Skip to content
Snippets Groups Projects
Select Git revision
  • 121-buffersystem-receiver-info-without-sender-ranks
  • 4-stable
  • CodegenForRefinement
  • RDM/UniformGrid
  • RDM/find_optimal_resource_distribution
  • ReinitializerTest
  • UpdatePhaseField
  • clang11
  • extendUniformGridGPU
  • farhata/benchmarkChallenge
  • fluidizedbed_showcase
  • free-surface
  • free-surface-new-curvature
  • master default protected
  • michis-moving-drop
  • removePhysics
  • setup_walberla_codegen
  • v4.2
  • v5.0dev
  • v4.1
  • v3.3
  • v3.2
  • v4.0dev
  • v3.1
24 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.014Mar131265428Feb2726131087653227Jan17141317Dec10928Nov222120191413116525Oct242318171413109827Sep2623191817929Aug2019Jun134314May8230Apr2923105226Mar20187628Feb27136231Jan302926252423221817921Dec19181513116430Nov282724232217109823Oct201719Sep1524Aug23228Jul26131130Jun2722161497126May252217121110943220Apr529Mar211714109828Feb2215873126Jan25242328Nov161493226Oct2420191714107528Sep16Aug29Jul2825221513118517Jun8131May244328Apr21201230Mar252423221716151411108743128Feb2521181710874227Jan20Dec177629Nov25242322191817161098643229Oct282522215130Sep2927232120161364331Aug2826242320191716131029Jul1376532130Jun292825242322Add a dry run method to check if the chosen decompositions are appropriate; and enhance 3D decomposition functions with descending order option (to be used for cell_per_block decomposition).RDM/find_optima…RDM/find_optimal_resource_distributionEnhance 3D decomposition functions with flexible conditions and improve candidate generation for optimal distributions.Refactor distribution logic to improve candidate generation.Refactor decomposition functions for improved balance and clarity; enhance candidate generation for 3D process distribution.[draft] change scripts to find a optimal distribution for Marenostrum.Divide search for decomposition to be able to obtain a list of possible decompositions.[draft] Add a function that looks for the ideal initial problem distribution for strong scaling.RDM/UniformGridRDM/UniformGridImprove documentation of python setup functions.Add another block decomposition method that does not use a greedy algorithm but tries to decompose the processes as equally as possible into 3 spatial directions.[skip ci] Adjust validation run configs to have meaningful test outputs.[draft] Choose Timers that are actually helpful.[skip ci] [debug] Test various timer accumulations to better unerstand how the waLBerla intern timer works.Fix completely wrong timings of GPU benchmark and start to refactor the setup script.Restructure GPU benchmark timing.fix overall timer for time steps per second computation. The counter does not need to be summed up over the MPI processes.Incrase the verbosity of the SQL file, to not only include the timing_pools, but also the timer that is used to compute the time steps per second metric.Reduce increase of outer iteration with increasing node number by a factor of 0.85, as doubling the number of nodes will likely not half the runtime, but be affected by communication. This `dissipation` like behaviour is asumed to be represented by this 0.85 factor.Fix some syntax error for the strong scaling part of the benchmark_config_RDM.py script for UniformGrid GPU on MareNostrum5-ACC.Fix some syntax error for the strong scaling part of the benchmark_config_RDM.py script for UniformGrid CPU on MareNostrum5-GPP.Add a todo note to domain_distribution setup, to improve the distribution at some point.Add partition_domain to python init to introduce it.Adjust cuda and domain_size parameters to not exceed the resources anymore for the GPU partition on MareNostrum5-ACC.Add a check for splitting and improve the sizing problem for strong skaling for the GPU partition on MareNostrum5-ACC.Add a check for splitting and improve the sizing problem for strong skaling for the CPU partition on MareNostrum5-GPP.Add function to compute an optimal distribution of cells_per_block between an given block decomposition at a given domain_size.Bugfix forgot ; while Multitasking.Fix MemoryStatistics output.Add memory usage logging to UniformGrid benchmarks.Fix small error that I missed the first time.Update benchmark configuration for UniformGrid CPU for RDM specifically on MareNostrum5-GPP.Update benchmark configuration for UniformGrid GPU for RDM specifically on MareNostrum5-ACC.MareNostrumACC setup [singleNode].Merge branch walberla:master into mastermastermasterMerge branch 'suffa/python_callback_fix' into 'master'Merge remote-tracking branch 'upstream/master'Refactor benchmark scenario creation to support multiple inner_outer_split configurations for weak and strong scaling benchmarks.Remove gui[no main fork] stop tracking the files that are outsourced to git walberla-config repository.extendUniformGr…extendUniformGridGPUMerge branch 'kemmler/remove_gui' into 'master'Remove gui