path; args.fShape->asPath(&path); path.transform(*args.fViewMatrix); fHeadDraw.fMatrix.setIdentity(); crop_path(path, fHeadDraw.fClipIBounds, &fHeadDraw.fPath); devBounds = fHeadDraw.fPath.getBounds(); } else { fHeadDraw.fMatrix = *args.fViewMatrix; args.fShape->asPath(&fHeadDraw.fPath); } fHeadDraw.fColor = color; // Can't call args.fPaint.getColor() because it has been std::move'd. // FIXME: intersect with clip bounds to (hopefully) improve batching. // (This is nontrivial due to assumptions in generating the octagon cover geometry.) this->setBounds(devBounds, GrOp::HasAABloat::kYes, GrOp::IsZeroArea::kNo); } CCPR::DrawPathsOp::~DrawPathsOp() { if (fOwningRTPendingPaths) { // Remove CCPR's dangling pointer to this Op before deleting it. fOwningRTPendingPaths->fDrawOps.remove(this); } SkDEBUGCODE(--fCCPR->fPendingDrawOpsCount); } GrDrawOp::RequiresDstTexture CCPR::DrawPathsOp::finalize(const GrCaps& caps, const GrAppliedClip* clip, GrPixelConfigIsClamped dstIsClamped) { SkASSERT(!fCCPR->fFlushing); // There should only be one single path draw in this Op right now. SkASSERT(1 == fInstanceCount); SkASSERT(&fHeadDraw == fTailDraw); GrProcessorSet::Analysis analysis = fProcessors.finalize(fHeadDraw.fColor, GrProcessorAnalysisCoverage::kSingleChannel, clip, false, caps, dstIsClamped, &fHeadDraw.fColor); return analysis.requiresDstTexture() ? RequiresDstTexture::kYes : RequiresDstTexture::kNo; } bool CCPR::DrawPathsOp::onCombineIfPossible(GrOp* op, const GrCaps& caps) { DrawPathsOp* that = op->cast<DrawPathsOp>(); SkASSERT(fCCPR == that->fCCPR); SkASSERT(!fCCPR->fFlushing); SkASSERT(fOwningRTPendingPaths); SkASSERT(fInstanceCount); SkASSERT(!that->fOwningRTPendingPaths || that->fOwningRTPendingPaths == fOwningRTPendingPaths); SkASSERT(that->fInstanceCount); if (this->getFillType() != that->getFillType() || fSRGBFlags != that->fSRGBFlags || fProcessors != that->fProcessors) { return false; } fTailDraw->fNext = &fOwningRTPendingPaths->fDrawsAllocator.push_back(that->fHeadDraw); fTailDraw = (that->fTailDraw == &that->fHeadDraw) ? fTailDraw->fNext : that->fTailDraw; this->joinBounds(*that); SkDEBUGCODE(fInstanceCount += that->fInstanceCount); SkDEBUGCODE(that->fInstanceCount = 0); return true; } void CCPR::DrawPathsOp::wasRecorded(GrRenderTargetOpList* opList) { SkASSERT(!fCCPR->fFlushing); SkASSERT(!fOwningRTPendingPaths); fOwningRTPendingPaths = &fCCPR->fRTPendingPathsMap[opList->uniqueID()]; fOwningRTPendingPaths->fDrawOps.addToTail(this); } bool GrCoverageCountingPathRenderer::canMakeClipProcessor(const SkPath& deviceSpacePath) const { if (!fDrawCachablePaths && !deviceSpacePath.isVolatile()) { return false; } if (SkPathPriv::ConicWeightCnt(deviceSpacePath)) { return false; } return true; } std::unique_ptr<GrFragmentProcessor> GrCoverageCountingPathRenderer::makeClipProcessor( GrProxyProvider* proxyProvider, uint32_t opListID, const SkPath& deviceSpacePath, const SkIRect& accessRect, int rtWidth, int rtHeight) { using