/wrfv2_fire/frame/module_io_quilt.F

!WRF:DRIVER_LAYER:IO
!
#define DEBUG_LVL 50
!#define mpi_x_comm_size(i,j,k)  Mpi_Comm_Size ( i,j,k ) ; write(0,*) __LINE__
#define mpi_x_comm_size(i,j,k)  Mpi_Comm_Size ( i,j,k )

! Workaround for bug in the IBM MPI implementation.  Look near the
! bottom of this file for an explanation.
#ifdef IBM_REDUCE_BUG_WORKAROUND
#define mpi_x_reduce(sb,rb,c,dt,op,r,com,ierr) reduce_add_integer(sb,rb,c,r,com)
#else
#define mpi_x_reduce(sb,rb,c,dt,op,r,com,ierr) MPI_Reduce(sb,rb,c,dt,op,r,com,ierr)
#endif

MODULE module_wrf_quilt
!<DESCRIPTION>
!<PRE>
! This module contains WRF-specific I/O quilt routines called by both 
! client (compute) and server (I/O quilt) tasks.  I/O quilt servers are 
! a run-time optimization that allow I/O operations, executed on the I/O 
! quilt server tasks, to be overlapped with useful computation, executed on 
! the compute tasks.  Since I/O operations are often quite slow compared to 
! computation, this performance optimization can increase parallel 
! efficiency.  
!
! Currently, one group of I/O servers can be specified at run-time.  Namelist 
! variable "nio_tasks_per_group" is used to specify the number of I/O server 
! tasks in this group.  In most cases, parallel efficiency is optimized when 
! the minimum number of I/O server tasks are used.  If memory needed to cache 
! I/O operations fits on a single processor, then set nio_tasks_per_group=1.  
! If not, increase the number of I/O server tasks until I/O operations fit in 
! memory.  In the future, multiple groups of I/O server tasks will be 
! supported.  The number of groups will be specified by namelist variable 
! "nio_groups".  For now, nio_groups must be set to 1.  Currently, I/O servers 
! only support overlap of output operations with computation.  Also, only I/O 
! packages that do no support native parallel I/O may be used with I/O server 
! tasks.  This excludes PHDF5 and MCEL.  
!
! In this module, the I/O quilt server tasks call package-dependent 
! WRF-specific I/O interfaces to perform I/O operations requested by the 
! client (compute) tasks.  All of these calls occur inside subroutine 
! quilt().  
! 
! The client (compute) tasks call package-independent WRF-specific "quilt I/O" 
! interfaces that send requests to the I/O quilt servers.  All of these calls 
! are made from module_io.F.  
!
! These routines have the same names and (roughly) the same arguments as those 
! specified in the WRF I/O API except that:
! - "Quilt I/O" routines defined in this file and called by routines in 
!   module_io.F have the "wrf_quilt_" prefix.
! - Package-dependent routines called from routines in this file are defined 
!   in the external I/O packages and have the "ext_" prefix.
!
! Both client (compute) and server tasks call routine init_module_wrf_quilt() 
! which then calls setup_quilt_servers() determine which tasks are compute 
! tasks and which are server tasks.  Before the end of init_module_wrf_quilt() 
! server tasks call routine quilt() and remain there for the rest of the model 
! run.  Compute tasks return from init_module_wrf_quilt() to perform model 
! computations.  
!
! See http://www.mmm.ucar.edu/wrf/WG2/software_2.0/IOAPI.doc for the latest
! version of the WRF I/O API.  This document includes detailed descriptions
! of subroutines and their arguments that are not duplicated here.
!</PRE>
!</DESCRIPTION>
  USE module_internal_header_util
  USE module_timing

  INTEGER, PARAMETER :: int_num_handles = 99
  INTEGER, PARAMETER :: max_servers = int_num_handles+1  ! why +1?
  LOGICAL, DIMENSION(int_num_handles) :: okay_to_write, int_handle_in_use, okay_to_commit
  INTEGER, DIMENSION(int_num_handles) :: int_num_bytes_to_write, io_form
  REAL, POINTER,SAVE :: int_local_output_buffer(:)
  INTEGER,      SAVE :: int_local_output_cursor
  LOGICAL          :: quilting_enabled
  LOGICAL          :: disable_quilt = .FALSE.
  INTEGER          :: prev_server_for_handle = -1
  INTEGER          :: server_for_handle(int_num_handles)
  INTEGER          :: reduced(2), reduced_dummy(2)
  LOGICAL, EXTERNAL :: wrf_dm_on_monitor

  INTEGER :: mpi_comm_avail,availrank
  LOGICAL :: in_avail=.false., poll_servers=.false.

  INTEGER nio_groups
#ifdef DM_PARALLEL
  INTEGER :: mpi_comm_local
  LOGICAL :: compute_node
  LOGICAL :: compute_group_master(max_servers)
  INTEGER :: mpi_comm_io_groups(max_servers)
  INTEGER :: nio_tasks_in_group
  INTEGER :: nio_tasks_per_group
  INTEGER :: ncompute_tasks
  INTEGER :: ntasks
  INTEGER :: mytask

  INTEGER, PARAMETER           :: onebyte = 1
  INTEGER comm_io_servers, iserver, hdrbufsize, obufsize
  INTEGER, DIMENSION(4096)     :: hdrbuf
  INTEGER, DIMENSION(int_num_handles)     :: handle
#endif

#ifdef IBM_REDUCE_BUG_WORKAROUND
! Workaround for bug in the IBM MPI implementation.  Look near the
! bottom of this file for an explanation.
  interface reduce_add_integer
     module procedure reduce_add_int_arr
     module procedure reduce_add_int_scl
  end interface
#endif

  CONTAINS

#if  defined(DM_PARALLEL)  &&  !defined( STUBMPI )
    INTEGER FUNCTION get_server_id ( dhandle )
!<DESCRIPTION>
! Logic in the client side to know which io server
! group to send to. If the unit corresponds to a file that's
! already been opened, then we have no choice but to send the
! data to that group again, regardless of whether there are
! other server-groups. If it's a new file, we can chose a new
! server group. I.e. opening a file locks it onto a server
! group. Closing the file unlocks it.
!</DESCRIPTION>
      IMPLICIT NONE
      INTEGER, INTENT(IN) :: dhandle
      IF ( dhandle .GE. 1 .AND. dhandle .LE. int_num_handles ) THEN
        IF ( server_for_handle ( dhandle ) .GE. 1 ) THEN
          get_server_id = server_for_handle ( dhandle )
        ELSE
           IF(poll_servers) THEN
              ! Poll server group masters to find an inactive I/O server group:
              call wrf_quilt_find_server(server_for_handle(dhandle))
           ELSE
              ! Server polling is disabled, so cycle through servers:
              prev_server_for_handle = mod ( prev_server_for_handle + 1 , nio_groups )
              server_for_handle( dhandle ) = prev_server_for_handle+1
           ENDIF
           get_server_id=server_for_handle(dhandle)
        ENDIF
      ELSE
         CALL wrf_message('module_io_quilt: get_server_id bad dhandle' )
      ENDIF
    END FUNCTION get_server_id
#endif

    SUBROUTINE set_server_id ( dhandle, value )
       IMPLICIT NONE
       INTEGER, INTENT(IN) :: dhandle, value
       IF ( dhandle .GE. 1 .AND. dhandle .LE. int_num_handles ) THEN
         server_for_handle(dhandle) = value
       ELSE
         CALL wrf_message('module_io_quilt: set_server_id bad dhandle' )
       ENDIF
    END SUBROUTINE set_server_id

    LOGICAL FUNCTION get_poll_servers() 
      implicit none
      get_poll_servers=poll_servers
    end FUNCTION get_poll_servers

#if defined( DM_PARALLEL ) && !defined( STUBMPI )
    SUBROUTINE int_get_fresh_handle( retval )
!<DESCRIPTION>
! Find an unused "client file handle" and return it in retval.
! The "client file handle" is used to remember how a file was opened
! so clients do not need to ask the I/O quilt servers for this information.
! It is also used as a file identifier in communications with the I/O
! server task.
!
! Note that client tasks know nothing about package-specific handles.
! Only the I/O quilt servers know about them.
!</DESCRIPTION>
      INTEGER i, retval
      retval = -1
      DO i = 1, int_num_handles
        IF ( .NOT. int_handle_in_use(i) )  THEN
          retval = i
          GOTO 33
        ENDIF
      ENDDO
33    CONTINUE
      IF ( retval < 0 )  THEN
        CALL wrf_error_fatal("frame/module_io_quilt.F: int_get_fresh_handle() can not")
      ENDIF
      int_handle_in_use(i) = .TRUE.
      NULLIFY ( int_local_output_buffer )
    END SUBROUTINE int_get_fresh_handle

    SUBROUTINE setup_quilt_servers ( nio_tasks_per_group,     &
                                     mytask,                  &
                                     ntasks,                  &
                                     n_groups_arg,            &
                                     nio,                     &
                                     mpi_comm_wrld,           &
                                     mpi_comm_local,          &
                                     mpi_comm_io_groups)
!<DESCRIPTION>
! Both client (compute) and server tasks call this routine to 
! determine which tasks are compute tasks and which are I/O server tasks.  
!
! Module variables MPI_COMM_LOCAL and MPI_COMM_IO_GROUPS(:) are set up to 
! contain MPI communicators as follows:  
!
! MPI_COMM_LOCAL is the Communicator for the local groups of tasks. For the 
! compute tasks it is the group of compute tasks; for a server group it the 
! communicator of tasks in the server group.
!
! Elements of MPI_COMM_IO_GROUPS are communicators that each contain one or 
! more compute tasks and a single I/O server assigned to those compute tasks.  
! The I/O server tasks is always the last task in these communicators.  
! On a compute task, which has a single associate in each of the server 
! groups, MPI_COMM_IO_GROUPS is treated as an array; each element corresponds 
! to a different server group. 
! On a server task only the first element of MPI_COMM_IO_GROUPS is used 
! because each server task is part of only one io_group.  
!
! I/O server tasks in each I/O server group are divided among compute tasks as 
! evenly as possible.  
!
! When multiple I/O server groups are used, each must have the same number of 
! tasks.  When the total number of extra I/O tasks does not divide evenly by 
! the number of io server groups requested, the remainder tasks are not used 
! (wasted).  
!
! For example, communicator membership for 18 tasks with nio_groups=2 and 
! nio_tasks_per_group=3 is shown below:  
!
!<PRE>
! Membership for MPI_COMM_LOCAL communicators:
!   COMPUTE TASKS:          0   1   2   3   4   5   6   7   8   9  10  11
!   1ST I/O SERVER GROUP:  12  13  14
!   2ND I/O SERVER GROUP:  15  16  17
!
! Membership for MPI_COMM_IO_GROUPS(1):  
!   COMPUTE TASKS 0, 3, 6, 9:   0   3   6   9  12
!   COMPUTE TASKS 1, 4, 7,10:   1   4   7  10  13
!   COMPUTE TASKS 2, 5, 8,11:   2   5   8  11  14
!   I/O SERVER TASK       12:   0   3   6   9  12
!   I/O SERVER TASK       13:   1   4   7  10  13
!   I/O SERVER TASK       14:   2   5   8  11  14
!   I/O SERVER TASK       15:   0   3   6   9  15
!   I/O SERVER TASK       16:   1   4   7  10  16
!   I/O SERVER TASK       17:   2   5   8  11  17
!
! Membership for MPI_COMM_IO_GROUPS(2):  
!   COMPUTE TASKS 0, 3, 6, 9:   0   3   6   9  15
!   COMPUTE TASKS 1, 4, 7,10:   1   4   7  10  16
!   COMPUTE TASKS 2, 5, 8,11:   2   5   8  11  17
!   I/O SERVER TASK       12:  ** not used **
!   I/O SERVER TASK       13:  ** not used **
!   I/O SERVER TASK       14:  ** not used **
!   I/O SERVER TASK       15:  ** not used **
!   I/O SERVER TASK       16:  ** not used **
!   I/O SERVER TASK       17:  ** not used **
!</PRE>
!</DESCRIPTION>
      USE module_configure
#ifdef DM_PARALLEL
      USE module_dm, ONLY : compute_mesh
#endif
      IMPLICIT NONE
      INCLUDE 'mpif.h'
      INTEGER,                      INTENT(IN)  :: nio_tasks_per_group, mytask, ntasks, &
                                                   n_groups_arg, mpi_comm_wrld
      INTEGER,  INTENT(OUT)                     :: mpi_comm_local, nio
      INTEGER, DIMENSION(100),      INTENT(OUT) :: mpi_comm_io_groups
! Local
      INTEGER                     :: i, j, ii, comdup, ierr, niotasks, n_groups, iisize
      INTEGER, DIMENSION(ntasks)  :: icolor
      CHARACTER*128 mess
      INTEGER :: io_form_setting
      INTEGER :: me
      INTEGER :: k, m, nprocx, nprocy
      LOGICAL :: reorder_mesh

!check the namelist and make sure there are no output forms specified
!that cannot be quilted
      CALL nl_get_io_form_history(1,   io_form_setting) ; call sokay( 'history', io_form_setting )
      CALL nl_get_io_form_restart(1,   io_form_setting) ; call sokay( 'restart', io_form_setting )
      CALL nl_get_io_form_auxhist1(1,  io_form_setting) ; call sokay( 'auxhist1', io_form_setting )
      CALL nl_get_io_form_auxhist2(1,  io_form_setting) ; call sokay( 'auxhist2', io_form_setting )
      CALL nl_get_io_form_auxhist3(1,  io_form_setting) ; call sokay( 'auxhist3', io_form_setting )
      CALL nl_get_io_form_auxhist4(1,  io_form_setting) ; call sokay( 'auxhist4', io_form_setting )
      CALL nl_get_io_form_auxhist5(1,  io_form_setting) ; call sokay( 'auxhist5', io_form_setting )
      CALL nl_get_io_form_auxhist6(1,  io_form_setting) ; call sokay( 'auxhist6', io_form_setting )
      CALL nl_get_io_form_auxhist7(1,  io_form_setting) ; call sokay( 'auxhist7', io_form_setting )
      CALL nl_get_io_form_auxhist8(1,  io_form_setting) ; call sokay( 'auxhist8', io_form_setting )
      CALL nl_get_io_form_auxhist9(1,  io_form_setting) ; call sokay( 'auxhist9', io_form_setting )
      CALL nl_get_io_form_auxhist10(1, io_form_setting) ; call sokay( 'auxhist10', io_form_setting )
      CALL nl_get_io_form_auxhist11(1, io_form_setting) ; call sokay( 'auxhist11', io_form_setting )

      n_groups = n_groups_arg
      IF ( n_groups .LT. 1 ) n_groups = 1

      compute_node = .TRUE.

!<DESCRIPTION>
! nio is number of io tasks per group.  If there arent enough tasks to satisfy
! the requirement that there be at least as many compute tasks as io tasks in
! each group, then just print a warning and dump out of quilting
!</DESCRIPTION>

      nio = nio_tasks_per_group
      ncompute_tasks = ntasks - (nio * n_groups)
      IF ( ncompute_tasks .LT. nio ) THEN 
        WRITE(mess,'("Not enough tasks to have ",I3," groups of ",I3," I/O tasks. No quilting.")')n_groups,nio
        nio            = 0
        ncompute_tasks = ntasks
      ELSE                                   
        WRITE(mess,'("Quilting with ",I3," groups of ",I3," I/O tasks.")')n_groups,nio
      ENDIF                                   
      CALL wrf_message(mess)

      IF ( nio .LT. 0 ) THEN
        nio = 0
      ENDIF
      IF ( nio .EQ. 0 ) THEN
        quilting_enabled = .FALSE.
        mpi_comm_local = mpi_comm_wrld
        mpi_comm_io_groups = mpi_comm_wrld
        RETURN
      ENDIF
      quilting_enabled = .TRUE.

! First construct the local communicators
! prepare to split the communicator by designating compute-only tasks
      DO i = 1, ncompute_tasks
        icolor(i) = 0
      ENDDO
      ii = 1
! and designating the groups of i/o tasks
      DO i = ncompute_tasks+1, ntasks, nio
        DO j = i, i+nio-1
          icolor(j) = ii
        ENDDO
        ii = ii+1
      ENDDO
      CALL MPI_Comm_dup(mpi_comm_wrld,comdup,ierr)
      CALL MPI_Comm_split(comdup,icolor(mytask+1),mytask,mpi_comm_local,ierr)

! Now construct the communicators for the io_groups
      CALL nl_get_reorder_mesh(1,reorder_mesh)
      IF ( reorder_mesh ) THEN
        reorder_mesh = .FALSE.
        CALL nl_set_reorder_mesh(1,reorder_mesh)
        CALL wrf_message('Warning: reorder_mesh does not work with quilting. Disabled reorder_mesh.')
      ENDIF
      ! assign the compute tasks to the i/o tasks in full rows
      CALL compute_mesh( ncompute_tasks, nprocx, nprocy )

      nio = min(nio,nprocy)
      m = mod(nprocy,nio)  ! divide up remainder, 1 row per, until gone
      ii = 1
      DO j = 1, nio, 1
         DO k = 1,nprocy/nio+min(m,1)
           DO i = 1, nprocx
             icolor(ii) = j - 1
             ii = ii + 1
           ENDDO
         ENDDO
         m = max(m-1,0)
      ENDDO

! ... and add the io servers as the last task in each group
      DO j = 1, n_groups
        ! TBH:  each I/O group will contain only one I/O server
        DO i = ncompute_tasks+1,ntasks
          icolor(i) = MPI_UNDEFINED
        ENDDO
        ii = 0
        DO i = ncompute_tasks+(j-1)*nio+1,ncompute_tasks+j*nio
          icolor(i) = ii
          ii = ii+1
        ENDDO
        CALL MPI_Comm_dup(mpi_comm_wrld,comdup,ierr)
        CALL MPI_Comm_split(comdup,icolor(mytask+1),mytask, &
                            mpi_comm_io_groups(j),ierr)
      ENDDO

#ifdef PNETCDF_QUILT
      if(poll_servers) then
         poll_servers=.false.
         call wrf_message('Warning: server polling does not work with pnetcdf_quilt.  Disabled poll_servers.')
      else
#endif
         if(nio_groups==1) then
            poll_servers=.false.
            call wrf_message('Server polling is useless with one io group.  Disabled poll_servers.')
         endif
#ifdef PNETCDF_QUILT
      endif
#endif

      if(poll_servers) then
         ! If server polling is enabled, we need to create mpi_comm_avail, 
         ! which contains the monitor process, and the I/O server master process
         ! for each I/O server group.  This will be used in the routines
         ! wrf_quilt_find_server and wrf_quilt_server_ready to find inactive
         ! I/O servers for new data handles in get_server_id.

