hi evevyone, recently,I Build paraview source with CMake and vs2015 with QT5.10,It build seccused,but when I use paraview.exe which created by vs2015 ,the paraview load a VTK File wait for few minutes,it is too slower than the published paraview.exe.Is there someone meet with this problem？ some one can tell me how to speed up this.thx!!!
Did you build in Debug mode? With MSVC, Debug builds are substantially slower than Release builds (which is the mode the published paraview.exe was built with). This is particularly true when it comes to parsing ASCII text files, which I am assuming your VTK file is. You can try to work around this by saving the VTK file as a binary file, which will be much faster to load in both Release and Debug modes. Or you can just rebuild in Release mode.
yes，I know Debug mode is slower than release builds,and i also had tried the release build mode,it is faster a little,but it also load vtk File slowly;yes, the vtk file is ASCII one,but when i use the published paraview.exe to load the same vtk file,it more fast. I dont konw when the paraview load the file,I clicked the Open File,selected my file path,in the pipeline brower can see the name of vtk,when i clicked the icon(which look like an eye),the sofware will be hold on and i must wait for about 2min,it can finish loading the model in the graphic view.
That’s interesting. I’m not sure what else might explain the slowdown.
thanks for your reply.I not sure… Is the MPI_module has relationship about files loading？Maybe some module is not available？
Not sure if this is the issue, but don’t forget about caching issues. The first time the dataset is loaded, it comes from disk. The second time, it probably comes from cache (depending on the size of the dataset). So, if you try your debug build first, and a Kitware release build second, the release build will be much faster. Always use the second load of your data for performance tests.
Yep it the second time load the vtk from cache,it is
Also performance much slowly(about 2min)than I used the published paraview.exe(just 15seconds).
签名由 网易邮箱大师 定制
MPI should not be a factor. The Kitware binaries are built with Intel’s Thread Building Blocks as the multithreading backend, which is used to accelerate some algorithms, but not the reader as far as I know. Could you compare the CPU usages between the two versions? If you have a multicore system and the Kitware ParaView uses all your cores at near 100% while your built version does not use all the cores, that could be the difference.