MustCheckBounds = GrCCClipProcessor::MustCheckBounds; SkASSERT(!fFlushing); SkASSERT(this->canMakeClipProcessor(deviceSpacePath)); ClipPath& clipPath = fRTPendingPathsMap[opListID].fClipPaths[deviceSpacePath.getGenerationID()]; if (clipPath.isUninitialized()) { // This ClipPath was just created during lookup. Initialize it. clipPath.init(proxyProvider, deviceSpacePath, accessRect, rtWidth, rtHeight); } else { clipPath.addAccess(accessRect); } bool mustCheckBounds = !clipPath.pathDevIBounds().contains(accessRect); return skstd::make_unique<GrCCClipProcessor>(&clipPath, MustCheckBounds(mustCheckBounds), deviceSpacePath.getFillType()); } void CCPR::ClipPath::init(GrProxyProvider* proxyProvider, const SkPath& deviceSpacePath, const SkIRect& accessRect, int rtWidth, int rtHeight) { SkASSERT(this->isUninitialized()); fAtlasLazyProxy = proxyProvider->createFullyLazyProxy( [this](GrResourceProvider* resourceProvider, GrSurfaceOrigin* outOrigin) { if (!resourceProvider) { return sk_sp<GrTexture>(); } SkASSERT(fHasAtlas); SkASSERT(!fHasAtlasTransform); GrTextureProxy* textureProxy = fAtlas ? fAtlas->textureProxy() : nullptr; if (!textureProxy || !textureProxy->instantiate(resourceProvider)) { fAtlasScale = fAtlasTranslate = {0, 0}; SkDEBUGCODE(fHasAtlasTransform = true); return sk_sp<GrTexture>(); } fAtlasScale = {1.f / textureProxy->width(), 1.f / textureProxy->height()}; fAtlasTranslate = {fAtlasOffsetX * fAtlasScale.x(), fAtlasOffsetY * fAtlasScale.y()}; if (kBottomLeft_GrSurfaceOrigin == textureProxy->origin()) { fAtlasScale.fY = -fAtlasScale.y(); fAtlasTranslate.fY = 1 - fAtlasTranslate.y(); } SkDEBUGCODE(fHasAtlasTransform = true); *outOrigin = textureProxy->origin(); return sk_ref_sp(textureProxy->priv().peekTexture()); }, GrProxyProvider::Renderable::kYes, kAlpha_half_GrPixelConfig); const SkRect& pathDevBounds = deviceSpacePath.getBounds(); if (SkTMax(pathDevBounds.height(), pathDevBounds.width()) > kPathCropThreshold) { // The path is too large. We need to crop it or analytic AA can run out of fp32 precision. crop_path(deviceSpacePath, SkIRect::MakeWH(rtWidth, rtHeight), &fDeviceSpacePath); } else { fDeviceSpacePath = deviceSpacePath; } deviceSpacePath.getBounds().roundOut(&fPathDevIBounds); fAccessRect = accessRect; } void GrCoverageCountingPathRenderer::preFlush(GrOnFlushResourceProvider* onFlushRP, const uint32_t* opListIDs, int numOpListIDs, SkTArray<sk_sp<GrRenderTargetContext>>* results) { using PathInstance = GrCCPathProcessor::Instance; SkASSERT(!fFlushing); SkASSERT(!fPerFlushIndexBuffer); SkASSERT(!fPerFlushVertexBuffer); SkASSERT(!fPerFlushInstanceBuffer); SkASSERT(!fPerFlushPathParser); SkASSERT(fPerFlushAtlases.empty()); SkDEBUGCODE(fFlushing = true); if (fRTPendingPathsMap.empty()) { return; // Nothing to draw. } fPerFlushResourcesAreValid = false; // Count the paths that are being flushed. int maxTotalPaths = 0, maxPathPoints = 0, numSkPoints = 0, numSkVerbs = 0; SkDEBUGCODE(int