fschuch / xcompact3d Goto Github PK

I do not think the +1 is necessary when the domain is periodic in a given direction.

The gravitational term is not computed if ilmn .eq. False (transeq.f90):

    if (ilmn) then
      !! Gravity
      if ((Fr**2).gt.zero) then
        call momentum_gravity(dux1, duy1, duz1, rho1(:,:,:,1) - one, one / Fr**2)
      endif
      do is = 1, numscalar
        call momentum_gravity(dux1, duy1, duz1, phi1(:,:,:,is), ri(is))
      enddo
    endif

That is a bug for the proper simulation of gravity currents.

SUBROUTINE calc_divu_constraint (navier.f90) needs review:

• CALL derxx (tc1, phi1(:,:,:,is), di1, sx, sfxp, ssxp, swxp, xsize(1), xsize(2), xsize(3), 1);
• CALL deryy (td2, phi2(:,:,:,is), di2, sy, sfyp, ssyp, swyp, ysize(1), ysize(2), ysize(3), 1);
• CALL derzz (tc3, phi3(:,:,:,is), di3, sz, sfzp, sszp, swzp, zsize(1), zsize(2), zsize(3), 1).

SUBROUTINE continuity_rhs_eq (transqe.f90) needs review:

• call derz (ta3, rho3, di3, sz, ffzp, fszp, fwzp, zsize(1), zsize(2), zsize(3), 1);
• call dery (ta2, rho2, di2, sy, ffyp, fsyp, fwyp, ppy, ysize(1), ysize(2), ysize(3), 1);
• call derx (drho1(:,:,:,1), rho1(:,:,:,1), & di1, sx, ffxp, fsxp, fwxp, xsize(1), xsize(2), xsize(3), 1);
• call derzz (ta3,rho3,di3,sz,sfzp,sszp,swzp,zsize(1),zsize(2),zsize(3),1);
• call deryy (ta2,rho2,di2,sy,sfyp,ssyp,swyp,ysize(1),ysize(2),ysize(3),1);
• call derxx (tb1,rho1,di1,sx,sfxp,ssxp,swxp,xsize(1),xsize(2),xsize(3),1).

It is probably a restart file added by mistake. Just have to find where and how to remove it.

The binary outputs were reorganized:

• A format .bin was added to binary fields, in order to be more friendly with any applications based on filename patterns;
• The files where arranged in folders according to their shape, i.e.:
  • ./data/3d_snapshots for 3d fields (nx, ny, nz);
  • ./data/xy_planes for planes (nx, ny);
  • ./data/xz_planes for planes (nx, nz);
  • ./data/yz_planes for planes (ny, nz).
• The filenames enumeration changed, now files receive the number according to the current timestep. The previous continuous counting is still possible by setting filenamedigits = 1 at .i3d (default is 0), in addition, the number of digits can be easily chosen with ifilenameformat (default is '(I9.9)'),

All changes aim to make postprocessing easier for tools like the new xcompact3d_toolbox. Reading a bunch of files is possible with just a fill lines of Python code since the shape is inferred from the directory name and the time from filenames.

The problem was tracked to init_sandbox, more specifically, to call decomp_2d_read_one(...), I'm investigating how to solve it.

The recommendation, for now, is to use only double-precision, i.e., make sure -DDOUBLE_PREC is included as DEFS at the Makefile.

It is overwritten when numscalar.ne.0 anyway, so, why have both?

It happens because some variables are allocated and initialized at init_sandbox, and it is not called when restarting.

We should allocate and initialize boundary conditions and volume integration in another place.

For consistency, we should force phi to zero inside the geometry, just like the lagrangian interpolator does.