I need to export complex vectors data from PETSc and read them in MATLAB.
However, I am encountering some issues with the data format and
interpretation in MATLAB.
code-snippet of the vector data export section:
// Assemble the vectors before exporting
ierr = VecAssemblyBegin(f); CHKERRQ(ierr);
ierr = VecAssemblyEnd(f); CHKERRQ(ierr);
PetscViewer viewerf;
// Save the complex vectors to binary files
ierr = PetscViewerBinaryOpen(PETSC_COMM_WORLD, "output_f.dat",
FILE_MODE_WRITE, &viewerf); CHKERRQ(ierr);
ierr = VecView(f, viewerf); CHKERRQ(ierr);
ierr = PetscViewerDestroy(&viewerf); CHKERRQ(ierr);
// Create vectors to store the magnitudes
Vec f_magnitude;
ierr = VecDuplicate(f, &f_magnitude); CHKERRQ(ierr);
// Get local portion of the vectors
const PetscScalar *f_array;
PetscScalar *f_magnitude_array;
PetscInt n_local;
ierr = VecGetLocalSize(f, &n_local); CHKERRQ(ierr);
ierr = VecGetArrayRead(f, &f_array); CHKERRQ(ierr);
ierr = VecGetArray(f_magnitude, &f_magnitude_array); CHKERRQ(ierr);
// Compute the magnitude for each element
for (int i = 0; i < n_local; i++) {
f_magnitude_array[i] = PetscAbsScalar(f_array[i]);
}
// Restore arrays
ierr = VecRestoreArrayRead(f, &f_array); CHKERRQ(ierr);
ierr = VecRestoreArray(f_magnitude, &f_magnitude_array); CHKERRQ(ierr);
// Assemble the magnitude vectors
ierr = VecAssemblyBegin(f_magnitude); CHKERRQ(ierr);
ierr = VecAssemblyEnd(f_magnitude); CHKERRQ(ierr);
// Save the magnitude vectors to binary files
PetscViewer viewerfmag;
ierr = PetscViewerBinaryOpen(PETSC_COMM_WORLD, "output_f_mag.dat",
FILE_MODE_WRITE, &viewerfmag); CHKERRQ(ierr);
ierr = VecView(f_magnitude, viewerfmag); CHKERRQ(ierr);
ierr = PetscViewerDestroy(&viewerfmag); CHKERRQ(ierr);
In MATLAB, I am using petscBinaryRead to read the data. The complex vectors
are read, but only the real part is available. The magnitude vectors are
however read as alternating zero and non-zero elements. What went wrong?
How can I export the data correctly to MATLAB-accessible format? (I have
not configured PETSc with Matlab as I was encountering library conflict
issues)
