Enhancing the Performance of Web-based Extended Reality with WebAssembly and WebGPU

Abstract

Abstract The increasingdemandforimmersiveweb-basedexperienceshasdriventhede- velopmentofadvancedbrowsertechnologiescapableofdeliveringrich,interactive content.Amongthese,WebAssemblyandWebGPUhaveemergedaspowerful toolsthatpromisenear-nativeperformanceandhigh-fidelityrenderingcapabil- ities, respectively.Thisresearchexploresthepracticalintegrationofthesetwo technologieswithinthecontextofweb-basedExtendedReality(XR)applications, focusingonperformanceandimplementationchallenges. The primaryobjectiveofthisstudyistoinvestigatehowWebAssemblyandWe- bGPU canbejointlyleveragedtoimplementperformantXRapplicationswithin the browser,andtodeterminewhethertheyoffertangibleadvantagesovertra- ditional WebGLandJavaScript-basedapproaches.Toexplorethis,aprototype system supportingbothVirtualReality(VR)andAugmentedReality(AR)was developed.WebAssemblywasutilizedforhandlingCPU-relatedtasks,suchas processingscenedata,managingXRsessiondata,andpreparingGPUcommands. Meanwhile,WebGPUwasemployedforrenderingcomplex3Dscenesandexecut- ing large-scalematrixcomputations.DuetothelackofofficialWebGPUsupport in theWebXRDeviceAPI,theprototypeadoptedWebGLasabridginglayerfor XR renderinglayerhandling,resultinginahybridrenderingpipelinethatwas critically evaluatedthroughoutthestudy. The studywasdrivenbyaseriesofresearchquestionsaimedatidentifyingsuit- able implementationapproaches,demonstratingapracticalintegrationstrategy, and assessingperformanceoutcomes.Acomprehensiveperformanceevaluation wasconductedacrossmultipledimensions,includingCPUusage,GPUworkload, frame ratestability,anddrawcallbehaviorundervaryingscenecomplexities. The resultsshowedthattheintegrationofWebAssemblyandWebGPUledto notable improvementsinCPUefficiency—especiallyinscenariosinvolvinghigh- volumematrixcomputations.However,itdidnotconsistentlyoutperformthe traditional WebGL-basedimplementationintermsofoverallperformance.This limitation mayhavecausedbytheoverheadintroducedbytheinteroperability