[akshaydeshpande]

Hi all,

I am trying to impose a boundary layer profile in a Mach 2.9 flow at the inlet of a flat plate in SU2 v6.0. For this purpose, I have created a function in solver_structure.cpp,
void CSolver::Read_Inlet_Profile(CGeometry *geometry, CConfig *config, unsigned short val_marker)

The function reads in a file containing the velocity and density values to data structures initialized in solver_structure.hpp. I allocate memory to the data structure in the same function as follows,
InletVelocity = new su2double[geometry->nVertex[val_marker]]
InletDensity = new su2double[geometry->nVertex[val_marker]]

I loop over the vertices of that boundary marker and read the data from a text file using getline and istringstream. I call Read_Inlet_Profile in solver_direct_mean.cpp in the function,
void CEulerSolver::BC_Supersonic_Inlet(CGeometry *geometry, CSolver **solver_container, CNumerics *conv_numerics, CNumerics *visc_numerics, CConfig *config, unsigned short val_marker).

To access the data structures, I have created accessing functions GetInletVelocity and GetInletDensity, which I call in the loop,
for (iVertex = 0; iVertex < geometry->nVertex[val_marker]; iVertex++), and appropriately modify the primitive variables in the array V_inlet.

This approach works when I run the code serially, but breaks down when I try to run on multiple processors. To try a parallel approach, I tried to follow the function,
void CSolver::Read_SU2_Restart_ASCII(CGeometry *geometry, CConfig *config, string val_filename) in solver_structure.cpp and have all the ranks read in the file simultaneously.

I believe my problem might be in assigning the velocity/density values to the right vertex after ParMETIS does grid partitioning. I have also tried SU2_MPI::Bcast by having only the root process read the file, but this approach fails too as the number of vertices each in processor are not equal.

I would appreciate any suggestions regarding this issue.

Thanks,

Akshay

[fpalacios]

Hi Akshay, sorry for the late reply,
Maybe you should take a look at
COutput::SetCp_InverseDesign(CSolver *solver_container, CGeometry *geometry, CConfig *config, unsigned long iExtIter) {}
In that subroutine SU2 reads and storages a target Cp distribution on a surface... it maybe be a comparable problem.

By the way, glad you are developing SU2... www.su2devsociety.org maybe be interesting for you.

Best,
Francisco

[akshaydeshpande]

Thanks! I'll take a look at it.