         ! The "in_avail" is set to true iff I am in the mpi_comm_avail.

         call mpi_comm_rank(mpi_comm_wrld,me,ierr)

         icolor=MPI_UNDEFINED
         in_avail=.false.

         if(wrf_dm_on_monitor()) then
            in_avail=.true. ! monitor process is in mpi_comm_avail
         endif
         icolor(1)=1

         do j=1,n_groups
            i=ncompute_tasks+j*nio-1
            if(me+1==i) then
               in_avail=.true. ! I/O server masters are in mpi_comm_avail
            endif
            icolor(i)=1
         enddo

         CALL MPI_Comm_dup(mpi_comm_wrld,comdup,ierr)
         CALL MPI_Comm_split(comdup,icolor(me+1),me, &
                             mpi_comm_avail,ierr)

         availrank=MPI_UNDEFINED
         if(in_avail) then
            call mpi_comm_rank(mpi_comm_avail,availrank,ierr)
         endif

      endif

      compute_group_master = .FALSE.
      compute_node         = .FALSE.

      DO j = 1, n_groups

         IF ( mytask .LT. ncompute_tasks .OR.                                                  &    ! I am a compute task
              (ncompute_tasks+(j-1)*nio .LE. mytask .AND. mytask .LT. ncompute_tasks+j*nio) &    ! I am the I/O server for this group
            ) THEN

         CALL MPI_Comm_Size( mpi_comm_io_groups(j) , iisize, ierr )
         ! Get the rank of this compute task in the compute+io 
         ! communicator to which it belongs
         CALL MPI_Comm_Rank( mpi_comm_io_groups(j) , me , ierr )

         ! If I am an I/O server for this group then make that group's
         ! communicator the first element in the mpi_comm_io_groups array 
         ! (I will ignore all of the other elements).
         IF (ncompute_tasks+(j-1)*nio .LE. mytask .AND. mytask .LT. ncompute_tasks+j*nio) THEN
            mpi_comm_io_groups(1) = mpi_comm_io_groups(j)
         ELSE
            compute_node = .TRUE.
            ! If I am a compute task, check whether I am the member of my 
            ! group that will communicate things that should be sent just 
            ! once (e.g. commands) to the IO server of my group.
            compute_group_master(j) = (me .EQ. 0)

!            IF( compute_group_master(j) ) WRITE(*,*) mytask,': ARPDBG : I will talk to IO server in group ',j
         ENDIF
         ENDIF
      ENDDO

    END SUBROUTINE setup_quilt_servers

    SUBROUTINE sokay ( stream, io_form )
    USE module_state_description
    CHARACTER*(*) stream
    CHARACTER*256 mess
    INTEGER io_form

    SELECT CASE (io_form)
#ifdef NETCDF
      CASE ( IO_NETCDF   )
         RETURN
#endif
#ifdef INTIO
      CASE ( IO_INTIO   )
         RETURN
#endif
#ifdef YYY
      CASE ( IO_YYY )
         RETURN
#endif
#ifdef GRIB1
      CASE ( IO_GRIB1 )
         RETURN
#endif
#ifdef GRIB2
      CASE ( IO_GRIB2 )
         RETURN
#endif
      CASE (0)
         RETURN
      CASE DEFAULT
         WRITE(mess,*)' An output format has been specified that is incompatible with quilting: io_form: ',io_form,' ',TRIM(stream)
         CALL wrf_error_fatal(mess)
    END SELECT
    END SUBROUTINE sokay

    SUBROUTINE quilt
!<DESCRIPTION>
! I/O server tasks call this routine and remain in it for the rest of the 
! model run.  I/O servers receive I/O requests from compute tasks and 
! perform requested I/O operations by calling package-dependent WRF-specific 
! I/O interfaces.  Requests are sent in the form of "data headers".  Each 
! request has a unique "header" message associated with it.  For requests that 
! contain large amounts of data, the data is appended to the header.  See 
! file module_internal_header_util.F for detailed descriptions of all 
! headers.  
!
! We wish to be able to link to different packages depending on whether
! the I/O is restart, initial, history, or boundary.
!</DESCRIPTION>
      USE module_state_description
      USE module_quilt_outbuf_ops
      IMPLICIT NONE
      INCLUDE 'mpif.h'
#include "intio_tags.h"
#include "wrf_io_flags.h"
      INTEGER itag, ninbuf, ntasks_io_group, ntasks_local_group, mytask_local, ierr
      INTEGER istat
      INTEGER mytask_io_group
      INTEGER   :: nout_set = 0
      INTEGER   :: obufsize, bigbufsize, chunksize, sz
      REAL, DIMENSION(1)      :: dummy
      INTEGER, ALLOCATABLE, DIMENSION(:) :: obuf, bigbuf
      REAL,    ALLOCATABLE, DIMENSION(:) :: RDATA
      INTEGER, ALLOCATABLE, DIMENSION(:) :: IDATA
      CHARACTER (LEN=512) :: CDATA
      CHARACTER (LEN=80) :: fname
      INTEGER icurs, hdrbufsize, itypesize, ftypesize, rtypesize, Status, fstat, io_form_arg
      INTEGER :: DataHandle, FieldType, Comm, IOComm, DomainDesc, code, Count
      INTEGER, DIMENSION(3) :: DomainStart , DomainEnd , MemoryStart , MemoryEnd , PatchStart , PatchEnd
      INTEGER :: dummybuf(1)
      INTEGER :: num_noops, num_commit_messages, num_field_training_msgs, hdr_tag
      CHARACTER (len=256) :: DateStr , Element, VarName, MemoryOrder , Stagger , DimNames(3), FileName, SysDepInfo, mess
      INTEGER, EXTERNAL :: use_package
      LOGICAL           :: stored_write_record, retval
      INTEGER iii, jjj, vid, CC, DD
      LOGICAL           :: call_server_ready

logical okay_to_w
character*120 sysline

! If we've been built with PNETCDF_QUILT defined then we use parallel I/O
! within the group of I/O servers rather than gathering the data onto the
! root I/O server. Unfortunately, this approach means that we can no-longer
! select different I/O layers for use with quilting at run time. ARPDBG.
! This code is sufficiently different that it is kept in the separate 
! quilt_pnc() routine.
#ifdef PNETCDF_QUILT
      CALL quilt_pnc()
      RETURN
#endif

! Call ext_pkg_ioinit() routines to initialize I/O packages.  
      SysDepInfo = " "
#ifdef NETCDF
      CALL ext_ncd_ioinit( SysDepInfo, ierr)
#endif
#ifdef INTIO
      CALL ext_int_ioinit( SysDepInfo, ierr )
#endif
#ifdef XXX
      CALL ext_xxx_ioinit( SysDepInfo, ierr)
#endif
#ifdef YYY
      CALL ext_yyy_ioinit( SysDepInfo, ierr)
#endif
#ifdef ZZZ
      CALL ext_zzz_ioinit( SysDepInfo, ierr)
#endif
#ifdef GRIB1
      CALL ext_gr1_ioinit( SysDepInfo, ierr)
#endif
#ifdef GRIB2
      CALL ext_gr2_ioinit( SysDepInfo, ierr)
#endif

      call_server_ready = .true. ! = true when the server is ready for a new file

      okay_to_commit = .false.
      stored_write_record = .false.
      ninbuf = 0
      ! get info. about the I/O server group that this I/O server task
      ! belongs to
      ! Last task in this I/O server group is the I/O server "root"
      ! The I/O server "root" actually writes data to disk
      ! TBH:  WARNING:  This is also implicit in the call to collect_on_comm().
      CALL mpi_x_comm_size( mpi_comm_io_groups(1), ntasks_io_group,    ierr )
      CALL MPI_COMM_RANK( mpi_comm_io_groups(1), mytask_io_group,    ierr )
      CALL mpi_x_comm_size( mpi_comm_local,        ntasks_local_group, ierr )
      CALL MPI_COMM_RANK( mpi_comm_local,        mytask_local,       ierr )

      CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )
      IF ( itypesize <= 0 ) THEN
        CALL wrf_error_fatal("external/RSL/module_dm.F: quilt: type size <= 0 invalid")
      ENDIF

! Work out whether this i/o server processor has one fewer associated compute proc than
! the most any processor has. Can happen when number of i/o tasks does not evenly divide
! the number of compute tasks. This is needed to keep the i/o tasks sychronized on the
! same message when they start commmunicating to stitch together an output.
!
! Compute processes associated with this task:
       CC = ntasks_io_group - 1
! Number of compute tasks per I/O task (less remainder)
       DD = ncompute_tasks / ntasks_local_group
!
! If CC-DD is 1 on servrs with the maximum number of compute clients, 
!             0 on servrs with one less than maximum


! infinite loop until shutdown message received
! This is the main request-handling loop.  I/O quilt servers stay in this loop 
! until the model run ends.  
okay_to_w = .false.
      DO WHILE (.TRUE.)  ! {

!<DESCRIPTION>
! Each I/O server receives requests from its compute tasks.  Each request
! is contained in a data header (see module_internal_header_util.F for
! detailed descriptions of data headers).
! Each request is sent in two phases.  First, sizes of all messages that 
! will be sent from the compute tasks to this I/O server are summed on the 
! I/O server via MPI_reduce().  The I/O server then allocates buffer "obuf" 
! and receives concatenated messages from the compute tasks in it via the 
! call to collect_on_comm().  Note that "sizes" are generally expressed in 
! *bytes* in this code so conversion to "count" (number of Fortran words) is 
! required for Fortran indexing and MPI calls.  
!</DESCRIPTION>

         if(poll_servers .and. call_server_ready) then
            call_server_ready=.false.
            ! Send a message to the monitor telling it we're ready
            ! for a new data handle.
            call wrf_quilt_server_ready()
         endif

        ! wait for info from compute tasks in the I/O group that we're ready to rock
        ! obufsize will contain number of *bytes*
!CALL start_timing()
        ! first element of reduced is obufsize, second is DataHandle 
        ! if needed (currently needed only for ioclose).
        reduced_dummy = 0
        CALL mpi_x_reduce( reduced_dummy, reduced, 2, MPI_INTEGER, MPI_SUM, mytask_io_group, mpi_comm_io_groups(1), ierr )
        obufsize = reduced(1)
!CALL end_timing("MPI_Reduce at top of forever loop") 
!JMDEBUGwrite(0,*)'obufsize = ',obufsize
! Negative obufsize will trigger I/O server exit.  
        IF ( obufsize .LT. 0 ) THEN
          IF ( obufsize .EQ. -100 ) THEN         ! magic number
#ifdef NETCDF
            CALL ext_ncd_ioexit( Status )
#endif
#ifdef INTIO
            CALL ext_int_ioexit( Status )
#endif
#ifdef XXX
            CALL ext_xxx_ioexit( Status )
#endif
#ifdef YYY
            CALL ext_yyy_ioexit( Status )
#endif
#ifdef ZZZ
            CALL ext_zzz_ioexit( Status )
#endif
#ifdef GRIB1
            CALL ext_gr1_ioexit( Status )
#endif
#ifdef GRIB2
            CALL ext_gr2_ioexit( Status )
#endif
            CALL wrf_message ( 'I/O QUILT SERVERS DONE' )
            CALL mpi_finalize(ierr)
            STOP
          ELSE
            WRITE(mess,*)'Possible 32-bit overflow on output server. Try larger nio_tasks_per_group in namelist.'
            CALL wrf_error_fatal(mess)
          ENDIF
        ENDIF

!        CALL start_timing()
! Obufsize of zero signals a close

! Allocate buffer obuf to be big enough for the data the compute tasks
! will send.  Note: obuf is size in *bytes* so we need to pare this 
! down, since the buffer is INTEGER.  
        IF ( obufsize .GT. 0 ) THEN
          ALLOCATE( obuf( (obufsize+1)/itypesize ) )

! let's roll; get the data from the compute procs and put in obuf
          CALL collect_on_comm_debug(__FILE__,__LINE__, mpi_comm_io_groups(1),        &
                                onebyte,                      &
                                dummy, 0,                     &
                                obuf, obufsize )
!          CALL end_timing( "quilt on server: collecting data from compute procs" )
        ELSE
          ! Necessarily, the compute processes send the ioclose signal,
          ! if there is one, after the iosync, which means they 
          ! will stall on the ioclose message waiting for the quilt 
          ! processes if we handle the way other messages are collected,
          ! using collect_on_comm.  This avoids this, but we need
          ! a special signal (obufsize zero) and the DataHandle
          ! to be closed. That handle is send as the second
          ! word of the io_close message received by the MPI_Reduce above.
          ! Then a header representing the ioclose message is constructed
          ! here and handled below as if it were received from the 
          ! compute processes. The clients (compute processes) must be
          ! careful to send this correctly (one compule process sends the actual
          ! handle and everone else sends a zero, so the result sums to 
          ! the value of the handle).
          !
          ALLOCATE( obuf( 4096 ) )
          ! DataHandle is provided as second element of reduced
          CALL int_gen_handle_header( obuf, obufsize, itypesize, &
                                      reduced(2) , int_ioclose )

          if(poll_servers) then 
             ! Once we're done closing, we need to tell the master
             ! process that we're ready for more data.
             call_server_ready=.true.
          endif
        ENDIF

!write(0,*)'calling init_store_piece_of_field'
! Now all messages received from the compute clients are stored in 
! obuf.  Scan through obuf and extract headers and field data and store in 
! internal buffers.  The scan is done twice, first to determine sizes of 
! internal buffers required for storage of headers and fields and second to 
! actually store the headers and fields.  This bit of code does not do the 
! "quilting" (assembly of patches into full domains).  For each field, it 
! simply concatenates all received patches for the field into a separate 
! internal buffer (i.e. one buffer per field).  Quilting is done later by 
! routine store_patch_in_outbuf().  
        CALL init_store_piece_of_field
        CALL mpi_type_size ( MPI_INTEGER , itypesize , ierr )
!write(0,*)'mpi_type_size returns ', itypesize
! Scan obuf the first time to calculate the size of the buffer required for 
! each field.  Calls to add_to_bufsize_for_field() accumulate sizes.  
        vid = 0
        icurs = itypesize
        num_noops = 0 
        num_commit_messages = 0 
        num_field_training_msgs = 0 
        DO WHILE ( icurs .lt. obufsize ) ! {
          hdr_tag = get_hdr_tag( obuf ( icurs / itypesize ) )
          SELECT CASE ( hdr_tag )
            CASE ( int_field )
              CALL int_get_write_field_header ( obuf(icurs/itypesize), hdrbufsize, itypesize, ftypesize,  &
                                                DataHandle , DateStr , VarName , Dummy , FieldType , Comm , IOComm, &
                                                DomainDesc , MemoryOrder , Stagger , DimNames ,              &
                                                DomainStart , DomainEnd ,                                    &
                                                MemoryStart , MemoryEnd ,                                    &
                                                PatchStart , PatchEnd )
              chunksize = (PatchEnd(1)-PatchStart(1)+1)*(PatchEnd(2)-PatchStart(2)+1)* &
                          (PatchEnd(3)-PatchStart(3)+1)*ftypesize

              IF ( DomainDesc .EQ. 333933 ) THEN  ! Training write, only one per group of tasks
                 IF ( num_field_training_msgs .EQ. 0 ) THEN
                   call add_to_bufsize_for_field( VarName, hdrbufsize )
!write(0,*) 'X-1', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
                 ENDIF
                 num_field_training_msgs = num_field_training_msgs + 1
              ELSE
                 call add_to_bufsize_for_field( VarName, hdrbufsize )
!write(0,*) 'X-2a', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
              ENDIF
              icurs = icurs + hdrbufsize

!write(0,*) 'X-1', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)

              ! If this is a real write (i.e. not a training write), accumulate
              ! buffersize for this field.
              IF ( DomainDesc .NE. 333933 ) THEN   ! magic number
!write(0,*) 'X-1a', chunksize, TRIM(VarName)
                call add_to_bufsize_for_field( VarName, chunksize )
                icurs = icurs + chunksize
              ENDIF
            CASE ( int_open_for_write_commit )  ! only one per group of tasks
              hdrbufsize = obuf(icurs/itypesize)
              IF (num_commit_messages.EQ.0) THEN
                call add_to_bufsize_for_field( 'COMMIT', hdrbufsize )
              ENDIF
              num_commit_messages = num_commit_messages + 1
              icurs = icurs + hdrbufsize
            CASE DEFAULT
              hdrbufsize = obuf(icurs/itypesize)

! This logic and the logic in the loop below is used to determine whether
! to send a noop records sent by the compute processes to allow to go
! through. The purpose is to make sure that the communications between this
! server and the other servers in this quilt group stay synchronized in
! the collection loop below, even when the servers are serving different
! numbers of clients. Here are some conditions:
! 
!   1. The number of compute clients served will not differ by more than 1
!   2. The servers with +1 number of compute clients begin with task 0
!      of mpi_comm_local, the commicator shared by this group of servers
! 
!   3. For each collective field or metadata output from the compute tasks,
!      there will be one record sent to the associated i/o server task. The
!      i/o server task collects these records and stores them contiguously
!      in a buffer (obuf) using collect_on_comm above.  Thus, obuf on this
!      server task will contain one record from each associated compute
!      task, in order.
! 
!   4. In the case of replicated output from the compute tasks
!      (e.g. put_dom_ti records and control records like
!      open_for_write_commit type records), compute task 0 is the only
!      one that sends the record. The other compute tasks send noop
!      records. Thus, obuf on server task zero will contain the output
!      record from task 0 followed by noop records from the rest of the
!      compute tasks associated with task 0.  Obuf on the other server
!      tasks will contain nothing but noop records.
! 
!   5. The logic below will not allow any noop records from server task 0.
!      It allows only one noop record from each of the other server tasks
!      in the i/o group.  This way, for replicated output, when the records
!      are collected on one server task below, using collect_on_comm on
!      mpi_comm_local, each task will provide exactly one record for each
!      call to collect_on_comm: 1 bona fide output record from server task
!      0 and noops from the rest.