numClipPaths = 0); for (int i = 0; i < numOpListIDs; ++i) { auto it = fRTPendingPathsMap.find(opListIDs[i]); if (fRTPendingPathsMap.end() == it) { continue; } const RTPendingPaths& rtPendingPaths = it->second; SkTInternalLList<DrawPathsOp>::Iter drawOpsIter; drawOpsIter.init(rtPendingPaths.fDrawOps, SkTInternalLList<DrawPathsOp>::Iter::kHead_IterStart); while (DrawPathsOp* op = drawOpsIter.get()) { for (const DrawPathsOp::SingleDraw* draw = op->head(); draw; draw = draw->fNext) { ++maxTotalPaths; maxPathPoints = SkTMax(draw->fPath.countPoints(), maxPathPoints); numSkPoints += draw->fPath.countPoints(); numSkVerbs += draw->fPath.countVerbs(); } drawOpsIter.next(); } maxTotalPaths += rtPendingPaths.fClipPaths.size(); SkDEBUGCODE(numClipPaths += rtPendingPaths.fClipPaths.size()); for (const auto& clipsIter : rtPendingPaths.fClipPaths) { const SkPath& path = clipsIter.second.deviceSpacePath(); maxPathPoints = SkTMax(path.countPoints(), maxPathPoints); numSkPoints += path.countPoints(); numSkVerbs += path.countVerbs(); } } if (!maxTotalPaths) { return; // Nothing to draw. } // Allocate GPU buffers. fPerFlushIndexBuffer = GrCCPathProcessor::FindIndexBuffer(onFlushRP); if (!fPerFlushIndexBuffer) { SkDebugf("WARNING: failed to allocate ccpr path index buffer.\n"); return; } fPerFlushVertexBuffer = GrCCPathProcessor::FindVertexBuffer(onFlushRP); if (!fPerFlushVertexBuffer) { SkDebugf("WARNING: failed to allocate ccpr path vertex buffer.\n"); return; } fPerFlushInstanceBuffer = onFlushRP->makeBuffer(kVertex_GrBufferType, maxTotalPaths * sizeof(PathInstance)); if (!fPerFlushInstanceBuffer) { SkDebugf("WARNING: failed to allocate path instance buffer. No paths will be drawn.\n"); return; } PathInstance* pathInstanceData = static_cast<PathInstance*>(fPerFlushInstanceBuffer->map()); SkASSERT(pathInstanceData); int pathInstanceIdx = 0; fPerFlushPathParser = sk_make_sp<GrCCPathParser>(maxTotalPaths, maxPathPoints, numSkPoints, numSkVerbs); SkDEBUGCODE(int skippedTotalPaths = 0); // Allocate atlas(es) and fill out GPU instance buffers. for (int i = 0; i < numOpListIDs; ++i) { auto it = fRTPendingPathsMap.find(opListIDs[i]); if (fRTPendingPathsMap.end() == it) { continue; } RTPendingPaths& rtPendingPaths = it->second; SkTInternalLList<DrawPathsOp>::Iter drawOpsIter; drawOpsIter.init(rtPendingPaths.fDrawOps, SkTInternalLList<DrawPathsOp>::Iter::kHead_IterStart); while (DrawPathsOp* op = drawOpsIter.get()) { pathInstanceIdx = op->setupResources(onFlushRP, pathInstanceData, pathInstanceIdx); drawOpsIter.next(); SkDEBUGCODE(skippedTotalPaths += op->numSkippedInstances_debugOnly()); } for (auto& clipsIter : rtPendingPaths.fClipPaths) { clipsIter.second.placePathInAtlas(this, onFlushRP, fPerFlushPathParser.get()); } } fPerFlushInstanceBuffer->unmap(); SkASSERT(pathInstanceIdx == maxTotalPaths - skippedTotalPaths - numClipPaths); if (!fPerFlushAtlases.empty()) { auto coverageCountBatchID = fPerFlushPathParser->closeCurrentBatch(); fPerFlushAtlases.back().setCoverageCountBatchID(coverageCountBatchID); } if (!fPerFlushPathParser->finalize(onFlushRP)) { SkDebugf("WARNING: failed to allocate GPU buffers for CCPR. No paths will be drawn.