Difference between revisions of "Adding dH/dt module"

From Interactive System for Ice sheet Simulation
Jump to: navigation, search
Line 1: Line 1:
 
This page contains step-by-step instructions for adding a new module to Glimmer-CISM. In this case, the module is a first-order, upwinding advection scheme for mass transport (dH/dt) using velocities calculated from a higher-order dynamics model. The procedure, however, is generic and could apply to adding almost any module.
 
This page contains step-by-step instructions for adding a new module to Glimmer-CISM. In this case, the module is a first-order, upwinding advection scheme for mass transport (dH/dt) using velocities calculated from a higher-order dynamics model. The procedure, however, is generic and could apply to adding almost any module.
  
===step 1 ===
+
===Step 1 ===
 
(i) Create a module '''for_upwind_advect_mod.F90''' that contains the necessary subroutines. For now, these will just be "stubs", which we will fill in later.
 
(i) Create a module '''for_upwind_advect_mod.F90''' that contains the necessary subroutines. For now, these will just be "stubs", which we will fill in later.
  
Line 59: Line 59:
  
 
end module fo_upwind_advect</source>
 
end module fo_upwind_advect</source>
 +
 +
'''NOTE''': This may seem like extra work ... why don't we just call the subroutine ''fo_upwind_advect_main'' directly from the main code? In this case, perhaps this is true. In general, however, having a "driver" routine is good practice. For example, there might be additional pre/post processing of variables that needs to take place before and after the call to the subroutine ''fo_upwind_advect_main''. Making space for that here rather than in the main program, where only the driver is called, makes more sense and keeps the main code cleaner and easier to understand.

Revision as of 10:55, 22 July 2009

This page contains step-by-step instructions for adding a new module to Glimmer-CISM. In this case, the module is a first-order, upwinding advection scheme for mass transport (dH/dt) using velocities calculated from a higher-order dynamics model. The procedure, however, is generic and could apply to adding almost any module.

Step 1

(i) Create a module for_upwind_advect_mod.F90 that contains the necessary subroutines. For now, these will just be "stubs", which we will fill in later.

module fo_upwind_advect_mod
 
! subroutines for mass advection scheme based on 1st order upwinding
 
    contains
 
!----------------------------------------------------------------------
 
    subroutine fo_upwind_advect_init( )
 
    ! initialization for 1st-order upwinding mass advection
 
    end subroutine fo_upwind_advect_init
 
!----------------------------------------------------------------------
 
    subroutine fo_upwind_advect_final( )
 
    ! finalization for 1st-order upwinding mass advection
 
    end subroutine fo_upwind_advect_final
 
!----------------------------------------------------------------------
 
    subroutine fo_upwind_advect_main( )
 
    ! 1st-order upwinding mass advection
 
    end subroutine fo_upwind_advect_main
 
!----------------------------------------------------------------------
 
end module fo_upwind_advect_mod

(ii) create another module for_upwind_advect.F90 that will contain the "driver" subroutine - that which is called from the main code, which in this case is glide.F90 ...

module fo_upwind_advect
 
    ! this is the driver routine for the 1st order, upwind mass transport scheme
 
    use fo_upwind_advect_mod, only : fo_upwind_advect_main
 
    implicit none
    private
    public :: fo_upwind_advect_driver
 
    contains
 
    subroutine fo_upwind_advect_driver( )
 
        call fo_upwind_advect_main( )
 
    end subroutine fo_upwind_advect_driver
 
end module fo_upwind_advect

NOTE: This may seem like extra work ... why don't we just call the subroutine fo_upwind_advect_main directly from the main code? In this case, perhaps this is true. In general, however, having a "driver" routine is good practice. For example, there might be additional pre/post processing of variables that needs to take place before and after the call to the subroutine fo_upwind_advect_main. Making space for that here rather than in the main program, where only the driver is called, makes more sense and keeps the main code cleaner and easier to understand.