              IF ((hdr_tag.EQ.int_noop.AND.mytask_local.NE.0.AND.num_noops.LE.0)  &
                  .OR.hdr_tag.NE.int_noop) THEN
                write(VarName,'(I5.5)')vid 
!write(0,*) 'X-2', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
                call add_to_bufsize_for_field( VarName, hdrbufsize )
                vid = vid+1
              ENDIF
              IF ( hdr_tag .EQ. int_noop ) num_noops = num_noops + 1
              icurs = icurs + hdrbufsize
          END SELECT
        ENDDO ! }
! Store the headers and field data in internal buffers.  The first call to 
! store_piece_of_field() allocates internal buffers using sizes computed by 
! calls to add_to_bufsize_for_field().  
        vid = 0
        icurs = itypesize
        num_noops = 0 
        num_commit_messages = 0 
        num_field_training_msgs = 0 
        DO WHILE ( icurs .lt. obufsize ) !{
!write(0,*) 'A icurs ', icurs, ' obufsize ', obufsize
          hdr_tag = get_hdr_tag( obuf ( icurs / itypesize ) )
          SELECT CASE ( hdr_tag )
            CASE ( int_field )
              CALL int_get_write_field_header ( obuf(icurs/itypesize), hdrbufsize, itypesize, ftypesize,  &
                                                DataHandle , DateStr , VarName , Dummy , FieldType , Comm , IOComm, &
                                                DomainDesc , MemoryOrder , Stagger , DimNames ,              &
                                                DomainStart , DomainEnd ,                                    &
                                                MemoryStart , MemoryEnd ,                                    &
                                                PatchStart , PatchEnd )
              chunksize = (PatchEnd(1)-PatchStart(1)+1)*(PatchEnd(2)-PatchStart(2)+1)* &
                          (PatchEnd(3)-PatchStart(3)+1)*ftypesize

              IF ( DomainDesc .EQ. 333933 ) THEN  ! Training write, only one per group of tasks
                 IF ( num_field_training_msgs .EQ. 0 ) THEN
                   call store_piece_of_field( obuf(icurs/itypesize), VarName, hdrbufsize )
!write(0,*) 'A-1', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
                 ENDIF
                 num_field_training_msgs = num_field_training_msgs + 1
              ELSE
                 call store_piece_of_field( obuf(icurs/itypesize), VarName, hdrbufsize )
!write(0,*) 'A-2a', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
              ENDIF
              icurs = icurs + hdrbufsize
              ! If this is a real write (i.e. not a training write), store
              ! this piece of this field.
              IF ( DomainDesc .NE. 333933 ) THEN   ! magic number
!write(0,*) 'A-1a', chunksize, TRIM(VarName),PatchStart(1:3),PatchEnd(1:3)
                call store_piece_of_field( obuf(icurs/itypesize), VarName, chunksize )
                icurs = icurs + chunksize
              ENDIF
            CASE ( int_open_for_write_commit )  ! only one per group of tasks
              hdrbufsize = obuf(icurs/itypesize)
              IF (num_commit_messages.EQ.0) THEN
                call store_piece_of_field( obuf(icurs/itypesize), 'COMMIT', hdrbufsize )
              ENDIF
              num_commit_messages = num_commit_messages + 1
              icurs = icurs + hdrbufsize
            CASE DEFAULT
              hdrbufsize = obuf(icurs/itypesize)
              IF ((hdr_tag.EQ.int_noop.AND.mytask_local.NE.0.AND.num_noops.LE.0)  &
                  .OR.hdr_tag.NE.int_noop) THEN
                write(VarName,'(I5.5)')vid 
!write(0,*) 'A-2b', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
                call store_piece_of_field( obuf(icurs/itypesize), VarName, hdrbufsize )
                vid = vid+1
              ENDIF
              IF ( hdr_tag .EQ. int_noop ) num_noops = num_noops + 1
              icurs = icurs + hdrbufsize
          END SELECT
        ENDDO !}

! Now, for each field, retrieve headers and patches (data) from the internal 
! buffers and collect them all on the I/O quilt server "root" task.
        CALL init_retrieve_pieces_of_field
! Retrieve header and all patches for the first field from the internal 
! buffers.  
        CALL retrieve_pieces_of_field ( obuf , VarName, obufsize, sz, retval )
! Sum sizes of all headers and patches (data) for this field from all I/O 
! servers in this I/O server group onto the I/O server "root".
        CALL mpi_x_reduce( sz, bigbufsize, 1, MPI_INTEGER, MPI_SUM, ntasks_local_group-1, mpi_comm_local, ierr )
!write(0,*)'seed: sz ',sz,' bigbufsize ',bigbufsize,' VarName ', TRIM(VarName),' retval ',retval

! Loop until there are no more fields to retrieve from the internal buffers.
        DO WHILE ( retval ) !{
#if 0
#else

! I/O server "root" allocates space to collect headers and fields from all
! other servers in this I/O server group.
          IF ( mytask_local .EQ. ntasks_local_group-1 ) THEN
            ALLOCATE( bigbuf( (bigbufsize+1)/itypesize ) )
          ENDIF

! Collect buffers and fields from all I/O servers in this I/O server group
! onto the I/O server "root"
          CALL collect_on_comm_debug2(__FILE__,__LINE__,Trim(VarName),        &
                                get_hdr_tag(obuf),sz,get_hdr_rec_size(obuf),  &
                                mpi_comm_local,                               &
                                onebyte,                                      &
                                obuf, sz,                                     &
                                bigbuf, bigbufsize )
! The I/O server "root" now handles collected requests from all compute 
! tasks served by this I/O server group (i.e. all compute tasks).  
          IF ( mytask_local .EQ. ntasks_local_group-1 ) THEN
!jjj = 4
!do iii = 1, ntasks_local_group
!  write(0,*)'i,j,tag,size ', iii, jjj, get_hdr_tag(bigbuf(jjj/4)),get_hdr_rec_size(bigbuf(jjj/4))
!  jjj = jjj + get_hdr_rec_size(bigbuf(jjj/4))
!enddo

            icurs = itypesize  ! icurs is a byte counter, but buffer is integer

            stored_write_record = .false.

! The I/O server "root" loops over the collected requests.  
            DO WHILE ( icurs .lt. bigbufsize ) !{
              CALL mpi_type_size ( MPI_INTEGER , itypesize , ierr )

!write(0,*)'B tag,size ',icurs,get_hdr_tag( bigbuf(icurs/itypesize) ),get_hdr_rec_size( bigbuf(icurs/itypesize) )
! The I/O server "root" gets the request out of the next header and
! handles it by, in most cases, calling the appropriate external I/O package
! interface.
              SELECT CASE ( get_hdr_tag( bigbuf(icurs/itypesize) ) )
! The I/O server "root" handles the "noop" (do nothing) request.  This is 
! actually quite easy.  "Noop" requests exist to help avoid race conditions.  
! In some cases, only one compute task will everything about a request so 
! other compute tasks send "noop" requests.  
                CASE ( int_noop )
                  CALL int_get_noop_header( bigbuf(icurs/itypesize), hdrbufsize, itypesize )
                  icurs = icurs + hdrbufsize

! The I/O server "root" handles the "put_dom_td_real" request.
                CASE ( int_dom_td_real )
                  CALL mpi_type_size( MPI_REAL, ftypesize, ierr )
                  ALLOCATE( RData( bigbuf(icurs/itypesize + 4 ) ) )      ! 5 is the count of data items for this record ; defined in collect_on_comm.c
                  CALL int_get_td_header( bigbuf(icurs/itypesize:), hdrbufsize, itypesize, ftypesize, &
                                          DataHandle, DateStr, Element, RData, Count, code )
                  icurs = icurs + hdrbufsize

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
                     CASE DEFAULT
                      Status = 0
                  END SELECT

                  DEALLOCATE( RData )
! The I/O server "root" handles the "put_dom_ti_real" request.
                CASE ( int_dom_ti_real )
!write(0,*)' int_dom_ti_real '
                  CALL mpi_type_size( MPI_REAL, ftypesize, ierr )
                  ALLOCATE( RData( bigbuf(icurs/itypesize + 4 ) ) )      ! 5 is the count of data items for this record ; defined in collect_on_comm.c
                  CALL int_get_ti_header( bigbuf(icurs/itypesize:), hdrbufsize, itypesize, ftypesize, &
                                          DataHandle, Element, RData, Count, code )
                  icurs = icurs + hdrbufsize

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
!write(0,*)'ext_ncd_put_dom_ti_real ',handle(DataHandle),TRIM(Element),RData,Status
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  DEALLOCATE( RData )

! The I/O server "root" handles the "put_dom_td_integer" request.
                CASE ( int_dom_td_integer )
!write(0,*)' int_dom_td_integer '
                  CALL mpi_type_size( MPI_INTEGER, ftypesize, ierr )
                  ALLOCATE( IData( bigbuf(icurs/itypesize + 4 ) ) )      ! 5 is the count of data items for this record ; defined in collect_on_comm.c
                  CALL int_get_td_header( bigbuf(icurs/itypesize:), hdrbufsize, itypesize, ftypesize, &
                                          DataHandle, DateStr, Element, IData, Count, code )
                  icurs = icurs + hdrbufsize

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  DEALLOCATE( IData )

! The I/O server "root" handles the "put_dom_ti_integer" request.
                CASE ( int_dom_ti_integer )
!write(0,*)' int_dom_ti_integer '

                  CALL mpi_type_size( MPI_INTEGER, ftypesize, ierr )
                  ALLOCATE( IData( bigbuf(icurs/itypesize + 4 ) ) )      ! 5 is the count of data items for this record ; defined in collect_on_comm.c
                  CALL int_get_ti_header( bigbuf(icurs/itypesize:), hdrbufsize, itypesize, ftypesize, &
                                          DataHandle, Element, IData, Count, code )
                  icurs = icurs + hdrbufsize
                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
!write(0,*)'ext_ncd_put_dom_ti_integer ',handle(DataHandle),TRIM(Element),IData,Status
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif

                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  DEALLOCATE( IData)
 
! The I/O server "root" handles the "set_time" request.
                CASE ( int_set_time )
!write(0,*)' int_set_time '
                  CALL int_get_ti_header_char( bigbuf(icurs/itypesize), hdrbufsize, itypesize, &
                                               DataHandle, Element, VarName, CData, code )
                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_set_time ( handle(DataHandle), TRIM(CData), Status)
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  icurs = icurs + hdrbufsize

! The I/O server "root" handles the "put_dom_ti_char" request.
                CASE ( int_dom_ti_char )
!write(0,*)' before int_get_ti_header_char '
                  CALL int_get_ti_header_char( bigbuf(icurs/itypesize), hdrbufsize, itypesize, &
                                               DataHandle, Element, VarName, CData, code )
!write(0,*)' after int_get_ti_header_char ',VarName

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  icurs = icurs + hdrbufsize

! The I/O server "root" handles the "put_var_ti_char" request.
                CASE ( int_var_ti_char )
!write(0,*)' int_var_ti_char '
                  CALL int_get_ti_header_char( bigbuf(icurs/itypesize), hdrbufsize, itypesize, &
                                               DataHandle, Element, VarName, CData, code )

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  icurs = icurs + hdrbufsize

                CASE ( int_ioexit )
! ioexit is now handled by sending negative message length to server
                  CALL wrf_error_fatal( &
                         "quilt: should have handled int_ioexit already")
! The I/O server "root" handles the "ioclose" request.
                CASE ( int_ioclose )
                  CALL int_get_handle_header( bigbuf(icurs/itypesize), hdrbufsize, itypesize, &
                                              DataHandle , code )
                  icurs = icurs + hdrbufsize

                  IF ( DataHandle .GE. 1 ) THEN
!JMDEBUGwrite(0,*)'closing DataHandle ',DataHandle,' io_form ',io_form(DataHandle)

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_inquire_filename( handle(DataHandle), fname, fstat, Status )
                      IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                        CALL ext_ncd_ioclose(handle(DataHandle),Status)
                      ENDIF
#endif
#ifdef PNETCDF
                    CASE ( IO_PNETCDF   )
                      CALL ext_pnc_inquire_filename( handle(DataHandle), fname, fstat, Status )
                      IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                        CALL ext_pnc_ioclose(handle(DataHandle),Status)
                      ENDIF
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_inquire_filename( handle(DataHandle), fname, fstat, Status )
                      IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                        CALL ext_int_ioclose(handle(DataHandle),Status)
                      ENDIF
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                      CALL ext_yyy_ioclose(handle(DataHandle),Status)
                    ENDIF
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                      CALL ext_gr1_ioclose(handle(DataHandle),Status)
                    ENDIF
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                      CALL ext_gr2_ioclose(handle(DataHandle),Status)
                    ENDIF
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT
                  ENDIF

! The I/O server "root" handles the "open_for_write_begin" request.
                CASE ( int_open_for_write_begin )

                  CALL int_get_ofwb_header( bigbuf(icurs/itypesize), hdrbufsize, itypesize, &
                                            FileName,SysDepInfo,io_form_arg,DataHandle )

!write(0,*)' int_open_for_write_begin itypesize ',itypesize,' itypesize ',itypesize
!write(0,*)' int_open_for_write_begin icurs ', icurs, hdrbufsize
!JMDEBUGwrite(0,*)' int_open_for_write_begin FileName ',TRIM(FileName) , ' DataHandle ', DataHandle
!write(0,*)' int_open_for_write_begin SysDepInfo ',TRIM(SysDepInfo) 
                  icurs = icurs + hdrbufsize
!write(0,*)' int_open_for_write_begin new icurs,tag,size ', icurs, get_hdr_tag( bigbuf(icurs/itypesize) ),get_hdr_rec_size( bigbuf(icurs/itypesize) )
                
                  io_form(DataHandle) = io_form_arg

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
!write(0,*)'ext_ncd_open_for_write_begin ',Trim(FileName),DataHandle,handle(DataHandle),Status
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
#endif
#ifdef YYY
                    CASE ( IO_YYY )
                       CALL ext_yyy_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
#endif
#ifdef GRIB1
                    CASE ( IO_GRIB1 )
                       CALL ext_gr1_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
#endif
#ifdef GRIB2
                    CASE ( IO_GRIB2 )
                       CALL ext_gr2_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT
                
                  okay_to_write(DataHandle) = .false.

! The I/O server "root" handles the "open_for_write_commit" request.
! In this case, the "okay_to_commit" is simply set to .true. so "write_field"
! requests will initiate writes to disk.  Actual commit will be done after
! all requests in this batch have been handled.
                CASE ( int_open_for_write_commit )

                  CALL int_get_handle_header( bigbuf(icurs/itypesize), hdrbufsize, itypesize, &
                                              DataHandle , code )
                  icurs = icurs + hdrbufsize
                  okay_to_commit(DataHandle) = .true.

