I guess the way to go, if you want to have full restart, is to first reduce the time step with the old resolution, save the three restart files (in the latest version only three files are necessary for order three) with the new time step but old resolution, and then restart... Philipp On 2017-05-23 22:56, [email protected] wrote:

Dear Juan Diego,

Good point.

I guess in this case I would just use the last restart file ----

The point of "full restart" is to give a carbon copy of the preceding run. If lx1 has changed, however, then the notion of a carbon copy is lost.

I do see, however, that there would be value to full restart with different lx1.

Paul

------------------------------------------------------------------------ *From:* [email protected] [[email protected]] on behalf of [email protected] [[email protected]] *Sent:* Tuesday, May 23, 2017 3:34 PM *To:* [email protected] *Subject:* Re: [Nek5000-users] Full restart with increasing lx ([email protected])

Hello,

I've been able to use the full restart function increasing lx1, but in my case I had to change the timestep size (DT) in order to satisfy the CFL condition, so the solution that was saved did not work properly with the full restart because it was obtained using a different DT. So this is something for you to consider.

Best,

Juan Diego

_______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users

Dear Hu, What information are you needing from another element. The usual approach, which appears relevant to your situation, is to recognize that communication is generally reciprocal: If element e needs data from element e', then e' also likely needs the same date from e. These operations are most readily handled through the gather-scatter operator, gs_op(), which can exchange data among participating processors. (All processors must participate, but some might not have data to exchange.) Are you trying to effect local averages? or, what? Many such operations can also be effected through existing spectral element operators, etc. hth, Paul ________________________________ From: Nek5000-users <[email protected]> on behalf of [email protected] <[email protected]> Sent: Monday, July 24, 2017 9:06:50 PM To: [email protected] Subject: [Nek5000-users] How to get another element information that in the different processor? Hi all, I want to get another element information when I handle one element, and the two elements maybe not in the same processor. How to get another element information that in the different processor? Below is some of my code: 1)for element e, get global element number; 2)get the correspond element number at the inlet; 3)call the ym1 value at the inlet element But I don't know how to get ym1(i,j,k,e_ele) in the e_nid processor. do e=1,nelv eg=lglel(e) call get_exyz(ex,ey,ez,eg,nelx,nely,nelz) e_in=(ez-1)*nelx*nely+(ey-1)*nelx+1 e_nid=gllnid(e_in) e_ele=gllel(e_in) c if (nid.eq.e_nid) then do k=1,nz1 do j=1,ny1 do i=1,nx1 temp=-ym1(i,j,k,e)*u_t/vis/A_plus cdyn(i,j,k,e)=cs0*(1-exp(temp)) enddo enddo enddo c endif enddo Best wishes! Hu _______________________________________________ Nek5000-users mailing list [email protected] https://lists.mcs.anl.gov/mailman/listinfo/nek5000-users