Fragment-Parallel Composite in addition to Filter Parallelism in Interactive Graphics A Feed-Forward Rendering Pipeline Composite & Filter Basic Idea

Fragment-Parallel Composite in addition to Filter Parallelism in Interactive Graphics A Feed-Forward Rendering Pipeline Composite & Filter Basic Idea www.phwiki.com

Fragment-Parallel Composite in addition to Filter Parallelism in Interactive Graphics A Feed-Forward Rendering Pipeline Composite & Filter Basic Idea

Thomas, Ryan, Contributing Writer has reference to this Academic Journal, PHwiki organized this Journal Fragment-Parallel Composite in addition to FilterAnjul Patney, Stanley Tzeng, in addition to John D. OwensUniversity of Cali as long as nia, DavisParallelism in Interactive GraphicsWell-expressed in hardware as well as APIsConsistently growing in degree & expressionMore in addition to more cores on upcoming GPUsFrom programmable shaders to pipelinesWe should rethink algorithms to exploit thisThis paper provides one exampleParallelization of composite/filter stagesA Feed-Forward Rendering PipelineGeometry ProcessingRasterizationCompositeFilterPrimitivesPixels

Charleston Southern University SC www.phwiki.com

This Particular University is Related to this Particular Journal

Composite & FilterInput: Unordered list of fragmentsOutputPixel colorsAssumptionNo fragments are discardedPixelBasic IdeaBasic IdeaPixel-ParallelProcessors

Basic IdeaFragment-ParallelProcessorsMotivationMost applications have low depth complexityPixel-level parallelism is sufficientWe are interested in applications withVery high depth complexityHigh variation in depth complexityFurtherFuture plat as long as ms will dem in addition to more parallelismHigh depth-complexity can limit pixel-parallelismMotivation

Related WorkOrder-Independent Transparency (OIT)Depth-Peeling [Everitt 01]One pass per transparent layerStencil-Routed A-buffer [Myers & Bavoil 07]One pass per 8 depth layers1Bucket Depth-Peeling [Liu et al. 09]One pass per up to 32 layers21 Maximum MSAA samples per pixel2 Maximum render targetsRelated WorkOrder-Independent Transparency (OIT)OIT using Direct3D 11 [Gruen et al. 10]Use fragment linked-listsPer-pixel sort in addition to compositeHair Self-Shadowing [Sintorn et al. 09]Each fragment computes its contributionAssumes constant opacityRelated WorkProgrammable Rendering PipelinesRenderAnts [Zhou et al. 09]Sort fragments globallyPer-pixel composite/filterFreePipe [Liu et al. 10]Sort fragments globallyPer-pixel composite/filter

Pixel-Parallel FormulationP: PixelS: SubsamplePixel-Parallel FormulationWorkload sizeDepends on number of fragmentsLimits the size of renderingDegree of parallelismDepends on number of pixels/subpixelsThese two may not always correspondFragment-Parallel FormulationP: PixelS: SubsampleP: PixelS: Subsample

Fragment-Parallel FormulationHow can this behavior be achievedRevisit the composite equationCs = 1C1 + (1-1){2C2+(1-2)( (N+(1-N)CB) }fragment 1 fragment 2 backgroundCs = 1.1.C1 + (1-1).2.C2 + (1-1)(1-2).3.C3 + + (1-1)(1-2) (1-k-1).i.Ck + + (1-1)(1-2) (1-N).CBFragment-Parallel FormulationLk is trivially parallel (local computation)Gk is the result of a scan operation (product)For the list of input fragmentsCompute G[ ] in addition to L[ ], multiplyPer as long as m reduction to add subpixel contributionsCs = G1.L1 + G2.L2 + G3.L3 GN.LN Gk = (1-1).(1-2) (1-k-1) Lk = k.CkFragment-Parallel FormulationFilter, as long as every pixel: This can be expressed as another reductionAfter multiplying with subpixel weights mCan be merged with previous reductionCp = Cs1.1 + Cs2.2 + + CsM.M

Fragment-Parallel Composite & FilterFinal AlgorithmTwo-key sort (Subpixel ID, depth)Segmented Scan (obtain Gk)Premultiply with weights (Lk, m)Segmented ReductionFragment-Parallel FormulationP: PixelS: SubsampleP: PixelS: SubsampleSegmented Scan (product)Segmented Reduction (sum)ImplementationHardware used: NVIDIA GeForce GTX 280We require fast Segmented Scan in addition to ReduceCUDPP library provides thatRestricts implementation to NVIDIA CUDANo direct access to hardware rasterizerWe wrote our own

Example System – PolygonsApplicationsGamesDepth Complexity1 to few tens of layersSuited to pixel-parallelFragment-parallel software rasterizerExample System – ParticlesApplicationsSimulations, gamesDepth ComplexityHundreds of layersHigh depth-varianceParticle-parallel sprite rasterizerExample System – VolumesApplicationsScientific VisualizationDepth ComplexityTens to Hundreds of layersLow depth-varianceMajor-axis-slice rasterizer

Thomas, Ryan Ranch & Coast Contributing Writer www.phwiki.com

Example System – ReyesApplicationsOffline renderingDepth ComplexityTens of layersModerate depth varianceData-parallel micropolygon rasterizerPer as long as mance ResultsPer as long as mance Variation

LimitationsIncreased memory trafficSeveral passes through CUDPP primitivesUnclear how to optimize as long as special casesThreshold opacityThreshold depth complexitySummary in addition to ConclusionParallel as long as mulation of composite equationMaps well to known primitivesCan be integrated with filterConsistent per as long as mance across varying workloadsFPC is applicable to future rendering pipelinesExploits higher degree of parallelismBetter related to size of rendering workloadA tool as long as building programmable pipelinesFuture WorkPer as long as manceReduction in memory trafficExtension to special-case scenesHybrid PPC-FPC as long as mulationsApplicationsIntegration with hardware rasterizerCinematic rendering, Photoshop

AcknowledgmentsNSF Award 0541448SciDAC Insitute as long as Ultrascale VisualizationNVIDIA Research Fellowship Equipment donated by NVIDIADiscussions in addition to FeedbackShubho Sengupta (UC Davis), Matt Pharr (Intel), Aaron Lefohn (Intel), Mike Houston (AMD)Anonymous reviewersImplementation assistanceJeff Stuart, Shubho SenguptaThanks!

Thomas, Ryan Contributing Writer

Thomas, Ryan is from United States and they belong to Ranch & Coast and they are from  Rancho Santa Fe, United States got related to this Particular Journal. and Thomas, Ryan deal with the subjects like Affluent Lifestyle; Regional Interest; Society

Journal Ratings by Charleston Southern University

This Particular Journal got reviewed and rated by Charleston Southern University and short form of this particular Institution is SC and gave this Journal an Excellent Rating.