! The I/O server "root" handles the "write_field" (int_field) request.
! If okay_to_write(DataHandle) is .true. then the patch in the
! header (bigbuf) is written to a globally-sized internal output buffer via
! the call to store_patch_in_outbuf().  Note that this is where the actual
! "quilting" (reassembly of patches onto a full-size domain) is done.  If
! okay_to_write(DataHandle) is .false. then external I/O package interfaces
! are called to write metadata for I/O formats that support native metadata.
!
! NOTE that the I/O server "root" will only see write_field (int_field)
! requests AFTER an "iosync" request.
                CASE ( int_field )
                  CALL mpi_type_size( MPI_INTEGER, ftypesize, ierr )
                  CALL int_get_write_field_header ( bigbuf(icurs/itypesize), hdrbufsize, itypesize, ftypesize,  &
                                                    DataHandle , DateStr , VarName , Dummy , FieldType , Comm , IOComm, &
                                                    DomainDesc , MemoryOrder , Stagger , DimNames ,              &
                                                    DomainStart , DomainEnd ,                                    &
                                                    MemoryStart , MemoryEnd ,                                    &
                                                    PatchStart , PatchEnd )
!write(0,*)' int_field ',TRIM(VarName),DataHandle,okay_to_write(DataHandle)
                  icurs = icurs + hdrbufsize

                  IF ( okay_to_write(DataHandle) ) THEN

!                    WRITE(0,*)'>>> ',TRIM(DateStr), ' ', TRIM(VarName), ' ', TRIM(MemoryOrder), ' ', &
!                        (PatchEnd(1)-PatchStart(1)+1)*(PatchEnd(2)-PatchStart(2)+1)*(PatchEnd(3)-PatchStart(3)+1)

                    IF ( FieldType .EQ. WRF_FLOAT .OR. FieldType .EQ. WRF_DOUBLE)  THEN
                      ! Note that the WRF_DOUBLE branch of this IF statement must come first since 
                      ! WRF_FLOAT is set equal to WRF_DOUBLE during autopromotion builds.  
                      IF ( FieldType .EQ. WRF_DOUBLE)  THEN
! this branch has not been tested TBH: 20050406
                        CALL mpi_type_size( MPI_DOUBLE_PRECISION, ftypesize, ierr )
                      ELSE
                        CALL mpi_type_size( MPI_REAL, ftypesize, ierr )
                      ENDIF
                      stored_write_record = .true.
                      CALL store_patch_in_outbuf ( bigbuf(icurs/itypesize), dummybuf, TRIM(DateStr), TRIM(VarName) , &
                                                   FieldType, TRIM(MemoryOrder), TRIM(Stagger), DimNames, &
                                                   DomainStart , DomainEnd , &
                                                   MemoryStart , MemoryEnd , &
                                                   PatchStart , PatchEnd )

                    ELSE IF ( FieldType .EQ. WRF_INTEGER ) THEN
                      CALL mpi_type_size( MPI_INTEGER, ftypesize, ierr )
                      stored_write_record = .true.
                      CALL store_patch_in_outbuf ( dummybuf, bigbuf(icurs/itypesize), TRIM(DateStr), TRIM(VarName) , &
                                                   FieldType, TRIM(MemoryOrder), TRIM(Stagger), DimNames, &
                                                   DomainStart , DomainEnd , &
                                                   MemoryStart , MemoryEnd , &
                                                   PatchStart , PatchEnd )
                    ELSE IF ( FieldType .EQ. WRF_LOGICAL ) THEN
                      ftypesize = LWORDSIZE
                    ENDIF
                    icurs = icurs + (PatchEnd(1)-PatchStart(1)+1)*(PatchEnd(2)-PatchStart(2)+1)* &
                                    (PatchEnd(3)-PatchStart(3)+1)*ftypesize
                  ELSE
                    SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                      CASE ( IO_NETCDF   )
                        CALL ext_ncd_write_field ( handle(DataHandle) , TRIM(DateStr) ,         &
                                   TRIM(VarName) , dummy , FieldType , Comm , IOComm,           &
                                   DomainDesc , TRIM(MemoryOrder) , TRIM(Stagger) , DimNames ,  &
                                   DomainStart , DomainEnd ,                                    &
                                   DomainStart , DomainEnd ,                                    &
                                   DomainStart , DomainEnd ,                                    &
                                   Status )
#endif
#if 0
! since this is training and the grib output doesn't need training, disable this branch.
#ifdef YYY
                 CASE ( IO_YYY )
                      CALL ext_YYY_write_field ( handle(DataHandle) , TRIM(DateStr) ,         &
                                 TRIM(VarName) , dummy , FieldType , Comm , IOComm,           &
                                 DomainDesc , TRIM(MemoryOrder) , TRIM(Stagger) , DimNames ,  &
                                 DomainStart , DomainEnd ,                                    &
                                 DomainStart , DomainEnd ,                                    &
                                 DomainStart , DomainEnd ,                                    &
                                 Status )
#endif
#endif
                      CASE DEFAULT
                        Status = 0
                    END SELECT
                  ENDIF
                CASE ( int_iosync )
                  CALL int_get_handle_header( bigbuf(icurs/itypesize), hdrbufsize, itypesize, &
                                            DataHandle , code )
                  icurs = icurs + hdrbufsize
                CASE DEFAULT
                  WRITE(mess,*)'quilt: bad tag: ',get_hdr_tag( bigbuf(icurs/itypesize) ),' icurs ',icurs/itypesize
                  CALL wrf_error_fatal( mess )
              END SELECT

            ENDDO !}
! Now, the I/O server "root" has finshed handling all commands from the latest
! call to retrieve_pieces_of_field().

            IF (stored_write_record) THEN
! If any fields have been stored in a globally-sized internal output buffer
! (via a call to store_patch_in_outbuf()) then call write_outbuf() to write
! them to disk now.
! NOTE that the I/O server "root" will only have called
! store_patch_in_outbuf() when handling write_field (int_field)
! commands which only arrive AFTER an "iosync" command.
!              CALL start_timing
              CALL write_outbuf ( handle(DataHandle), use_package(io_form(DataHandle))) 
!              CALL end_timing( "quilt: call to write_outbuf" ) 
            ENDIF

! If one or more "open_for_write_commit" commands were encountered from the
! latest call to retrieve_pieces_of_field() then call the package-specific
! routine to do the commit.
            IF (okay_to_commit(DataHandle)) THEN

              SELECT CASE (use_package(io_form(DataHandle)))
#ifdef NETCDF
                CASE ( IO_NETCDF   )
                  CALL ext_ncd_inquire_filename( handle(DataHandle), fname, fstat, Status )
                  IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                    CALL ext_ncd_open_for_write_commit(handle(DataHandle),Status)
                    okay_to_write(DataHandle) = .true.
                  ENDIF
#endif
#ifdef INTIO
                CASE ( IO_INTIO   )
                  CALL ext_int_inquire_filename( handle(DataHandle), fname, fstat, Status )
                  IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                    CALL ext_int_open_for_write_commit(handle(DataHandle),Status)
                    okay_to_write(DataHandle) = .true.
                  ENDIF
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                       CALL ext_yyy_open_for_write_commit(handle(DataHandle),Status)
                       okay_to_write(DataHandle) = .true.
                    ENDIF
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                       CALL ext_gr1_open_for_write_commit(handle(DataHandle),Status)
                       okay_to_write(DataHandle) = .true.
                    ENDIF
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                       CALL ext_gr2_open_for_write_commit(handle(DataHandle),Status)
                       okay_to_write(DataHandle) = .true.
                    ENDIF
#endif

                CASE DEFAULT
                  Status = 0
              END SELECT

            okay_to_commit(DataHandle) = .false.
          ENDIF
          DEALLOCATE( bigbuf )
        ENDIF
#endif

! Retrieve header and all patches for the next field from the internal 
! buffers.  
        CALL retrieve_pieces_of_field ( obuf , VarName, obufsize, sz, retval )
! Sum sizes of all headers and patches (data) for this field from all I/O 
! servers in this I/O server group onto the I/O server "root".
        CALL mpi_x_reduce( sz, bigbufsize, 1, MPI_INTEGER,MPI_SUM, ntasks_local_group-1,mpi_comm_local, ierr )
! Then, return to the top of the loop to collect headers and data from all 
! I/O servers in this I/O server group onto the I/O server "root" and handle 
! the next batch of commands.  
      END DO !}

      DEALLOCATE( obuf )

      ! flush output files if needed
      IF (stored_write_record) THEN
!         CALL start_timing()
        SELECT CASE ( use_package(io_form) )
#ifdef NETCDF
          CASE ( IO_NETCDF   )
            CALL ext_ncd_iosync( handle(DataHandle), Status )
#endif
#ifdef XXX
          CASE ( IO_XXX   )
            CALL ext_xxx_iosync( handle(DataHandle), Status )
#endif
#ifdef YYY
          CASE ( IO_YYY   )
            CALL ext_yyy_iosync( handle(DataHandle), Status )
#endif
#ifdef ZZZ
          CASE ( IO_ZZZ   )
            CALL ext_zzz_iosync( handle(DataHandle), Status )
#endif
#ifdef GRIB1
          CASE ( IO_GRIB1   )
            CALL ext_gr1_iosync( handle(DataHandle), Status )
#endif
#ifdef GRIB2
          CASE ( IO_GRIB2   )
            CALL ext_gr2_iosync( handle(DataHandle), Status )
#endif
#ifdef INTIO
          CASE ( IO_INTIO   )
            CALL ext_int_iosync( handle(DataHandle), Status )
#endif
          CASE DEFAULT
            Status = 0
        END SELECT
!CALL end_timing( "quilt: flush" )
      ENDIF

      END DO ! }

    END SUBROUTINE quilt

    SUBROUTINE quilt_pnc
!<DESCRIPTION>
! Same as quilt() routine except that _all_ of the IO servers that call it
! actually write data to disk using pNetCDF. This version is only used when 
! the  code is compiled with PNETCDF_QUILT defined.
!</DESCRIPTION>
      USE module_state_description
      USE module_quilt_outbuf_ops
      IMPLICIT NONE
      INCLUDE 'mpif.h'
#include "intio_tags.h"
#include "wrf_io_flags.h"
      INTEGER itag, ninbuf, ntasks_io_group, ntasks_local_group, mytask_local, ierr
      INTEGER istat
      INTEGER mytask_io_group
      INTEGER   :: nout_set = 0
      INTEGER   :: obufsize, bigbufsize, chunksize, sz
      REAL,                 DIMENSION(1) :: dummy
      INTEGER, ALLOCATABLE, DIMENSION(:) :: obuf, bigbuf
      REAL,    ALLOCATABLE, DIMENSION(:) :: RDATA
      INTEGER, ALLOCATABLE, DIMENSION(:) :: IDATA
      CHARACTER (LEN=512) :: CDATA
      CHARACTER (LEN=80) :: fname
      INTEGER icurs, hdrbufsize, itypesize, ftypesize, rtypesize, Status, fstat, io_form_arg
      INTEGER :: DataHandle, FieldType, Comm, IOComm, DomainDesc, code, Count
      INTEGER, DIMENSION(3) :: DomainStart , DomainEnd , MemoryStart , MemoryEnd , PatchStart , PatchEnd
      INTEGER :: dummybuf(1)
      INTEGER :: num_noops, num_commit_messages, num_field_training_msgs, hdr_tag
      CHARACTER (len=256) :: DateStr , Element, VarName, MemoryOrder , Stagger , DimNames(3), FileName, SysDepInfo, mess
      INTEGER, EXTERNAL :: use_package
      LOGICAL           :: stored_write_record, retval, written_record
      INTEGER iii, jjj, vid, CC, DD

!      logical okay_to_w
!      character*120 sysline

! Call ext_pkg_ioinit() routines to initialize I/O packages.  
      SysDepInfo = " "
#ifdef NETCDF
      CALL ext_ncd_ioinit( SysDepInfo, ierr)
#endif
#ifdef PNETCDF_QUILT
      CALL ext_pnc_ioinit( SysDepInfo, ierr)
#endif
#ifdef INTIO
      CALL ext_int_ioinit( SysDepInfo, ierr )
#endif
#ifdef XXX
      CALL ext_xxx_ioinit( SysDepInfo, ierr)
#endif
#ifdef YYY
      CALL ext_yyy_ioinit( SysDepInfo, ierr)
#endif
#ifdef ZZZ
      CALL ext_zzz_ioinit( SysDepInfo, ierr)
#endif
#ifdef GRIB1
      CALL ext_gr1_ioinit( SysDepInfo, ierr)
#endif
#ifdef GRIB2
      CALL ext_gr2_ioinit( SysDepInfo, ierr)
#endif

      okay_to_commit = .false.
      stored_write_record = .false.
      ninbuf = 0
      ! get info. about the I/O server group that this I/O server task
      ! belongs to
      CALL mpi_x_comm_size( mpi_comm_io_groups(1), ntasks_io_group,    ierr )
      CALL MPI_COMM_RANK( mpi_comm_io_groups(1), mytask_io_group,    ierr )
      CALL mpi_x_comm_size( mpi_comm_local,        ntasks_local_group, ierr )
      CALL MPI_COMM_RANK( mpi_comm_local,        mytask_local,       ierr )

      CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )
      IF ( itypesize <= 0 ) THEN
        CALL wrf_error_fatal("external/RSL/module_dm.F: quilt: type size <= 0 invalid")
      ENDIF

! Work out whether this i/o server processor has one fewer associated compute proc than
! the most any processor has. Can happen when number of i/o tasks does not evenly divide
! the number of compute tasks. This is needed to keep the i/o tasks sychronized on the
! same message when they start commmunicating to stitch together an output.
!
! Compute processes associated with this task:
       CC = ntasks_io_group - 1
! Number of compute tasks per I/O task (less remainder)
       DD = ncompute_tasks / ntasks_local_group
!
! If CC-DD is 1 on servrs with the maximum number of compute clients, 
!             0 on servrs with one less than maximum


! infinite loop until shutdown message received
! This is the main request-handling loop.  I/O quilt servers stay in this loop 
! until the model run ends.  
!okay_to_w = .false.
      DO WHILE (.TRUE.)  ! {

!<DESCRIPTION>
! Each I/O server receives requests from its compute tasks.  Each request
! is contained in a data header (see module_internal_header_util.F for
! detailed descriptions of data headers).
! Each request is sent in two phases.  First, sizes of all messages that 
! will be sent from the compute tasks to this I/O server are summed on the 
! I/O server via MPI_reduce().  The I/O server then allocates buffer "obuf" 
! and receives concatenated messages from the compute tasks in it via the 
! call to collect_on_comm().  Note that "sizes" are generally expressed in 
! *bytes* in this code so conversion to "count" (number of Fortran words) is 
! required for Fortran indexing and MPI calls.  
!</DESCRIPTION>
        ! wait for info from compute tasks in the I/O group that we're ready to rock
        ! obufsize will contain number of *bytes*
!CALL start_timing
        ! first element of reduced is obufsize, second is DataHandle 
        ! if needed (currently needed only for ioclose).
        reduced_dummy = 0
        CALL mpi_x_reduce( reduced_dummy, reduced, 2, MPI_INTEGER, MPI_SUM, mytask_io_group, mpi_comm_io_groups(1), ierr )
        obufsize = reduced(1)
!CALL end_timing("MPI_Reduce at top of forever loop") 
!JMDEBUGwrite(0,*)'obufsize = ',obufsize
! Negative obufsize will trigger I/O server exit.  
        IF ( obufsize .LT. 0 ) THEN
          IF ( obufsize .EQ. -100 ) THEN         ! magic number
#ifdef NETCDF
            CALL ext_ncd_ioexit( Status )
#endif
#ifdef PNETCDF_QUILT
            CALL ext_pnc_ioexit( Status )
#endif
#ifdef INTIO
            CALL ext_int_ioexit( Status )
#endif
#ifdef XXX
            CALL ext_xxx_ioexit( Status )
#endif
#ifdef YYY
            CALL ext_yyy_ioexit( Status )
#endif
#ifdef ZZZ
            CALL ext_zzz_ioexit( Status )
#endif
#ifdef GRIB1
            CALL ext_gr1_ioexit( Status )
#endif
#ifdef GRIB2
            CALL ext_gr2_ioexit( Status )
#endif
            CALL wrf_message ( 'I/O QUILT SERVERS DONE' )
            CALL mpi_finalize(ierr)
            STOP
          ELSE
            WRITE(mess,*)'Possible 32-bit overflow on output server. Try larger nio_tasks_per_group in namelist.'
            CALL wrf_error_fatal(mess)
          ENDIF
        ENDIF

!        CALL start_timing
! Obufsize of zero signals a close

! Allocate buffer obuf to be big enough for the data the compute tasks
! will send.  Note: obuf is size in *bytes* so we need to pare this 
! down, since the buffer is INTEGER.  
        IF ( obufsize .GT. 0 ) THEN
          ALLOCATE( obuf( (obufsize+1)/itypesize ) )

! let's roll; get the data from the compute procs and put in obuf
          CALL collect_on_comm_debug(__FILE__,__LINE__, mpi_comm_io_groups(1),        &
                                onebyte,                      &
                                dummy, 0,                     &
                                obuf, obufsize )
!          CALL end_timing( "quilt on server: collecting data from compute procs" )
        ELSE
          ! Necessarily, the compute processes send the ioclose signal,
          ! if there is one, after the iosync, which means they 
          ! will stall on the ioclose message waiting for the quilt 
          ! processes if we handle the way other messages are collected,
          ! using collect_on_comm.  This avoids this, but we need
          ! a special signal (obufsize zero) and the DataHandle
          ! to be closed. That handle is send as the second
          ! word of the io_close message received by the MPI_Reduce above.
          ! Then a header representing the ioclose message is constructed
          ! here and handled below as if it were received from the 
          ! compute processes. The clients (compute processes) must be
          ! careful to send this correctly (one compule process sends the actual
          ! handle and everone else sends a zero, so the result sums to 
          ! the value of the handle).
          !
          ALLOCATE( obuf( 4096 ) )
          ! DataHandle is provided as second element of reduced
          CALL int_gen_handle_header( obuf, obufsize, itypesize, &
                                      reduced(2) , int_ioclose )
        ENDIF

