Custom Query (986 matches)

Corrected a bug in step2d_LF_AM3.h with subscript out range for local arrays DiaU2rhs and DiaV2rhs. This bug was introduced recently in ​src:ticket:403.

Many thanks to Joe Jurisa and Carlos Teixeira for ​reporting this problem.

NOTICE: The file varinfo.dat was changed in ​src:ticket:403 to include the new definitions for the additional diagnostics variables. Make sure that you use the latest version of varinfo.dat when DIAGNOSTICS_TS and/or DIAGNOSTICE_UV are activated.

Corrected a bug in get_ngfld.F and get_ngfldr.F when the scale_factor and add_offset NetCDF attributes are used to compress the data. The generic routines in mod_netcdf.F already process these paramemeters to uncompress the data. So the additional scaling in get_ngfld.F and get_ngfldr.F routines is only related with Fscale(ifield,ng) and not compression:

            Fmin=A(1)*Fscale(ifield,ng)
            Fmax=A(1)*Fscale(ifield,ng)
            ic=0
            DO j=1,Jrec
              DO i=Istr,Iend
                ic=ic+1
                Aval=A(ic)*Fscale(ifield,ng)
                Fmin=MIN(Fmin,Aval)
                Fmax=MAX(Fmax,Aval)
                Fout(i,j,Tindex)=Aval
              END DO
            END DO

Notice that the Fscale factor is a different issue and doesn't have nothing to do with the compression of data but field units manipulation from varinfo.dat. Usually, Fscale=1.0 for mostly all input fields.

Many thanks to Mark Hadfield for ​reporting this problem.

The same correction is also needed for get_2dfld.F, get_2dfldr.F, get_3dfld.F, and get_3dfldr.F when processing point-data. However, it very unlikely to have a input NetCDF file with point-data that needs to be compressed as byte and scaled with the scale_factor and add_offset attributes.

During the coupling of ROMS to SWAN, in the file SWAN/Src/waves_coupler.F, there is a call

      CALL GlobalSegMap_Ordpnts (GSMapSWAN, MyRank, points)

This call uses MCT to obtain the local grid points for each tile. The corresponding mct routine allocates memory for the variable points, and then computes those values. In the waves_coupler routine, I also had allocated the variable points and then deallocated the memory. However, since both MCT and I were allocating the memory, it led to a slow but steady increase in memory cache. This only seemed to occur on some compilers for certain applications, and was extremely difficult to identify. The correction is to comment out lines 295 and 297 in waves_coupler.F. This changes the code from

!
!  Allocate attribute vector array used to export/import data.
!
      allocate ( avdata(gsmsize),stat=ierr )
      allocate ( points(gsmsize),stat=ierr )
      avdata=0.0_r8
      points=0

change to:

!
!  Allocate attribute vector array used to export/import data.
!
      allocate ( avdata(gsmsize),stat=ierr )
!!    allocate ( points(gsmsize),stat=ierr )
      avdata=0.0_r8
!!    points=0

I tried this on ifort and pgi and it works fine.