Setting fluxp == 2 is supposed to cause all flux moments to be output to the flux file. However, the Fortran reference implementation of SUBROUTINE output_flux_file erroneously prints multiple copies of moment 1 (the isotropic component) where moments 2 through cmom (the anisotropic components) are supposed to be.
I think the offending statement is here on line 332 of output.f90, which only reads from flux0 instead of conditionally reading from flux0 or fluxm depending on the value of l. Note that the output_send calls on line 322 do correctly implement this conditional logic:
IF ( l == 1 ) THEN
CALL output_send ( mtag, flux0(:,:,kloc,g) )
ELSE
CALL output_send ( mtag, fluxm(l-1,:,:,kloc,g) )
END IFHowever, in non-MPI runs, output_send is a no-op, causing the program to revert to faulty single-rank behavior. In addition, the FORMAT statement on line 370 uses the I1 specifier:
323 FORMAT( /, 2X, 'Moment = ', I1 )
This causes the erroneous output Moment = * to be printed whenever the moment index is 10 or greater (which commonly occurs in 2D and 3D problems whenever nmom >= 4). The I0 specifier should be used instead to allow arbitrarily large integers to be printed without unnecessary padding.
Setting
fluxp == 2is supposed to cause all flux moments to be output to the flux file. However, the Fortran reference implementation ofSUBROUTINE output_flux_fileerroneously prints multiple copies of moment 1 (the isotropic component) where moments 2 throughcmom(the anisotropic components) are supposed to be.I think the offending statement is here on line 332 of output.f90, which only reads from
flux0instead of conditionally reading fromflux0orfluxmdepending on the value ofl. Note that theoutput_sendcalls on line 322 do correctly implement this conditional logic:However, in non-MPI runs,
output_sendis a no-op, causing the program to revert to faulty single-rank behavior. In addition, theFORMATstatement on line 370 uses theI1specifier:This causes the erroneous output
Moment = *to be printed whenever the moment index is 10 or greater (which commonly occurs in 2D and 3D problems whenevernmom >= 4). TheI0specifier should be used instead to allow arbitrarily large integers to be printed without unnecessary padding.