!write(0,*)'calling init_store_piece_of_field'
! Now all messages received from the compute clients are stored in 
! obuf.  Scan through obuf and extract headers and field data and store in 
! internal buffers.  The scan is done twice, first to determine sizes of 
! internal buffers required for storage of headers and fields and second to 
! actually store the headers and fields.  This bit of code does not do any 
! "quilting" (assembly of patches into full domains).  For each field, it 
! simply writes all received patches for the field to disk.
! ARPDBG we can vastly reduce the number of writes to disk by stitching
! any contiguous patches together first. Has implications for synchronisation
! of pNetCDF calls though.
        CALL init_store_piece_of_field
        CALL mpi_type_size ( MPI_INTEGER , itypesize , ierr )
!write(0,*)'mpi_type_size returns ', itypesize
! Scan obuf the first time to calculate the size of the buffer required for 
! each field.  Calls to add_to_bufsize_for_field() accumulate sizes.  
        vid = 0
        icurs = itypesize
        num_noops = 0 
        num_commit_messages = 0 
        num_field_training_msgs = 0 
        DO WHILE ( icurs .lt. obufsize ) ! {
          hdr_tag = get_hdr_tag( obuf ( icurs / itypesize ) )
          SELECT CASE ( hdr_tag )
            CASE ( int_field )
              CALL int_get_write_field_header ( obuf(icurs/itypesize), hdrbufsize, itypesize, ftypesize,  &
                                                DataHandle , DateStr , VarName , Dummy , FieldType , Comm , IOComm, &
                                                DomainDesc , MemoryOrder , Stagger , DimNames ,              &
                                                DomainStart , DomainEnd ,                                    &
                                                MemoryStart , MemoryEnd ,                                    &
                                                PatchStart , PatchEnd )
              chunksize = (PatchEnd(1)-PatchStart(1)+1)*(PatchEnd(2)-PatchStart(2)+1)* &
                          (PatchEnd(3)-PatchStart(3)+1)*ftypesize

              IF ( DomainDesc .EQ. 333933 ) THEN  ! Training write, only one per group of tasks
                 IF ( num_field_training_msgs .EQ. 0 ) THEN
                   call add_to_bufsize_for_field( VarName, hdrbufsize )
!write(0,*) 'X-1', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
                 ENDIF
                 num_field_training_msgs = num_field_training_msgs + 1
              ELSE
                 call add_to_bufsize_for_field( VarName, hdrbufsize )
!write(0,*) 'X-2a', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
              ENDIF
              icurs = icurs + hdrbufsize

!write(0,*) 'X-1', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)

              ! If this is a real write (i.e. not a training write), accumulate
              ! buffersize for this field.
              IF ( DomainDesc .NE. 333933 ) THEN   ! magic number
!write(0,*) 'X-1a', chunksize, TRIM(VarName)
                call add_to_bufsize_for_field( VarName, chunksize )
                icurs = icurs + chunksize
              ENDIF
            CASE ( int_open_for_write_commit )  ! only one per group of tasks
              hdrbufsize = obuf(icurs/itypesize)
              IF (num_commit_messages.EQ.0) THEN
                call add_to_bufsize_for_field( 'COMMIT', hdrbufsize )
              ENDIF
              num_commit_messages = num_commit_messages + 1
              icurs = icurs + hdrbufsize
            CASE DEFAULT
              hdrbufsize = obuf(icurs/itypesize)

! This logic and the logic in the loop below is used to determine whether
! to send a noop records sent by the compute processes to allow to go
! through. The purpose is to make sure that the communications between this
! server and the other servers in this quilt group stay synchronized in
! the collection loop below, even when the servers are serving different
! numbers of clients. Here are some conditions:
! 
!   1. The number of compute clients served will not differ by more than 1
!   2. The servers with +1 number of compute clients begin with task 0
!      of mpi_comm_local, the commicator shared by this group of servers
! 
!   3. For each collective field or metadata output from the compute tasks,
!      there will be one record sent to the associated i/o server task. The
!      i/o server task collects these records and stores them contiguously
!      in a buffer (obuf) using collect_on_comm above.  Thus, obuf on this
!      server task will contain one record from each associated compute
!      task, in order.
! ! 
!   4. In the case of replicated output from the compute tasks
!      (e.g. put_dom_ti records and control records like
!      open_for_write_commit type records), only compute tasks for which 
!      (compute_group_master == .TRUE) send the record. The other compute 
!      tasks send noop records. This is done so that each server task 
!      receives exactly one record plus noops from the other compute tasks. 
! 
!   5. Logic below does not allow any noop records through since each IO 
!      server task now receives a valid record (from the 'compute-group master'
!      when doing replicated output
              IF (hdr_tag.NE.int_noop) THEN
                write(VarName,'(I5.5)')vid 
!write(0,*) 'X-2', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
                call add_to_bufsize_for_field( VarName, hdrbufsize )
                vid = vid+1
              ENDIF
              IF ( hdr_tag .EQ. int_noop ) num_noops = num_noops + 1
              icurs = icurs + hdrbufsize

          END SELECT
        ENDDO ! }
! Store the headers and field data in internal buffers.  The first call to 
! store_piece_of_field() allocates internal buffers using sizes computed by 
! calls to add_to_bufsize_for_field().  
        vid = 0
        icurs = itypesize
        num_noops = 0 
        num_commit_messages = 0 
        num_field_training_msgs = 0 
        DO WHILE ( icurs .lt. obufsize ) !{
!write(0,*) 'A icurs ', icurs, ' obufsize ', obufsize
          hdr_tag = get_hdr_tag( obuf ( icurs / itypesize ) )
          SELECT CASE ( hdr_tag )
            CASE ( int_field )
              CALL int_get_write_field_header ( obuf(icurs/itypesize), hdrbufsize, itypesize, ftypesize,  &
                                                DataHandle , DateStr , VarName , Dummy , FieldType , Comm , IOComm, &
                                                DomainDesc , MemoryOrder , Stagger , DimNames ,              &
                                                DomainStart , DomainEnd ,                                    &
                                                MemoryStart , MemoryEnd ,                                    &
                                                PatchStart , PatchEnd )
              chunksize = (PatchEnd(1)-PatchStart(1)+1)*(PatchEnd(2)-PatchStart(2)+1)* &
                          (PatchEnd(3)-PatchStart(3)+1)*ftypesize

              IF ( DomainDesc .EQ. 333933 ) THEN  ! Training write, only one per group of tasks
                 IF ( num_field_training_msgs .EQ. 0 ) THEN
                   call store_piece_of_field( obuf(icurs/itypesize), VarName, hdrbufsize )
!write(0,*) 'A-1', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
                 ENDIF
                 num_field_training_msgs = num_field_training_msgs + 1
              ELSE
                 call store_piece_of_field( obuf(icurs/itypesize), VarName, hdrbufsize )
!write(0,*) 'A-2a', icurs, hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
              ENDIF
              icurs = icurs + hdrbufsize
              ! If this is a real write (i.e. not a training write), store
              ! this piece of this field.
              IF ( DomainDesc .NE. 333933 ) THEN   ! magic number
                call store_piece_of_field( obuf(icurs/itypesize), VarName, chunksize )
                icurs = icurs + chunksize
!write(0,*) 'A-1a',TRIM(VarName),' icurs ',icurs,PatchStart(1:3),PatchEnd(1:3)
              ENDIF
            CASE ( int_open_for_write_commit )  ! only one per group of tasks
              hdrbufsize = obuf(icurs/itypesize)
              IF (num_commit_messages.EQ.0) THEN
                call store_piece_of_field( obuf(icurs/itypesize), 'COMMIT', hdrbufsize )
              ENDIF
              num_commit_messages = num_commit_messages + 1
              icurs = icurs + hdrbufsize
            CASE DEFAULT
              hdrbufsize = obuf(icurs/itypesize)
              IF (hdr_tag.NE.int_noop) THEN

                write(VarName,'(I5.5)')vid 
!write(0,*) 'A-2b', hdrbufsize, get_hdr_tag( obuf ( icurs / itypesize ) ) , get_hdr_rec_size( obuf ( icurs / itypesize ) ), TRIM(VarName)
                call store_piece_of_field( obuf(icurs/itypesize), VarName, hdrbufsize )
                vid = vid+1
              ENDIF
              IF ( hdr_tag .EQ. int_noop ) num_noops = num_noops + 1
              icurs = icurs + hdrbufsize
          END SELECT
       ENDDO !} while(icurs < obufsize)

! Now, for each field, retrieve headers and patches (data) from the internal 
! buffers
       CALL init_retrieve_pieces_of_field
! Retrieve header and all patches for the first field from the internal 
! buffers.  
       CALL retrieve_pieces_of_field ( obuf , VarName, obufsize, sz, retval )
       written_record = .false.

! Loop until there are no more fields to retrieve from the internal buffers.
       DO WHILE ( retval ) !{

! This I/O server now handles the collected requests from the compute 
! tasks it serves

            icurs = itypesize  ! icurs is a byte counter, but buffer is integer

            stored_write_record = .false.

! ALL I/O servers in this group loop over the collected requests they have
! received.
            DO WHILE ( icurs .lt. sz)! bigbufsize ) !{

! The I/O server gets the request out of the next header and
! handles it by, in most cases, calling the appropriate external I/O package
! interface.
!write(0,*)__FILE__,__LINE__,'get_hdr_tag ',icurs,sz,get_hdr_tag( obuf(icurs/itypesize) )
              SELECT CASE ( get_hdr_tag( obuf(icurs/itypesize) ) )
! The I/O server handles the "noop" (do nothing) request.  This is 
! actually quite easy.  "Noop" requests exist to help avoid race conditions.  
                CASE ( int_noop )
                  CALL int_get_noop_header( obuf(icurs/itypesize), &
                                            hdrbufsize, itypesize )
                  icurs = icurs + hdrbufsize

! The I/O server "root" handles the "put_dom_td_real" request.
                CASE ( int_dom_td_real )
                  CALL mpi_type_size( MPI_REAL, ftypesize, ierr )
                  ALLOCATE( RData( obuf(icurs/itypesize + 4 ) ) )      ! 5 is the count of data items for this record ; defined in collect_on_comm.c
                  CALL int_get_td_header( obuf(icurs/itypesize:), hdrbufsize, itypesize, ftypesize, &
                                          DataHandle, DateStr, Element, RData, Count, code )
                  icurs = icurs + hdrbufsize

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                    CASE (IO_PNETCDF  )
                      CALL ext_pnc_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_dom_td_real( handle(DataHandle),TRIM(Element),TRIM(DateStr),RData, Count, Status )
#endif
                     CASE DEFAULT
                      Status = 0
                  END SELECT

                  DEALLOCATE( RData )
! Every I/O server handles the "put_dom_ti_real" request.
                CASE ( int_dom_ti_real )

                  CALL mpi_type_size( MPI_REAL, ftypesize, ierr )
                  ALLOCATE( RData( obuf(icurs/itypesize + 4 ) ) )      ! 5 is the count of data items for this record ; defined in collect_on_comm.c
                  CALL int_get_ti_header( obuf(icurs/itypesize:), hdrbufsize, itypesize, ftypesize, &
                                          DataHandle, Element, RData, Count, code )
                  icurs = icurs + hdrbufsize

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                    CASE (IO_PNETCDF  )
                      CALL ext_pnc_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_dom_ti_real( handle(DataHandle),TRIM(Element), RData, Count, Status )
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  DEALLOCATE( RData )

! Every I/O server handles the "put_dom_td_integer" request.
                CASE ( int_dom_td_integer )

                  CALL mpi_type_size( MPI_INTEGER, ftypesize, ierr )
                  ALLOCATE( IData( obuf(icurs/itypesize + 4 ) ) )      ! 5 is the count of data items for this record ; defined in collect_on_comm.c
                  CALL int_get_td_header( obuf(icurs/itypesize:), hdrbufsize, itypesize, ftypesize, &
                                          DataHandle, DateStr, Element, IData, Count, code )
                  icurs = icurs + hdrbufsize

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                  CASE (IO_PNETCDF  )
                      CALL ext_pnc_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef NETCDF
                   CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef INTIO
                   CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef YYY
                   CASE ( IO_YYY )
                      CALL ext_yyy_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef GRIB1
                   CASE ( IO_GRIB1 )
                      CALL ext_gr1_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
#ifdef GRIB2
                   CASE ( IO_GRIB2 )
                      CALL ext_gr2_put_dom_td_integer( handle(DataHandle),TRIM(Element), Trim(DateStr), IData, Count, Status )
#endif
                   CASE DEFAULT
                      Status = 0
                   END SELECT

                   DEALLOCATE( IData )

! Every I/O server handles the "put_dom_ti_integer" request.
                CASE ( int_dom_ti_integer )

                  CALL mpi_type_size( MPI_INTEGER, ftypesize, ierr )
                  ALLOCATE( IData( obuf(icurs/itypesize + 4 ) ) )      ! 5 is the count of data items for this record ; defined in collect_on_comm.c
                  CALL int_get_ti_header( obuf(icurs/itypesize:), hdrbufsize, itypesize, ftypesize, &
                                          DataHandle, Element, IData, Count, code )
                  icurs = icurs + hdrbufsize
                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                    CASE (IO_PNETCDF  )
                      CALL ext_pnc_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_put_dom_ti_integer( handle(DataHandle),TRIM(Element), IData, Count, Status )
#endif

                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  DEALLOCATE( IData)
 
! Every I/O server  handles the "set_time" request.
                CASE ( int_set_time )

                  CALL int_get_ti_header_char( obuf(icurs/itypesize), hdrbufsize, itypesize, &
                                               DataHandle, Element, VarName, CData, code )
                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_set_time ( handle(DataHandle), TRIM(CData), Status)
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT

                  icurs = icurs + hdrbufsize

! Every I/O server handles the "put_dom_ti_char" request.
                CASE ( int_dom_ti_char )

                  CALL int_get_ti_header_char( obuf(icurs/itypesize), hdrbufsize, itypesize, &
                                               DataHandle, Element, VarName, CData, code )

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                    CASE (IO_PNETCDF  )
                      CALL ext_pnc_put_dom_ti_char ( handle(DataHandle), TRIM(Element), Trim(CData), Status)
#endif
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
#ifdef YYY
                   CASE ( IO_YYY )
                      CALL ext_yyy_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
#ifdef GRIB1
                   CASE ( IO_GRIB1 )
                      CALL ext_gr1_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
#ifdef GRIB2
                   CASE ( IO_GRIB2 )
                      CALL ext_gr2_put_dom_ti_char ( handle(DataHandle), TRIM(Element), TRIM(CData), Status)
#endif
                   CASE DEFAULT
                      Status = 0
                   END SELECT

                  icurs = icurs + hdrbufsize

! Every I/O server handles the "put_var_ti_char" request.
                CASE ( int_var_ti_char )

                  CALL int_get_ti_header_char( obuf(icurs/itypesize), hdrbufsize, itypesize, &
                                               DataHandle, Element, VarName, CData, code )

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                    CASE (IO_PNETCDF  )
                      CALL ext_pnc_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status )
#endif
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
#ifdef YYY
                   CASE ( IO_YYY )
                      CALL ext_yyy_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
#ifdef GRIB1
                   CASE ( IO_GRIB1 )
                      CALL ext_gr1_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
#ifdef GRIB2
                   CASE ( IO_GRIB2 )
                      CALL ext_gr2_put_var_ti_char ( handle(DataHandle), TRIM(Element), TRIM(VarName), TRIM(CData), Status)
#endif
                   CASE DEFAULT
                      Status = 0
                   END SELECT

                  icurs = icurs + hdrbufsize

                CASE ( int_ioexit )
! ioexit is now handled by sending negative message length to server
                  CALL wrf_error_fatal( &
                         "quilt: should have handled int_ioexit already")
! Every I/O server handles the "ioclose" request.
                CASE ( int_ioclose )
                  CALL int_get_handle_header( obuf(icurs/itypesize), hdrbufsize, itypesize, &
                                              DataHandle , code )
                  icurs = icurs + hdrbufsize

                  IF ( DataHandle .GE. 1 ) THEN

                     SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                    CASE ( IO_PNETCDF   )
                      CALL ext_pnc_inquire_filename( handle(DataHandle), fname, fstat, Status )
                      IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                        CALL ext_pnc_ioclose(handle(DataHandle),Status)
                      ENDIF
#endif
#ifdef NETCDF
                     CASE ( IO_NETCDF   )
                        CALL ext_ncd_inquire_filename( handle(DataHandle), fname, fstat, Status )
                        IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                           CALL ext_ncd_ioclose(handle(DataHandle),Status)
                        ENDIF
#endif
#ifdef INTIO
                     CASE ( IO_INTIO   )
                        CALL ext_int_inquire_filename( handle(DataHandle), fname, fstat, Status )
                        IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                           CALL ext_int_ioclose(handle(DataHandle),Status)
                        ENDIF
#endif
#ifdef YYY
                     CASE ( IO_YYY )
                        CALL ext_yyy_inquire_filename( handle(DataHandle), fname, fstat, Status )
                        IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                           CALL ext_yyy_ioclose(handle(DataHandle),Status)
                        ENDIF
#endif
#ifdef GRIB1
                     CASE ( IO_GRIB1 )
                        CALL ext_gr1_inquire_filename( handle(DataHandle), fname, fstat, Status )
                        IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                           CALL ext_gr1_ioclose(handle(DataHandle),Status)
                        ENDIF
#endif
#ifdef GRIB2
                     CASE ( IO_GRIB2 )
                        CALL ext_gr2_inquire_filename( handle(DataHandle), fname, fstat, Status )
                        IF ( fstat .EQ. WRF_FILE_OPENED_FOR_WRITE .OR. fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                           CALL ext_gr2_ioclose(handle(DataHandle),Status)
                        ENDIF
#endif
                     CASE DEFAULT
                        Status = 0
                     END SELECT
                  ENDIF

! Every I/O server handles the "open_for_write_begin" request.
                CASE ( int_open_for_write_begin )

                  CALL int_get_ofwb_header( obuf(icurs/itypesize), hdrbufsize, itypesize, &
                                            FileName,SysDepInfo,io_form_arg,DataHandle )

!write(0,*)' int_open_for_write_begin itypesize ',itypesize,' itypesize ',itypesize
!write(0,*)' int_open_for_write_begin icurs ', icurs, hdrbufsize
!JMDEBUGwrite(0,*)' int_open_for_write_begin FileName ',TRIM(FileName) , ' DataHandle ', DataHandle
!write(0,*)' int_open_for_write_begin SysDepInfo ',TRIM(SysDepInfo) 
                  icurs = icurs + hdrbufsize
!write(0,*)' int_open_for_write_begin new icurs,tag,size ', icurs, get_hdr_tag( bigbuf(icurs/itypesize) ),get_hdr_rec_size( bigbuf(icurs/itypesize) )
                
                  io_form(DataHandle) = io_form_arg

                  SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                    CASE (IO_PNETCDF  )
                      CALL ext_pnc_open_for_write_begin(FileName,mpi_comm_local,mpi_comm_local,SysDepInfo,handle(DataHandle),Status )
#endif
#ifdef NETCDF
                    CASE ( IO_NETCDF   )
                      CALL ext_ncd_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
!write(0,*)'ext_ncd_open_for_write_begin ',Trim(FileName),DataHandle,handle(DataHandle),Status
#endif
#ifdef INTIO
                    CASE ( IO_INTIO   )
                      CALL ext_int_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
#endif
#ifdef YYY
                    CASE ( IO_YYY )
                       CALL ext_yyy_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
#endif
#ifdef GRIB1
                    CASE ( IO_GRIB1 )
                       CALL ext_gr1_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
#endif
#ifdef GRIB2
                    CASE ( IO_GRIB2 )
                       CALL ext_gr2_open_for_write_begin(FileName,Comm,IOComm,SysDepInfo,handle(DataHandle),Status)
#endif
                    CASE DEFAULT
                      Status = 0
                  END SELECT
                
                  okay_to_write(DataHandle) = .false.

