Merge branch 'atmosphere/acc_atm_bdy_set_scalars_work' into develop (PR #1265)

This merge enables execution of the atm_bdy_set_scalars_work subroutine on GPUs
through the addition of OpenACC directives.

* atmosphere/acc_atm_bdy_set_scalars_work:
  Add OpenACC directives to enable executing atm_bdy_set_scalars_work on GPUs.

Author: mgduda

src/core_atmosphere/dynamics/mpas_atm_time_integration.F

@@ -7099,13 +7099,20 @@ subroutine atm_bdy_set_scalars_work( scalars_driving, scalars_new, &
 
       !---
 
+      MPAS_ACC_TIMER_START('atm_bdy_set_scalars_work [ACC_data_xfer]')
+      !$acc enter data copyin(scalars_new, scalars_driving)
+      MPAS_ACC_TIMER_STOP('atm_bdy_set_scalars_work [ACC_data_xfer]')
+
+      !$acc parallel default(present)
+      !$acc loop gang worker
       do iCell = cellSolveStart, cellSolveEnd ! threaded over cells
 
          if ( bdyMaskCell(iCell) > nRelaxZone) then ! specified zone
 
             !  update the specified-zone values
             !
 !DIR$ IVDEP
+           !$acc loop vector collapse(2)
             do k=1,nVertLevels
                do iScalar = 1, num_scalars
                   scalars_new(iScalar,k,iCell) = scalars_driving(iScalar,k,iCell)
@@ -7115,6 +7122,12 @@ subroutine atm_bdy_set_scalars_work( scalars_driving, scalars_new, &
          end if
 
       end do  ! updates now in temp storage
+      !$acc end parallel
+
+      MPAS_ACC_TIMER_START('atm_bdy_set_scalars_work [ACC_data_xfer]')
+      !$acc exit data copyout(scalars_new)
+      !$acc exit data delete(scalars_driving)
+      MPAS_ACC_TIMER_STOP('atm_bdy_set_scalars_work [ACC_data_xfer]')
 
    end subroutine atm_bdy_set_scalars_work