Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
No matching results
An error occurred while fetching branches. Retry the search.
An error occurred while fetching tags. Retry the search.
Created with Raphaël 2.2.027Aug26252423222120191716151413976532122Jul20191514131198765432128Jun22201918131211109876542131May30171412109876543123Apr19181712529Mar282422198654127Feb26222120543229Jan19171110921Dec191817141110976543230Nov2927262520131211109654130Oct292827Merge branch 'develop' into matrices-gpuTunnig experiments - new versio of CUDA boundary traverser, simpler and using more CUDA blocks.Implemented comparison of boundary conditions kernel.Implementing additional code for performance comparison.Heat equation benchmark prepared for testing of traversers.Optimized CUDA traverser for boundary entities in 2D -> speed-up almost 20%.Fixe of boundary conditions in HeatEquationBenchmarkProblem.Fix in tunning ExplicitUpdater.Fixed handling of the boundary conditions for explicit solver - introduced a method Problem::applyBoundaryConditions.SharedPointer of TraverserUserData replaced just by TraverserUserData -> very small speed-up.Tunning experiments - test of replacement od UserDataPoiner by UserData.Tunning experiments - implemented SimpleCell -> still keeping quite good efficiency.Optimized interior entity processor of ExplicitUpdater -> small speed-up.Restored --offline-mode flag.Tunning experiments - included explicit updater entyties processor.Tunning experiments - introduced explicit updater user data.Tunning experiments - 'pure-c' benchmark copied to 'tunning' and refactored to work with operator and boundary conditions.Fixed cmake for build with Python.Added flag to turn of build with Python.Added flag to turn off build with Python.Tunning experiments - replaced MeshFunction u by Real* u -> no speed-up.Tuning experiments - avoided use of MeshEntity -> no speed-up.Extracted source code for the tunning experiments.Parallel version for CUDA of FSM is implemented.Merge branch 'develop' into operators-optimizationsCUDA Unified memroy is turned off by default - it is very slow currently.Implemented CUDA SharedPointer with CUDA Unified Memory. Does not work with sm30 and it is very slow.SharedPointer.h splitted into SharedPointer[Cuda/Host/Mic].h.All pointers moved to Pointer subdir.Merge branch 'cmake' into developFixed installation of SubdomainOverlapsGetter.hpp.Merge branch 'common-data' into developBoundaryMeshFunction added.Added --offline-build option to skip update of external Cmake projectsDisabled linking to the VTK library and enabled all mesh tests for CUDA even for nvccRemoved check for HAVE_CUSPARSEDeclared all custom build options in CMakeLists.txtRefactoring of MeshFunction.Merge branch 'periodic-bc' of geraldine.fjfi.cvut.cz:/local/projects/tnl/tnl into periodic-bcFix of comment in DistributedGrid.