! Every I/O server handles the "open_for_write_commit" request.
! In this case, the "okay_to_commit" is simply set to .true. so "write_field"
! (int_field) requests will initiate writes to disk.  Actual commit will be done after
! all requests in this batch have been handled.
                CASE ( int_open_for_write_commit )

                  CALL int_get_handle_header( obuf(icurs/itypesize), hdrbufsize, itypesize, &
                                              DataHandle , code )
                  icurs = icurs + hdrbufsize
                  okay_to_commit(DataHandle) = .true.

! Every I/O server handles the "write_field" (int_field) request.
! If okay_to_write(DataHandle) is .true. then the patch in the
! header (bigbuf) is written to disk using pNetCDF.  Note that this is where the actual
! "quilting" (reassembly of patches onto a full-size domain) is done.  If
! okay_to_write(DataHandle) is .false. then external I/O package interfaces
! are called to write metadata for I/O formats that support native metadata.
!
! NOTE that the I/O servers will only see write_field (int_field)
! requests AFTER an "iosync" request.
                CASE ( int_field )
                  CALL mpi_type_size( MPI_INTEGER, ftypesize, ierr )
                  CALL int_get_write_field_header ( obuf(icurs/itypesize), hdrbufsize, itypesize, ftypesize,  &
                                                    DataHandle , DateStr , VarName , Dummy , FieldType , Comm , IOComm, &
                                                    DomainDesc , MemoryOrder , Stagger , DimNames ,              &
                                                    DomainStart , DomainEnd ,                                    &
                                                    MemoryStart , MemoryEnd ,                                    &
                                                    PatchStart , PatchEnd )
!write(0,*)' int_field ',TRIM(VarName),DataHandle,okay_to_write(DataHandle)
                  icurs = icurs + hdrbufsize

                  IF ( okay_to_write(DataHandle) ) THEN

!!$                    WRITE(0,FMT="('>>> ',(A),1x,(A),1x,A2,I6,1x,3('[',I3,',',I3,'] '))") &
!!$                          TRIM(DateStr), TRIM(VarName), TRIM(MemoryOrder), &
!!$                        (PatchEnd(1)-PatchStart(1)+1)*(PatchEnd(2)-PatchStart(2)+1)*(PatchEnd(3)-PatchStart(3)+1), &
!!$PatchStart(1),PatchEnd(1),PatchStart(2),PatchEnd(2),PatchStart(3),PatchEnd(3)
!!$                    WRITE(0,FMT="('>>> ',(A),1x,(A),1x,I6,1x,3('[',I3,',',I3,'] '))") &
!!$                          TRIM(DateStr), TRIM(VarName),  DomainDesc, &
!!$                          DomainStart(1),DomainEnd(1),DomainStart(2),DomainEnd(2),DomainStart(3),DomainEnd(3)

                    IF ( FieldType .EQ. WRF_FLOAT .OR. FieldType .EQ. WRF_DOUBLE)  THEN
                      ! Note that the WRF_DOUBLE branch of this IF statement must come first since 
                      ! WRF_FLOAT is set equal to WRF_DOUBLE during autopromotion builds.  
                      IF ( FieldType .EQ. WRF_DOUBLE)  THEN
! this branch has not been tested TBH: 20050406
                        CALL mpi_type_size( MPI_DOUBLE_PRECISION, ftypesize, ierr )
                      ELSE
                        CALL mpi_type_size( MPI_REAL, ftypesize, ierr )
                      ENDIF

#ifdef PNETCDF_QUILT
!                      WRITE(mess,FMT="('>>> ',(A),1x,(A),1x,I6,1x,3('[',I3,',',I3,'] '))") &
!                          TRIM(DateStr), TRIM(VarName),  DomainDesc, &
!                          DomainStart(1),DomainEnd(1), &
!                          DomainStart(2),DomainEnd(2),DomainStart(3),DomainEnd(3)
!                      CALL wrf_message(mess)

                      CALL store_patch_in_outbuf_pnc(obuf(icurs/itypesize), &
                                                     dummybuf, TRIM(DateStr), &
                                                     TRIM(VarName) , &
                                                     FieldType,      &
                                                     TRIM(MemoryOrder), &
                                                     TRIM(Stagger), &
                                                     DimNames, &
                                                     DomainStart , DomainEnd ,&
                                                     MemoryStart , MemoryEnd ,&
                                                     PatchStart , PatchEnd, &
                                                     ntasks_io_group-1 )
                      stored_write_record = .true.

!!$                      IF(VarName .eq. "PSFC")THEN
!!$                         CALL dump_real_array_c(obuf(icurs/itypesize), DomainStart,&
!!$                                                DomainEnd, PatchStart, PatchEnd,   &
!!$                                                mytask_local, DomainDesc)
!!$                      ENDIF

#endif
                    ELSE IF ( FieldType .EQ. WRF_INTEGER ) THEN
                      CALL mpi_type_size( MPI_INTEGER, ftypesize, ierr )
#ifdef PNETCDF_QUILT
                      CALL store_patch_in_outbuf_pnc ( dummybuf,             & 
                                                   obuf(icurs/itypesize) ,   &
                                                   TRIM(DateStr) ,           &
                                                   TRIM(VarName) ,           &
                                                   FieldType,                &
                                                   TRIM(MemoryOrder) ,       &
                                                   TRIM(Stagger), DimNames,  &
                                                   DomainStart , DomainEnd , &
                                                   MemoryStart , MemoryEnd , &
                                                   PatchStart , PatchEnd   , &
                                                   ntasks_io_group-1 )
                      stored_write_record = .true.
#endif
                    ELSE IF ( FieldType .EQ. WRF_LOGICAL ) THEN
                      ftypesize = LWORDSIZE
                    ENDIF

                    icurs = icurs + (PatchEnd(1)-PatchStart(1)+1)* &
                                    (PatchEnd(2)-PatchStart(2)+1)* &
                                    (PatchEnd(3)-PatchStart(3)+1)*ftypesize

                  ELSE ! Write metadata only (or do 'training'?)

                    SELECT CASE (use_package(io_form(DataHandle)))

#ifdef PNETCDF_QUILT
                      CASE ( IO_PNETCDF )
                        CALL ext_pnc_write_field ( handle(DataHandle) , TRIM(DateStr),        &
                                   TRIM(VarName) , dummy , FieldType , mpi_comm_local , mpi_comm_local,         &
                                   DomainDesc , TRIM(MemoryOrder) , TRIM(Stagger), DimNames , &
                                   DomainStart , DomainEnd ,                                  &
                                   MemoryStart , MemoryEnd ,                                  &
                                   PatchStart ,  PatchEnd,                                  &
                                   Status )
#endif
#ifdef NETCDF
                      CASE ( IO_NETCDF   )
                        CALL ext_ncd_write_field ( handle(DataHandle) , TRIM(DateStr) ,         &
                                   TRIM(VarName) , dummy , FieldType , Comm , IOComm,           &
                                   DomainDesc , TRIM(MemoryOrder) , TRIM(Stagger) , DimNames ,  &
                                   DomainStart , DomainEnd ,                                    &
                                   DomainStart , DomainEnd ,                                    &
                                   DomainStart , DomainEnd ,                                    &
                                   Status )
#endif
#if 0
! since this is training and the grib output doesn't need training, disable this branch.
#ifdef YYY
                 CASE ( IO_YYY )
                      CALL ext_YYY_write_field ( handle(DataHandle) , TRIM(DateStr) ,         &
                                 TRIM(VarName) , dummy , FieldType , Comm , IOComm,           &
                                 DomainDesc , TRIM(MemoryOrder) , TRIM(Stagger) , DimNames ,  &
                                 DomainStart , DomainEnd ,                                    &
                                 DomainStart , DomainEnd ,                                    &
                                 DomainStart , DomainEnd ,                                    &
                                 Status )
#endif
#endif
                      CASE DEFAULT
                        Status = 0
                    END SELECT
                  ENDIF
                CASE ( int_iosync )
                  CALL int_get_handle_header( obuf(icurs/itypesize), hdrbufsize, itypesize, &
                                            DataHandle , code )
                  icurs = icurs + hdrbufsize
                CASE DEFAULT
                  WRITE(mess,*)'quilt: bad tag: ',                            &
                               get_hdr_tag( obuf(icurs/itypesize) ),' icurs ',&
                               icurs/itypesize
                  CALL wrf_error_fatal( mess )
              END SELECT

            ENDDO !}
! Now, we have finshed handling all commands from the latest
! call to retrieve_pieces_of_field().

            IF (stored_write_record) THEN
! If any field patches have been stored in internal output buffers
! (via a call to store_patch_in_outbuf_pnc()) then call write_outbuf_pnc() 
! to write them to disk now.
! NOTE that the I/O server will only have called
! store_patch_in_outbuf() when handling write_field (int_field)
! commands which only arrive AFTER an "iosync" command.
!              CALL start_timing
#ifdef PNETCDF_QUILT
              CALL write_outbuf_pnc( handle(DataHandle), &
                                     use_package(io_form(DataHandle)), &
                                     mpi_comm_local, mytask_local,     &
                                     ntasks_local_group) 
#endif
!              CALL end_timing( "quilt_pnc: call to write_outbuf_pnc" ) 
              stored_write_record = .false.
              written_record = .true.
            ENDIF

! If one or more "open_for_write_commit" commands were encountered from the
! latest call to retrieve_pieces_of_field() then call the package-specific
! routine to do the commit.
            IF (okay_to_commit(DataHandle)) THEN

              SELECT CASE (use_package(io_form(DataHandle)))
#ifdef PNETCDF_QUILT
                CASE ( IO_PNETCDF   )
                  CALL ext_pnc_inquire_filename( handle(DataHandle), fname, fstat, Status )
                  IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                    CALL ext_pnc_open_for_write_commit(handle(DataHandle),Status)
                    okay_to_write(DataHandle) = .true.
                  ENDIF
#endif
#ifdef NETCDF
                CASE ( IO_NETCDF   )
                  CALL ext_ncd_inquire_filename( handle(DataHandle), fname, fstat, Status )
                  IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                    CALL ext_ncd_open_for_write_commit(handle(DataHandle),Status)
                    okay_to_write(DataHandle) = .true.
                  ENDIF
#endif
#ifdef INTIO
                CASE ( IO_INTIO   )
                  CALL ext_int_inquire_filename( handle(DataHandle), fname, fstat, Status )
                  IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                    CALL ext_int_open_for_write_commit(handle(DataHandle),Status)
                    okay_to_write(DataHandle) = .true.
                  ENDIF
#endif
#ifdef YYY
                 CASE ( IO_YYY )
                    CALL ext_yyy_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                       CALL ext_yyy_open_for_write_commit(handle(DataHandle),Status)
                       okay_to_write(DataHandle) = .true.
                    ENDIF
#endif
#ifdef GRIB1
                 CASE ( IO_GRIB1 )
                    CALL ext_gr1_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                       CALL ext_gr1_open_for_write_commit(handle(DataHandle),Status)
                       okay_to_write(DataHandle) = .true.
                    ENDIF
#endif
#ifdef GRIB2
                 CASE ( IO_GRIB2 )
                    CALL ext_gr2_inquire_filename( handle(DataHandle), fname, fstat, Status )
                    IF ( fstat .EQ. WRF_FILE_OPENED_NOT_COMMITTED ) THEN
                       CALL ext_gr2_open_for_write_commit(handle(DataHandle),Status)
                       okay_to_write(DataHandle) = .true.
                    ENDIF
#endif

                CASE DEFAULT
                  Status = 0
              END SELECT

            okay_to_commit(DataHandle) = .false.
          ENDIF
!!endif

! Retrieve header and all patches for the next field from the internal 
! buffers.  
        CALL retrieve_pieces_of_field ( obuf , VarName, obufsize, sz, retval )
      END DO !}

      DEALLOCATE( obuf )

      ! flush output files if needed
      IF (written_record) THEN
!CALL start_timing
        SELECT CASE ( use_package(io_form) )
#ifdef PNETCDF_QUILT
          CASE ( IO_PNETCDF   )
            CALL ext_pnc_iosync( handle(DataHandle), Status )
#endif
          CASE DEFAULT
            Status = 0
        END SELECT
        written_record = .false.
!CALL end_timing( "quilt_pnc: flush" )
      ENDIF

      END DO ! }

    END SUBROUTINE quilt_pnc

! end of #endif of DM_PARALLEL
#endif

    SUBROUTINE init_module_wrf_quilt
!<DESCRIPTION>
! Both client (compute) and server tasks call this routine to initialize the 
! module.  Routine setup_quilt_servers() is called from this routine to 
! determine which tasks are compute tasks and which are server tasks.  Server 
! tasks then call routine quilt() and remain there for the rest of the model 
! run.  Compute tasks return from init_module_wrf_quilt() to perform model 
! computations.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
      IMPLICIT NONE
      INCLUDE 'mpif.h'
      INTEGER i
      NAMELIST /namelist_quilt/ nio_tasks_per_group, nio_groups, poll_servers
      INTEGER ntasks, mytask, ierr, io_status
#  if defined(_OPENMP) && defined(MPI2_THREAD_SUPPORT)
      INTEGER thread_support_provided, thread_support_requested
#endif
      INTEGER mpi_comm_here, temp_poll
      LOGICAL mpi_inited
      LOGICAL esmf_coupling

!TODO:  Change this to run-time switch
#ifdef ESMFIO
      esmf_coupling = .TRUE.
#else
      esmf_coupling = .FALSE.
#endif

      quilting_enabled = .FALSE.
      IF ( disable_quilt ) RETURN

      DO i = 1,int_num_handles
        okay_to_write(i) = .FALSE.
        int_handle_in_use(i) = .FALSE.
        server_for_handle(i) = 0 
        int_num_bytes_to_write(i) = 0
      ENDDO

      CALL MPI_INITIALIZED( mpi_inited, ierr )
      IF ( .NOT. mpi_inited ) THEN
#  if defined(_OPENMP) && defined(MPI2_THREAD_SUPPORT)
        thread_support_requested = MPI_THREAD_FUNNELED
        CALL mpi_init_thread ( thread_support_requested, thread_support_provided, ierr )
        IF ( thread_support_provided .lt. thread_support_requested ) THEN
           CALL WRF_ERROR_FATAL( "failed to initialize MPI thread support")
        ENDIF
#  else
        CALL mpi_init ( ierr )
#  endif
        CALL wrf_set_dm_communicator( MPI_COMM_WORLD )
        CALL wrf_termio_dup
      ENDIF
      CALL wrf_get_dm_communicator( mpi_comm_here )

      CALL MPI_Comm_rank ( mpi_comm_here, mytask, ierr ) ;
      CALL mpi_x_comm_size ( mpi_comm_here, ntasks, ierr ) ;

      IF ( mytask .EQ. 0 ) THEN
        OPEN ( unit=27, file="namelist.input", form="formatted", status="old" )
        nio_groups = 1
        nio_tasks_per_group  = 0
        poll_servers = .false.
        READ ( 27 , NML = namelist_quilt, IOSTAT=io_status )
        IF (io_status .NE. 0) THEN
          CALL wrf_error_fatal( "ERROR reading namelist namelist_quilt" )
        ENDIF
        CLOSE ( 27 )
        IF ( esmf_coupling ) THEN
          IF ( nio_tasks_per_group > 0 ) THEN
            CALL wrf_error_fatal("frame/module_io_quilt.F: cannot use "// &
                                 "ESMF coupling with quilt tasks") ;
          ENDIF
        ENDIF
        if(poll_servers) then
           temp_poll=1
        else
           temp_poll=0
        endif
      ENDIF

      CALL mpi_bcast( nio_tasks_per_group  , 1 , MPI_INTEGER , 0 , mpi_comm_here, ierr )
      CALL mpi_bcast( nio_groups , 1 , MPI_INTEGER , 0 , mpi_comm_here, ierr )
      CALL mpi_bcast( temp_poll , 1 , MPI_INTEGER , 0 , mpi_comm_here, ierr )

      poll_servers = (temp_poll == 1)

      CALL setup_quilt_servers( nio_tasks_per_group,            &
                                mytask,               &
                                ntasks,               &
                                nio_groups,           &
                                nio_tasks_in_group,   &
                                mpi_comm_here,        &
                                mpi_comm_local,       &
                                mpi_comm_io_groups)

       ! provide the communicator for the integration tasks to RSL
       IF ( compute_node ) THEN
          CALL wrf_set_dm_communicator( mpi_comm_local )
#ifdef HWRF
          call ATM_SET_COMM(mpi_comm_local)
#endif
       ELSE
#ifdef HWRF
          call ATM_LEAVE_COUPLING()
#endif
          CALL quilt    ! will not return on io server tasks
       ENDIF
#endif
      RETURN
    END SUBROUTINE init_module_wrf_quilt


#ifdef IBM_REDUCE_BUG_WORKAROUND

    ! These three subroutines re-implement MPI_Reduce on MPI_INTEGER
    ! with OP=MPI_ADD.

