<p><b>mpetersen@lanl.gov</b> 2010-10-29 10:57:23 -0600 (Fri, 29 Oct 2010)</p><p>TRUNK COMMIT, sw core<br>
<br>
Removing unnecessary if statements from time_integration module.<br>
See attached emails for description. I am making these same revisions <br>
in the ocean core, but in the topography branch that will be merged<br>
soon.<br>
<br>
Date: Thu, 28 Oct 2010 15:48:25 -0600 (MDT)<br>
From: Mark Petersen <mpetersen@lanl.gov><br>
To: MPAS developers list <mpas-developers@ucar.edu><br>
Subject: [mpas-developers] nVertices+1<br>
<br>
Michael and others,<br>
<br>
In the sw and ocean core, there is a section (pasted below) with the line<br>
if (verticesOnEdge(1,iEdge) <= nVertices) then<br>
<br>
I understand that a cell numbered nCells+1 is outside of the local domain.<br>
But all vertices are always within the local domain, right? That is, I<br>
think verticesOnEdge <= nVertices is always true.<br>
<br>
If you agree, I will remove the if lines. Actually, this was already done<br>
in core_hyd_atmos, so I can just copy it over to sw, and make similar<br>
changes to the ocean core.<br>
<br>
Date: Thu, 28 Oct 2010 16:48:37 -0600<br>
From: Michael Duda <duda@ucar.edu><br>
To: Mark Petersen <mpetersen@lanl.gov><br>
Cc: MPAS developers list <mpas-developers@ucar.edu><br>
Subject: Re: [mpas-developers] nVertices+1<br>
<br>
Hi, Mark.<br>
<br>
I agree -- it should never be the case that we have an edge in the local<br>
domain (owned + ghost edges) but not both of its endpoints; in other<br>
words, verticesOnEdge(:,1:nEdges) should never be greater than<br>
nVertices. Please feel free to remove these if-tests whenever it's<br>
convenient!<br>
<br>
Cheers,<br>
Michael<br>
<br>
<br>
Date: Thu, 28 Oct 2010 16:23:33 -0600 (MDT)<br>
From: Mark Petersen <mpetersen@lanl.gov><br>
To: MPAS developers list <mpas-developers@ucar.edu><br>
Subject: [mpas-developers] two more if statements<br>
<br>
<br>
Here are two more places to remove if statements, where they have remained<br>
in the sw and ocean core but have already been removed from the hyd_atm<br>
core. Again, I think they are unnecessary. If everyone agrees, I will<br>
update the sw and ocean cores<br>
<br>
Date: Thu, 28 Oct 2010 16:55:47 -0600<br>
From: Michael Duda <duda@ucar.edu><br>
To: Mark Petersen <mpetersen@lanl.gov><br>
Cc: MPAS developers list <mpas-developers@ucar.edu><br>
Subject: Re: [mpas-developers] two more if statements<br>
<br>
Hi, Mark.<br>
<br>
I also agree that the tests in both of the sections you've identified<br>
below should also be safe to remove. In both cases below, iEdge or eoe<br>
may indeed reference an edge outside the local block, but this reference<br>
would have been switched to the index of the "garbage edge" during<br>
domain setup, and we expect that the values from the garbage edge should<br>
never influence the values of owned cells/edges/vertices. Please feel free<br>
to eliminate these if-tests, too, and thanks for taking such a careful<br>
look through the code!<br>
<br>
Cheers,<br>
Michael<br>
</p><hr noshade><pre><font color="gray">Modified: trunk/mpas/src/core_sw/module_time_integration.F
===================================================================
--- trunk/mpas/src/core_sw/module_time_integration.F 2010-10-28 02:11:18 UTC (rev 584)
+++ trunk/mpas/src/core_sw/module_time_integration.F 2010-10-29 16:57:23 UTC (rev 585)
@@ -779,16 +779,10 @@
!
circulation(:,:) = 0.0
do iEdge=1,nEdges
- if (verticesOnEdge(1,iEdge) <= nVertices) then
- do k=1,nVertLevels
- circulation(k,verticesOnEdge(1,iEdge)) = circulation(k,verticesOnEdge(1,iEdge)) - dcEdge(iEdge) * u(k,iEdge)
- end do
- end if
- if (verticesOnEdge(2,iEdge) <= nVertices) then
- do k=1,nVertLevels
- circulation(k,verticesOnEdge(2,iEdge)) = circulation(k,verticesOnEdge(2,iEdge)) + dcEdge(iEdge) * u(k,iEdge)
- end do
- end if
+ do k=1,nVertLevels
+ circulation(k,verticesOnEdge(1,iEdge)) = circulation(k,verticesOnEdge(1,iEdge)) - dcEdge(iEdge) * u(k,iEdge)
+ circulation(k,verticesOnEdge(2,iEdge)) = circulation(k,verticesOnEdge(2,iEdge)) + dcEdge(iEdge) * u(k,iEdge)
+ end do
end do
do iVertex=1,nVertices
do k=1,nVertLevels
@@ -845,11 +839,9 @@
do iEdge = 1,nEdges
do i=1,nEdgesOnEdge(iEdge)
eoe = edgesOnEdge(i,iEdge)
- if (eoe <= nEdges) then
- do k = 1,nVertLevels
- v(k,iEdge) = v(k,iEdge) + weightsOnEdge(i,iEdge) * u(k, eoe)
- end do
- end if
+ do k = 1,nVertLevels
+ v(k,iEdge) = v(k,iEdge) + weightsOnEdge(i,iEdge) * u(k, eoe)
+ end do
end do
end do
@@ -902,20 +894,19 @@
!
! Compute pv at the edges
- ! ( this computes pv_edge at all edges bounding real cells and distance-1 ghost cells )
+ ! ( this computes pv_edge at all edges bounding real cells )
!
pv_edge(:,:) = 0.0
do iVertex = 1,nVertices
do i=1,grid % vertexDegree
- iEdge = edgesOnVertex(i,iVertex)
- if(iEdge <= nEdges) then
- do k=1,nVertLevels
+ iEdge = edgesOnVertex(i,iVertex)
+ do k=1,nVertLevels
pv_edge(k,iEdge) = pv_edge(k,iEdge) + 0.5 * pv_vertex(k,iVertex)
- enddo
- endif
+ end do
end do
end do
+
!
! Modify PV edge with upstream bias.
!
</font>
</pre>