[Nek5000-users] Question on interp_v

[nek5000-users at lists.mcs.anl.gov]

Hi all,
I recently got some results I didn’t quite understand from the interpolation subroutines. I set up a simplified minimal example that shows the problem.
I add the following to the hemi example usrchk:
REAL*8 my_xyz(ldim,100)
&      ,my_uvw(ldim,100)

(…)

 DO i=1,100
        my_xyz(1,i)=xm1(i,1,1,1)
        my_xyz(2,i)=ym1(i,1,1,1)
        my_xyz(3,i)=zm1(i,1,1,1)
ENDDO
CALL interp_v(my_uvw,my_xyz,100)
DO i=1,100
           WRITE(*,*) '-------------------------------'
           WRITE(*,*) 'pos',my_xyz(1:3,i)
           WRITE(*,*) 'vx',vx(i,1,1,1),my_uvw(1,i),
&                     vx(i,1,1,1)-my_uvw(1,i)
           WRITE(*,*) 'vy',vy(i,1,1,1),my_uvw(2,i),
 &                     vy(i,1,1,1)-my_uvw(2,i)
           WRITE(*,*) 'vz',vz(i,1,1,1),my_uvw(3,i),
&                     vz(i,1,1,1)-my_uvw(3,i)
ENDDO
I would expect that interpolating on the grid points themselves returns the value of velocity there (plus perhaps some numerical error), but the results seem to be more or less random (only right order of magnitude). Am I using the subroutine wrong?
Jan