\n"); return; } // Draw the atlas(es). GrTAllocator<GrCCAtlas>::Iter atlasIter(&fPerFlushAtlases); while (atlasIter.next()) { if (auto rtc = atlasIter.get()->finalize(onFlushRP, fPerFlushPathParser)) { results->push_back(std::move(rtc)); } } fPerFlushResourcesAreValid = true; } int CCPR::DrawPathsOp::setupResources(GrOnFlushResourceProvider* onFlushRP, GrCCPathProcessor::Instance* pathInstanceData, int pathInstanceIdx) { GrCCPathParser* parser = fCCPR->fPerFlushPathParser.get(); const GrCCAtlas* currentAtlas = nullptr; SkASSERT(fInstanceCount > 0); SkASSERT(-1 == fBaseInstance); fBaseInstance = pathInstanceIdx; for (const SingleDraw* draw = this->head(); draw; draw = draw->fNext) { // parsePath gives us two tight bounding boxes: one in device space, as well as a second // one rotated an additional 45 degrees. The path vertex shader uses these two bounding // boxes to generate an octagon that circumscribes the path. SkRect devBounds, devBounds45; parser->parsePath(draw->fMatrix, draw->fPath, &devBounds, &devBounds45); SkIRect devIBounds; devBounds.roundOut(&devIBounds); int16_t offsetX, offsetY; GrCCAtlas* atlas = fCCPR->placeParsedPathInAtlas(onFlushRP, draw->fClipIBounds, devIBounds, &offsetX, &offsetY); if (!atlas) { SkDEBUGCODE(++fNumSkippedInstances); continue; } if (currentAtlas != atlas) { if (currentAtlas) { this->addAtlasBatch(currentAtlas, pathInstanceIdx); } currentAtlas = atlas; } const SkMatrix& m = draw->fMatrix; pathInstanceData[pathInstanceIdx++] = { devBounds, devBounds45, {{m.getScaleX(), m.getSkewY(), m.getSkewX(), m.getScaleY()}}, {{m.getTranslateX(), m.getTranslateY()}}, {{offsetX, offsetY}}, draw->fColor}; } SkASSERT(pathInstanceIdx == fBaseInstance + fInstanceCount - fNumSkippedInstances); if (currentAtlas) { this->addAtlasBatch(currentAtlas, pathInstanceIdx); } return pathInstanceIdx; } void CCPR::ClipPath::placePathInAtlas(GrCoverageCountingPathRenderer* ccpr, GrOnFlushResourceProvider* onFlushRP, GrCCPathParser* parser) { SkASSERT(!this->isUninitialized()); SkASSERT(!fHasAtlas); parser->parseDeviceSpacePath(fDeviceSpacePath); fAtlas = ccpr->placeParsedPathInAtlas(onFlushRP, fAccessRect, fPathDevIBounds, &fAtlasOffsetX, &fAtlasOffsetY); SkDEBUGCODE(fHasAtlas = true); } GrCCAtlas* GrCoverageCountingPathRenderer::placeParsedPathInAtlas( GrOnFlushResourceProvider* onFlushRP, const SkIRect& clipIBounds, const SkIRect& pathIBounds, int16_t* atlasOffsetX, int16_t* atlasOffsetY) { using ScissorMode = GrCCPathParser::ScissorMode; ScissorMode scissorMode; SkIRect clippedPathIBounds; if (clipIBounds.contains(pathIBounds)) { clippedPathIBounds = pathIBounds; scissorMode = ScissorMode::kNonScissored; } else if (clippedPathIBounds.intersect(clipIBounds, pathIBounds)) { scissorMode = ScissorMode::kScissored; } else { fPerFlushPathParser->discardParsedPath(); return nullptr; } SkIPoint16 atlasLocation; int h = clippedPathIBounds.height(), w = clippedPathIBounds.width(); if (fPerFlushAtlases.empty() || !fPerFlushAtlases.back().addRect(w, h, &atlasLocation)) { if (!fPerFlushAtlases.empty()) { // The atlas is out of room and can't grow any bigger. auto coverageCountBatchID = fPerFlushPathParser->closeCurrentBatch(); fPerFlushAtlases.back().setCoverageCountBatchID(coverageCountBatchID); } fPerFlushAtlases.emplace_back(*onFlushRP->caps(), w, h).addRect(w, h, &atlasLocation); } *atlasOffsetX = atlasLocation.x() - static_cast<int16_t>(clippedPathIBounds.left()); *atlasOffsetY = atlasLocation.y() - static_cast<int16_t>(clippedPathIBounds.top()); fPerFlushPathParser->saveParsedPath(scissorMode, clippedPathIBounds, *atlasOffsetX, *atlasOffsetY); return &fPerFlushAtlases.back(); } void CCPR::DrawPathsOp::onExecute(GrOpFlushState* flushState) { SkASSERT(fCCPR->fFlushing); SkASSERT(flushState->rtCommandBuffer()); if (!fCCPR->fPerFlushResourcesAreValid) { return; // Setup failed. } SkASSERT(fBaseInstance >= 0); // Make sure setupResources has been called. GrPipeline::InitArgs initArgs; initArgs.fFlags = fSRGBFlags; initArgs.fProxy = flushState->drawOpArgs().fProxy; initArgs.fCaps = &flushState->caps(); initArgs.fResourceProvider = flushState->resourceProvider(); initArgs.fDstProxy = flushState->drawOpArgs().fDstProxy; GrPipeline pipeline(initArgs, std::move(fProcessors), flushState->detachAppliedClip()); int baseInstance = fBaseInstance; for (int i = 0; i < fAtlasBatches.count(); baseInstance = fAtlasBatches[i++].fEndInstanceIdx) { const AtlasBatch& batch = fAtlasBatches[i]; SkASSERT(batch.fEndInstanceIdx > baseInstance); if (!batch.fAtlas->textureProxy()) { continue; // Atlas failed to allocate. } GrCCPathProcessor pathProc(flushState->resourceProvider(), sk_ref_sp(batch.fAtlas->textureProxy()), this->getFillType()); GrMesh mesh(GrCCPathProcessor::MeshPrimitiveType(flushState->caps())); mesh.setIndexedInstanced(fCCPR->fPerFlushIndexBuffer.get(), GrCCPathProcessor::NumIndicesPerInstance(flushState->caps()), fCCPR->fPerFlushInstanceBuffer.get(), batch.fEndInstanceIdx - baseInstance, baseInstance); mesh.setVertexData(fCCPR->fPerFlushVertexBuffer.get()); flushState->rtCommandBuffer()->draw(pipeline, pathProc, &mesh, nullptr, 1, this->bounds()); } SkASSERT(baseInstance == fBaseInstance + fInstanceCount - fNumSkippedInstances); } void GrCoverageCountingPathRenderer::postFlush(GrDeferredUploadToken, const uint32_t* opListIDs, int numOpListIDs) { SkASSERT(fFlushing); fPerFlushAtlases.reset(); fPerFlushPathParser.reset(); fPerFlushInstanceBuffer.reset(); fPerFlushVertexBuffer.reset(); fPerFlushIndexBuffer.reset(); // We wait to erase these until after flush, once Ops and FPs are done accessing their data. for (int i = 0; i < numOpListIDs; ++i) { fRTPendingPathsMap.erase(opListIDs[i]); } SkDEBUGCODE(fFlushing = false); }