1) file size
2) file handling in paraview: loading speed in paraview
record macro (record timings --> how fast does it read it, does it buffer in the animations)
3) writing speed (how long does it take to write the files)
4) [hdf5 vs. vtu]
5) scaling of vtu --> time w.r.t. to processes
(Kratos:
barrier before postprocessing starts
barrier after postprocessing ends)
average & maximum time of processes/rank
Task:
[x] include timing output per process (rank) in vtu_output.cpp (Kratos script)
[x] prepare simulation to be run on SuperMUC with more cores
[ ] write to a single file (append) -- providing necessary offset
Additional info from meeting:
Kratos --> build_timer e.g. buildTimer.ElapsedSeconds(), timer.h, check out vtk_ouptput.cpp ~line 116, timer.cpp,
(C++ STL std::chrono), VtkOutput::PrintOutput(..) timer_start, time_end = time - time_start;
timing_file.append(time_end);
21/12/2020
PRESENTATION comparisons: vtu & vtk
1) file size 2) file handling in paraview: loading speed in paraview record macro (record timings --> how fast does it read it, does it buffer in the animations)
3) writing speed (how long does it take to write the files) 4) [hdf5 vs. vtu] 5) scaling of vtu --> time w.r.t. to processes (Kratos: barrier before postprocessing starts barrier after postprocessing ends) average & maximum time of processes/rank
Task:
Additional info from meeting:
Kratos --> build_timer e.g. buildTimer.ElapsedSeconds(), timer.h, check out vtk_ouptput.cpp ~line 116, timer.cpp, (C++ STL std::chrono), VtkOutput::PrintOutput(..) timer_start, time_end = time - time_start; timing_file.append(time_end);
ghost level --> per file, not per node