    ! This is a workaround for a bug in the IBM MPI implementation.
    ! Some MPI processes will get stuck in MPI_Reduce and not
    ! return until the PREVIOUS I/O server group finishes writing.

    ! This workaround replaces the MPI_Reduce call with many 
    ! MPI_Send and MPI_Recv calls that perform the sum on the
    ! root of the communicator.

    ! There are two reduce routines: one for a sum of scalars
    ! and one for a sum of arrays.  The get_reduce_tag generates
    ! MPI tags for the communication.

  integer function get_reduce_tag(root,comm)
    implicit none
    include 'mpif.h'
    integer, intent(in) :: comm,root
    integer :: i,j, tag, here
    integer :: ierr,me,size

    integer, pointer :: nexttags(:)
    integer, target :: dummy(1)
    character(255) :: message
    integer(kind=4) :: comm4,hashed

    integer, parameter :: hashsize = 113 ! should be prime, >max_servers+1
    integer, parameter :: tagloop = 100000 ! number of tags reserved per communicator
    integer, parameter :: origin = 1031102 ! lowest tag number we'll use
    integer, save :: nexttag=origin ! next tag to use for a new communicator
    integer, save :: comms(hashsize)=-1, firsttag(hashsize)=0, curtag(hashsize)=0

    ! If integers are not four bytes, this implementation will still
    ! work, but it may be inefficient (O(N) lookup instead of O(1)).
    ! To fix that, an eight byte hash function would be needed, but
    ! integers are four bytes in WRF, so that is not a problem right
    ! now.

    comm4=comm
    call int_hash(comm4,hashed)
    hashed=mod(abs(hashed),hashsize)+1
    if(hashed<0) call wrf_error_fatal('hashed<0')

    do i=0,hashsize-1
       j=1+mod(i+hashed-1,hashsize)

       if(firsttag(j)/=0 .and. comms(j)==comm) then
          ! Found the communicator
          if(curtag(j)-firsttag(j) >= tagloop) then
             ! Hit the max tag number so we need to reset.
             ! To make sure >tagloop reduces don't happen 
             ! before someone finishes an old reduce, we 
             ! have an MPI_Barrier here.
             !call wrf_message('Hit tagloop limit so calling mpi_barrier in get_reduce_tag...')
             call mpi_barrier(comm,ierr)
             if(ierr/=0) call wrf_error_fatal('cannot call mpi_barrier')
             !call wrf_message('  ...back from mpi_barrier in get_reduce_tag.')

             curtag(j)=firsttag(j)
          endif

          tag=curtag(j)
          curtag(j)=tag+1
          get_reduce_tag=tag
          return
       endif
    enddo


    ! ==================== HANDLE NEW COMMUNICATORS ====================

    !write(message,'("Found a new communicator ",I0," in get_reduce_tag, so making a tag range for it")') comm

    ! If we get here, the communicator is new to us, so we need
    ! to add it to the hash and give it a new tag.  

    ! First, figure out where we'll put the tag in the hashtable
    here=-1
    do i=0,hashsize-1
       j=1+mod(i+hashed-1,hashsize)

       if(firsttag(j)==0) then
          here=j
          exit
       endif
    enddo
    if(here==-1) call wrf_error_fatal('no room in hashtable; increase hashsize in get_reduce_tag (should be >max_servers+1)')

    ! Now, find out the new tag's number.  To do this, we need to
    ! get the next tag number that is not used by any ranks.

    call mpi_comm_rank(comm,me,ierr)
    if(ierr/=0) call wrf_error_fatal('cannot call mpi_comm_rank')

    call mpi_comm_size(comm,size,ierr)
    if(ierr/=0) call wrf_error_fatal('cannot call mpi_comm_size')

    if(me==root) then
       allocate(nexttags(size))
    else
       nexttags=>dummy
    endif

    call mpi_gather(nexttag,1,MPI_INTEGER,nexttags,1,MPI_INTEGER,root,comm,ierr)
    if(ierr/=0) call wrf_error_fatal('cannot call mpi_gather')

    if(me==root) then
       nexttag=max(nexttag,maxval(nexttags))
       deallocate(nexttags)
    endif
    call mpi_bcast(nexttag,1,MPI_INTEGER,root,comm,ierr)

    comms(here)=comm
    firsttag(here)=nexttag
    curtag(here)=nexttag
    get_reduce_tag=nexttag

    !write(message,'("Stored comm ",I0," with tag ",I0,"=",I0," in hash element ",I0)') &
    !     comms(here),firsttag(here),curtag(here),here
    !call wrf_message(message)

    nexttag=nexttag+tagloop

  end function get_reduce_tag
  subroutine reduce_add_int_scl(send,recv,count,root,comm)
    implicit none
    include 'mpif.h'
    integer, intent(in) :: count,root,comm
    integer, intent(inout) :: recv
    integer, intent(in) :: send
    integer :: me, size, ierr, you, temp, tag
    character*255 :: message
    if(root<0) call wrf_error_fatal('root is less than 0')

    tag=get_reduce_tag(root,comm)

    !write(message,'("Send/recv to tag ",I0)') tag
    !call wrf_message(message)

    call mpi_comm_rank(comm,me,ierr)
    if(ierr/=0) call wrf_error_fatal('cannot call mpi_comm_rank')

    call mpi_comm_size(comm,size,ierr)
    if(ierr/=0) call wrf_error_fatal('cannot call mpi_comm_size')

    if(root>=size) call wrf_error_fatal('root is beyond highest communicator rank')

    if(me==root) then
       recv=send
       do you=0,size-2
          call mpi_recv(temp,1,MPI_INTEGER,MPI_ANY_SOURCE,tag,comm,MPI_STATUS_IGNORE,ierr)
          if(ierr/=0) call wrf_error_fatal('error calling mpi_recv')
          recv=recv+temp
       enddo
    else
       call mpi_send(send,1,MPI_INTEGER,root,tag,comm,ierr)
       if(ierr/=0) call wrf_error_fatal('error calling mpi_send')
    endif
  end subroutine reduce_add_int_scl
  subroutine reduce_add_int_arr(sendbuf,recvbuf,count,root,comm)
    implicit none
    include 'mpif.h'
    integer, intent(in) :: count,root,comm
    integer, intent(in) :: sendbuf(count)
    integer, intent(inout) :: recvbuf(count)
    integer :: me, size, ierr, you, tempbuf(count), tag
    character*255 :: message

    if(root<0) call wrf_error_fatal('root is less than 0')

    tag=get_reduce_tag(root,comm)

    !write(message,'("Send/recv to tag ",I0)') tag
    !call wrf_message(message)

    call mpi_comm_rank(comm,me,ierr)
    if(ierr/=0) call wrf_error_fatal('cannot call mpi_comm_rank')

    call mpi_comm_size(comm,size,ierr)
    if(ierr/=0) call wrf_error_fatal('cannot call mpi_comm_size')

    if(root>=size) call wrf_error_fatal('root is beyond highest communicator rank')

    if(me==root) then
       recvbuf=sendbuf
       do you=0,size-2
          call mpi_recv(tempbuf,count,MPI_INTEGER,MPI_ANY_SOURCE,tag,comm,MPI_STATUS_IGNORE,ierr)
          if(ierr/=0) call wrf_error_fatal('error calling mpi_recv')
          recvbuf=recvbuf+tempbuf
       enddo
    else
       call mpi_send(sendbuf,count,MPI_INTEGER,root,tag,comm,ierr)
       if(ierr/=0) call wrf_error_fatal('error calling mpi_send')
    endif
  end subroutine reduce_add_int_arr
#endif


END MODULE module_wrf_quilt

!<DESCRIPTION>
! Remaining routines in this file are defined outside of the module
! either to defeat arg/param type checking or to avoid an explicit use
! dependence.
!</DESCRIPTION>

SUBROUTINE disable_quilting
!<DESCRIPTION>
! Call this in programs that you never want to be quilting (e.g. real)
! Must call before call to init_module_wrf_quilt().  
!</DESCRIPTION>
  USE module_wrf_quilt
  disable_quilt = .TRUE.
  RETURN
END SUBROUTINE disable_quilting

LOGICAL FUNCTION  use_output_servers()
!<DESCRIPTION>
! Returns .TRUE. if I/O quilt servers are in-use for write operations.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
  USE module_wrf_quilt
  use_output_servers = quilting_enabled
  RETURN
END FUNCTION use_output_servers

LOGICAL FUNCTION  use_input_servers()
!<DESCRIPTION>
! Returns .TRUE. if I/O quilt servers are in-use for read operations.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
  USE module_wrf_quilt
  use_input_servers = .FALSE.
  RETURN
END FUNCTION use_input_servers

SUBROUTINE wrf_quilt_open_for_write_begin( FileName , Comm_compute, Comm_io, SysDepInfo, &
                                     DataHandle , io_form_arg, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to begin data definition ("training") phase
! for writing to WRF dataset FileName.  io_form_arg indicates file format.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  USE module_state_description, ONLY: IO_PNETCDF
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  CHARACTER *(*), INTENT(IN)  :: FileName
  INTEGER ,       INTENT(IN)  :: Comm_compute , Comm_io
  CHARACTER *(*), INTENT(IN)  :: SysDepInfo
  INTEGER ,       INTENT(OUT) :: DataHandle
  INTEGER ,       INTENT(IN)  :: io_form_arg
  INTEGER ,       INTENT(OUT) :: Status
! Local
  CHARACTER*132   :: locFileName, locSysDepInfo
  INTEGER i, itypesize, tasks_in_group, ierr, comm_io_group
  REAL dummy
  INTEGER, EXTERNAL :: use_package

  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_open_for_write_begin' ) 
  CALL int_get_fresh_handle(i)
  okay_to_write(i) = .false.
  DataHandle = i

  locFileName = FileName
  locSysDepInfo = SysDepInfo

  CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )

  SELECT CASE(use_package(io_form_arg))

#ifdef PNETCDF_QUILT
  CASE(IO_PNETCDF)
     IF(compute_group_master(1)) THEN
        CALL int_gen_ofwb_header( hdrbuf, hdrbufsize, itypesize, &
                                  locFileName,locSysDepInfo,io_form_arg,&
                                  DataHandle )
     ELSE
        CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
     END IF
#endif
  CASE DEFAULT

     IF ( wrf_dm_on_monitor() ) THEN
        CALL int_gen_ofwb_header( hdrbuf, hdrbufsize, itypesize, &
                                  locFileName,locSysDepInfo,io_form_arg,DataHandle )
     ELSE
        CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
     ENDIF

  END SELECT

  iserver = get_server_id ( DataHandle )
!JMDEBUGwrite(0,*)'wrf_quilt_open_for_write_begin iserver = ', iserver
  CALL get_mpi_comm_io_groups( comm_io_group , iserver )
!JMDEBUGwrite(0,*)'wrf_quilt_open_for_write_begin comm_io_group  = ', comm_io_group 

  CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )
!JMDEBUGwrite(0,*)'mpi_x_comm_size tasks_in_group ',tasks_in_group, ierr

!!JMTIMING  CALL start_timing
  ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
  reduced = 0
  reduced(1) = hdrbufsize 
#ifdef PNETCDF_QUILT
  IF ( compute_group_master(1) ) reduced(2) = i
#else
  IF ( wrf_dm_on_monitor() )  reduced(2) = i 
#endif
  CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )
!!JMTIMING   CALL end_timing("MPI_Reduce in wrf_quilt_open_for_write_begin")

  ! send data to the i/o processor
  CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                        onebyte,                       &
                        hdrbuf, hdrbufsize , &
                        dummy, 0 )

  Status = 0


#endif
  RETURN  
END SUBROUTINE wrf_quilt_open_for_write_begin

SUBROUTINE wrf_quilt_open_for_write_commit( DataHandle , Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to switch an internal flag to enable output
! for the dataset referenced by DataHandle.  The call to
! wrf_quilt_open_for_write_commit() must be paired with a call to
! wrf_quilt_open_for_write_begin().
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  INTEGER ,       INTENT(IN ) :: DataHandle
  INTEGER ,       INTENT(OUT) :: Status
  INTEGER i, itypesize, tasks_in_group, ierr, comm_io_group
  REAL dummy

  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_open_for_write_commit' ) 
  IF ( DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles ) THEN
    IF ( int_handle_in_use( DataHandle ) ) THEN
      okay_to_write( DataHandle ) = .true.
    ENDIF
  ENDIF

  CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )

#ifdef PNETCDF_QUILT
!ARP Only want one command to be received by each IO server when using
!ARP parallel IO
  IF(compute_group_master(1)) THEN
     CALL int_gen_handle_header( hdrbuf, hdrbufsize, itypesize, &
                                 DataHandle, int_open_for_write_commit )
  ELSE
     CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
  END IF
#else

  IF ( wrf_dm_on_monitor() ) THEN
     CALL int_gen_handle_header( hdrbuf, hdrbufsize, itypesize, &
                                 DataHandle, int_open_for_write_commit )
  ELSE
     CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
  ENDIF
#endif

  iserver = get_server_id ( DataHandle )
  CALL get_mpi_comm_io_groups( comm_io_group , iserver )

  CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )

!!JMTIMING  CALL start_timing
  ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
  reduced = 0
  reduced(1) = hdrbufsize 
#ifdef PNETCDF_QUILT
  IF ( compute_group_master(1) ) reduced(2) = DataHandle
#else
  IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
  CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )
!!JMTIMING   CALL end_timing("MPI_Reduce in wrf_quilt_open_for_write_commit")

  ! send data to the i/o processor
  CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                        onebyte,                       &
                        hdrbuf, hdrbufsize , &
                        dummy, 0 )

  Status = 0

#endif
  RETURN  
END SUBROUTINE wrf_quilt_open_for_write_commit

SUBROUTINE wrf_quilt_open_for_read ( FileName , Comm_compute, Comm_io, SysDepInfo, &
                               DataHandle , Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to open WRF dataset FileName for reading.
! This routine is called only by client (compute) tasks.  
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  CHARACTER *(*), INTENT(IN)  :: FileName
  INTEGER ,       INTENT(IN)  :: Comm_compute , Comm_io
  CHARACTER *(*), INTENT(IN)  :: SysDepInfo
  INTEGER ,       INTENT(OUT) :: DataHandle
  INTEGER ,       INTENT(OUT) :: Status

  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_open_for_read' ) 
  DataHandle = -1
  Status = -1
  CALL wrf_error_fatal ( "frame/module_io_quilt.F: wrf_quilt_open_for_read not yet supported" )
#endif
  RETURN  
END SUBROUTINE wrf_quilt_open_for_read

SUBROUTINE wrf_quilt_inquire_opened ( DataHandle, FileName , FileStatus, Status )
!<DESCRIPTION>
! Inquire if the dataset referenced by DataHandle is open.
! Does not require communication with I/O servers.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  IMPLICIT NONE
#include "wrf_io_flags.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER *(*), INTENT(IN)  :: FileName
  INTEGER ,       INTENT(OUT) :: FileStatus
  INTEGER ,       INTENT(OUT) :: Status

  Status = 0

  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_inquire_opened' ) 
  IF ( DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles ) THEN
    IF ( int_handle_in_use( DataHandle ) ) THEN
      IF ( okay_to_write( DataHandle ) ) THEN
        FileStatus = WRF_FILE_OPENED_FOR_WRITE
      ENDIF
    ENDIF
  ENDIF
  Status = 0
  
#endif
  RETURN
END SUBROUTINE wrf_quilt_inquire_opened

SUBROUTINE wrf_quilt_inquire_filename ( DataHandle, FileName , FileStatus, Status )
!<DESCRIPTION>
! Return the Filename and FileStatus associated with DataHandle.
! Does not require communication with I/O servers.
!
! Note that the current implementation does not actually return FileName.
! Currenlty, WRF does not use this returned value.  Fixing this would simply
! require saving the file names on the client tasks in an array similar to
! okay_to_write().
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  IMPLICIT NONE
#include "wrf_io_flags.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER *(*), INTENT(OUT) :: FileName
  INTEGER ,       INTENT(OUT) :: FileStatus
  INTEGER ,       INTENT(OUT) :: Status
  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_inquire_filename' ) 
  Status = 0
  IF ( DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles ) THEN
    IF ( int_handle_in_use( DataHandle ) ) THEN
      IF ( okay_to_write( DataHandle ) ) THEN
        FileStatus = WRF_FILE_OPENED_FOR_WRITE
      ELSE
        FileStatus = WRF_FILE_OPENED_NOT_COMMITTED
      ENDIF
    ELSE
        FileStatus = WRF_FILE_NOT_OPENED
    ENDIF
    Status = 0
    FileName = "bogusfornow"
  ELSE
    Status = -1
  ENDIF
#endif
  RETURN
END SUBROUTINE wrf_quilt_inquire_filename

