bulk_flux itermax

[jcwarner]

Small tiling bug in Nonlinear/bulk_flux.F due to variable Itermax. The variable Itermax is set in an i loop. Itermax takes a value of 1 or 3, depending on the last setting in the i loop.

After that i loop completes, we have

        DO Iter=1,IterMax
          DO i=Istr-1,IendR
.....

Well, the number of iterations for any given i-tile will depend on the last value of Itermax. Thus, the values in an i-tile will depend on how it is tiled.

Solution: dimension Itermax(private_1d_scratch_array) and compute Itermax(i). Then change the order of the do loops as:

        DO i=Istr-1,IendR
          DO Iter=1,IterMax(i)
.....

or just set Itermax = 3 and be done with it (?).

[arango]

Yes, good catch! Well, the original COARE code was 1-dimensional so the problem was when it was converted to two-dimensional quantities. I do agree with you that the best way to solve this problem is to set Itermax = 3 for the similarity theory look tables. This is the less complicated solution for the adjoint version of this routine.