SUBROUTINE wrf_quilt_iosync ( DataHandle, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to synchronize the disk copy of a dataset
! with memory buffers.
!
! After the "iosync" header (request) is sent to the I/O quilt server,
! the compute tasks will then send the entire contents (headers and data) of
! int_local_output_buffer to their I/O quilt server.  This communication is
! done in subroutine send_to_io_quilt_servers().  After the I/O quilt servers
! receive this data, they will write all accumulated fields to disk.
!
! Significant time may be required for the I/O quilt servers to organize
! fields and write them to disk.  Therefore, the "iosync" request should be
! sent only when the compute tasks are ready to run for a while without
! needing to communicate with the servers.  Otherwise, the compute tasks
! will end up waiting for the servers to finish writing to disk, thus wasting
! any performance benefits of having servers at all.
!
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if  defined( DM_PARALLEL ) && ! defined (STUBMPI) 
  USE module_wrf_quilt
  IMPLICIT NONE
  include "mpif.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  INTEGER ,       INTENT(OUT) :: Status

  INTEGER locsize , itypesize
  INTEGER ierr, tasks_in_group, comm_io_group, dummy, i

  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_iosync' ) 

!  CALL start_timing
  IF ( associated ( int_local_output_buffer ) ) THEN

    iserver = get_server_id ( DataHandle )
    CALL get_mpi_comm_io_groups( comm_io_group , iserver )

    CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )

    locsize = int_num_bytes_to_write(DataHandle)

!    CALL start_timing
    ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
    reduced = 0
    reduced(1) = locsize 
#ifdef PNETCDF_QUILT
! ARP Only want one command per IOServer if doing parallel IO
    IF ( compute_group_master(1) ) reduced(2) = DataHandle
#else
    IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
    CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )
!    CALL end_timing("MPI_Reduce in wrf_quilt_iosync")

    ! send data to the i/o processor
#ifdef DEREF_KLUDGE
    CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                          onebyte,                       &
                          int_local_output_buffer(1), locsize , &
                          dummy, 0 )
#else
    CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                          onebyte,                       &
                          int_local_output_buffer, locsize , &
                          dummy, 0 )
#endif


    int_local_output_cursor = 1
!    int_num_bytes_to_write(DataHandle) = 0
    DEALLOCATE ( int_local_output_buffer )
    NULLIFY ( int_local_output_buffer )
  ELSE
    CALL wrf_message ("frame/module_io_quilt.F: wrf_quilt_iosync: no buffer allocated")
  ENDIF
!  CALL end_timing("wrf_quilt_iosync")
  Status = 0
#endif
  RETURN
END SUBROUTINE wrf_quilt_iosync

SUBROUTINE wrf_quilt_ioclose ( DataHandle, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to close the dataset referenced by
! DataHandle.
! This routine also clears the client file handle and, if needed, deallocates
! int_local_output_buffer.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && ! defined( STUBMPI) 
  USE module_wrf_quilt
  USE module_timing
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  INTEGER ,       INTENT(OUT) :: Status
  INTEGER i, itypesize, tasks_in_group, comm_io_group, ierr
  REAL dummy

!!JMTIMING  CALL start_timing
  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_ioclose' ) 
  CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )

! If we're using pnetcdf then each IO server will need to receive the 
! handle just once as there is
! no longer a reduce over the IO servers to get it.
#ifdef PNETCDF_QUILT
  IF ( compute_group_master(1) )THEN
     CALL int_gen_handle_header( hdrbuf, hdrbufsize, itypesize, &
                                 DataHandle, int_ioclose )
  ELSE
     CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
  ENDIF
#else
  IF ( wrf_dm_on_monitor() ) THEN
     CALL int_gen_handle_header( hdrbuf, hdrbufsize, itypesize, &
                                 DataHandle , int_ioclose )
  ELSE
     CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
  ENDIF
#endif

  iserver = get_server_id ( DataHandle )
  CALL get_mpi_comm_io_groups( comm_io_group , iserver )

  CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )

!!JMTIMING  CALL start_timing
  ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
  reduced = 0
#ifdef PNETCDF_QUILT
! If we're using pnetcdf then each IO server will need the handle as there is
! no longer a reduce over the IO servers to get it.
  IF ( compute_group_master(1) ) reduced(2) = DataHandle
#else
  IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
  CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )
!!JMTIMING   CALL end_timing("MPI_Reduce in ioclose")

#if 0
  ! send data to the i/o processor
!!JMTIMING  CALL start_timing
  CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                        onebyte,                       &
                        hdrbuf, hdrbufsize , &
                        dummy, 0 )
!!JMTIMING   CALL end_timing("collect_on_comm in io_close")
#endif

  int_handle_in_use(DataHandle) = .false.
  CALL set_server_id( DataHandle, 0 ) 
  okay_to_write(DataHandle) = .false.
  okay_to_commit(DataHandle) = .false.
  int_local_output_cursor = 1
  int_num_bytes_to_write(DataHandle) = 0
  IF ( associated ( int_local_output_buffer ) ) THEN
    DEALLOCATE ( int_local_output_buffer )
    NULLIFY ( int_local_output_buffer )
  ENDIF

  Status = 0
!!JMTIMING   CALL end_timing( "wrf_quilt_ioclose" )

#endif
  RETURN
END SUBROUTINE wrf_quilt_ioclose

SUBROUTINE wrf_quilt_ioexit( Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to shut down the WRF I/O system.
! Do not call any wrf_quilt_*() routines after this routine has been called.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && ! defined (STUBMPI ) 
  USE module_wrf_quilt
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  INTEGER ,       INTENT(OUT) :: Status
  INTEGER                     :: DataHandle, actual_iserver
  INTEGER i, itypesize, tasks_in_group, comm_io_group, me, ierr 
  REAL dummy

  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_ioexit' ) 
  CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )

!ARPDBG - potential bug. Have no access to what type of IO is being used for
! this data so if PNETCDF_QUILT is defined then we assume that's what's being used.
#ifdef PNETCDF_QUILT
!ARP Send the ioexit message just once to each IOServer when using parallel IO
  IF( compute_group_master(1) ) THEN
     CALL int_gen_handle_header( hdrbuf, hdrbufsize, itypesize, &
                                 DataHandle, int_ioexit )
  ELSE
     CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
  END IF
#else

  IF ( wrf_dm_on_monitor() ) THEN
     CALL int_gen_handle_header( hdrbuf, hdrbufsize, itypesize, &
                                 DataHandle , int_ioexit )  ! Handle is dummy
  ELSE
     CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
  ENDIF
#endif

  DO iserver = 1, nio_groups
    if(poll_servers) then
       ! We're using server polling mode, so we must call
       ! *_find_server to receive the mpi_ssend sent by the servers,
       ! otherwise WRF will hang at the mpi_x_reduce below.

       call wrf_quilt_find_server(actual_iserver)

       ! The actual_iserver is now set to the next available I/O server.
       ! That may not be the same as iserver, but that's okay as long
       ! as we run through this loop exactly nio_groups times.
    else
       ! Not using server polling, so just access servers in numeric order.
       actual_iserver=iserver
    endif

    CALL get_mpi_comm_io_groups( comm_io_group , actual_iserver )

    CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )
    CALL mpi_comm_rank( comm_io_group , me , ierr )

! BY SENDING A NEGATIVE SIZE WE GET THE SERVERS TO SHUT DOWN
    hdrbufsize = -100 
    reduced = 0
    IF ( me .eq. 0 ) reduced(1) = hdrbufsize 
    CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )

  ENDDO
  Status = 0

#endif
  RETURN  
END SUBROUTINE wrf_quilt_ioexit

SUBROUTINE wrf_quilt_get_next_time ( DataHandle, DateStr, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to return the next time stamp.
! This is not yet supported.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && ! defined (STUBMPI) 
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*)               :: DateStr
  INTEGER                     :: Status
#endif
  RETURN
END SUBROUTINE wrf_quilt_get_next_time

SUBROUTINE wrf_quilt_get_previous_time ( DataHandle, DateStr, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to return the previous time stamp.
! This is not yet supported.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && ! defined (STUBMPI)
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*)               :: DateStr
  INTEGER                     :: Status
#endif
  RETURN
END SUBROUTINE wrf_quilt_get_previous_time

SUBROUTINE wrf_quilt_set_time ( DataHandle, Data,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to set the time stamp in the dataset
! referenced by DataHandle.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  USE module_state_description, ONLY: IO_PNETCDF
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Data
  INTEGER                     :: Status
  INTEGER i, itypesize, tasks_in_group, ierr, comm_io_group
  REAL dummy
  INTEGER                 :: Count
  INTEGER, EXTERNAL       :: use_package
!
  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_set_time' )

  IF ( DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles ) THEN
    IF ( int_handle_in_use( DataHandle ) ) THEN
      CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )
      Count = 0   ! there is no count for character strings

!ARPDBG - potential bug. Have no access to what type of IO is being used for
! this data so if PNETCDF_QUILT is defined then we assume that's what's being used.
#ifdef PNETCDF_QUILT
      IF(compute_group_master(1) )THEN
! Only want to send one time header to each IO server as 
! can't tell that's what they are on the IO servers themselves - therefore use
! the compute_group_master process.
         CALL int_gen_ti_header_char( hdrbuf, hdrbufsize, itypesize, &
                                      DataHandle, "TIMESTAMP", "", Data, int_set_time )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      END IF
#else
      IF ( wrf_dm_on_monitor() ) THEN
         CALL int_gen_ti_header_char( hdrbuf, hdrbufsize, itypesize, &
                                      DataHandle, "TIMESTAMP", "", Data, int_set_time )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#endif

      iserver = get_server_id ( DataHandle )
      CALL get_mpi_comm_io_groups( comm_io_group , iserver )
      CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )

      ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
      reduced = 0
      reduced(1) = hdrbufsize 
#ifdef PNETCDF_QUILT
      IF ( compute_group_master(1) ) reduced(2) = DataHandle
#else
      IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
      CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )
      ! send data to the i/o processor
      CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                            onebyte,                       &
                            hdrbuf, hdrbufsize , &
                            dummy, 0 )
    ENDIF
  ENDIF

#endif
RETURN
END SUBROUTINE wrf_quilt_set_time

SUBROUTINE wrf_quilt_get_next_var ( DataHandle, VarName, Status )
!<DESCRIPTION>
! When reading, instruct the I/O quilt servers to return the name of the next
! variable in the current time frame.
! This is not yet supported.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*)               :: VarName
  INTEGER                     :: Status
#endif
  RETURN
END SUBROUTINE wrf_quilt_get_next_var

SUBROUTINE wrf_quilt_get_dom_ti_real ( DataHandle,Element,   Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! independent domain metadata named "Element"
! from the open dataset described by DataHandle.
! Metadata of type real are
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  

! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  REAL,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Outcount
  INTEGER                     :: Status
  CALL wrf_message('wrf_quilt_get_dom_ti_real not supported yet')
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_ti_real 

SUBROUTINE wrf_quilt_put_dom_ti_real ( DataHandle,Element,   Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time independent
! domain metadata named "Element"
! to the open dataset described by DataHandle.
! Metadata of type real are
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  REAL ,          INTENT(IN)  :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
!Local
  CHARACTER*132   :: locElement
  INTEGER i, typesize, itypesize, tasks_in_group, ierr, comm_io_group
  REAL dummy
!
!!JMTIMING  CALL start_timing
  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_put_dom_ti_real' ) 
  CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )
  locElement = Element

  IF ( DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles ) THEN
    IF ( int_handle_in_use( DataHandle ) ) THEN
      CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )
      CALL MPI_TYPE_SIZE( MPI_REAL, typesize, ierr )

#ifdef PNETCDF_QUILT
      IF ( compute_group_master(1) ) THEN
         CALL int_gen_ti_header( hdrbuf, hdrbufsize, itypesize, typesize, &
                                 DataHandle, locElement, Data, Count, int_dom_ti_real )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#else
      IF ( wrf_dm_on_monitor() ) THEN
         CALL int_gen_ti_header( hdrbuf, hdrbufsize, itypesize, typesize, &
                                 DataHandle, locElement, Data, Count, int_dom_ti_real )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#endif

      iserver = get_server_id ( DataHandle )
      CALL get_mpi_comm_io_groups( comm_io_group , iserver )
      CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )

!!JMTIMING      CALL start_timing
      ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
      reduced = 0
      reduced(1) = hdrbufsize 
#ifdef PNETCDF_QUILT
      IF( compute_group_master(1) )  reduced(2) = DataHandle
#else
      IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
      CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )
!!JMTIMING       CALL end_timing("MPI_Reduce in wrf_quilt_put_dom_ti_real")
      ! send data to the i/o processor
      CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                            onebyte,                       &
                            hdrbuf, hdrbufsize , &
                            dummy, 0 )
    ENDIF
  ENDIF

  Status = 0
!!JMTIMING   CALL end_timing("wrf_quilt_put_dom_ti_real")
#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_ti_real 

SUBROUTINE wrf_quilt_get_dom_ti_double ( DataHandle,Element,   Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! independent domain metadata named "Element"
! from the open dataset described by DataHandle.
! Metadata of type double are
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  real*8                      :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
  CALL wrf_error_fatal('wrf_quilt_get_dom_ti_double not supported yet')
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_ti_double 

SUBROUTINE wrf_quilt_put_dom_ti_double ( DataHandle,Element,   Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time independent
! domain metadata named "Element"
! to the open dataset described by DataHandle.
! Metadata of type double are
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  REAL*8 ,        INTENT(IN)  :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
  CALL wrf_error_fatal('wrf_quilt_put_dom_ti_double not supported yet')
#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_ti_double 

SUBROUTINE wrf_quilt_get_dom_ti_integer ( DataHandle,Element,   Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! independent domain metadata named "Element"
! from the open dataset described by DataHandle.
! Metadata of type integer are
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  integer                     :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                      :: OutCount
  INTEGER                     :: Status
  CALL wrf_message('wrf_quilt_get_dom_ti_integer not supported yet')
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_ti_integer 

SUBROUTINE wrf_quilt_put_dom_ti_integer ( DataHandle,Element,   Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time independent
! domain metadata named "Element"
! to the open dataset described by DataHandle.
! Metadata of type integer are
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  USE module_state_description, ONLY: IO_PNETCDF
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  INTEGER ,       INTENT(IN)  :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
! Local
  CHARACTER*132   :: locElement
  INTEGER i, typesize, itypesize, tasks_in_group, ierr, comm_io_group
  REAL dummy
  INTEGER, EXTERNAL :: use_package
!

!!JMTIMING  CALL start_timing
  locElement = Element

  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_put_dom_ti_integer' ) 

  IF ( DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles ) THEN
    IF ( int_handle_in_use( DataHandle ) ) THEN
      CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )
      CALL MPI_TYPE_SIZE( MPI_INTEGER, typesize, ierr )

!ARPDBG - potential bug. Have no access to what type of IO is being used for
! this data so if PNETCDF_QUILT is defined then we assume that's what's being used.
#ifdef PNETCDF_QUILT
      IF ( compute_group_master(1) )THEN
         CALL int_gen_ti_header( hdrbuf, hdrbufsize, itypesize, typesize, &
                                 DataHandle, locElement, Data, Count,     &
                                 int_dom_ti_integer )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#else
      IF ( wrf_dm_on_monitor() ) THEN
         CALL int_gen_ti_header( hdrbuf, hdrbufsize, itypesize, typesize, &
                                 DataHandle, locElement, Data, Count,     &
                                 int_dom_ti_integer )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#endif

      iserver = get_server_id ( DataHandle )
      CALL get_mpi_comm_io_groups( comm_io_group , iserver )
      CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )

!!JMTIMING      CALL start_timing
      ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
      reduced = 0
      reduced(1) = hdrbufsize 
#ifdef PNETCDF_QUILT
      IF ( compute_group_master(1) ) reduced(2) = DataHandle
#else
      IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
      CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )

!!JMTIMING       CALL end_timing("MPI_Reduce in wrf_quilt_put_dom_ti_integer")
      ! send data to the i/o processor
      CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                            onebyte,                       &
                            hdrbuf, hdrbufsize , &
                            dummy, 0 )
    ENDIF
  ENDIF
  CALL wrf_debug ( DEBUG_LVL, 'returning from wrf_quilt_put_dom_ti_integer' ) 
!!JMTIMING   CALL end_timing("wrf_quilt_put_dom_ti_integer" )

#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_ti_integer 

SUBROUTINE wrf_quilt_get_dom_ti_logical ( DataHandle,Element,   Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! independent domain metadata named "Element"
! from the open dataset described by DataHandle.
! Metadata of type logical are
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  logical                     :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                      :: OutCount
  INTEGER                     :: Status
!  CALL wrf_message('wrf_quilt_get_dom_ti_logical not supported yet')
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_ti_logical 

SUBROUTINE wrf_quilt_put_dom_ti_logical ( DataHandle,Element,   Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time independent
! domain metadata named "Element"
! to the open dataset described by DataHandle.
! Metadata of type logical are
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  logical ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
! Local
  INTEGER i
  INTEGER one_or_zero(Count)

  DO i = 1, Count
    IF ( Data(i) ) THEN
      one_or_zero(i) = 1
    ELSE
      one_or_zero(i) = 0
    ENDIF
  ENDDO

  CALL wrf_quilt_put_dom_ti_integer ( DataHandle,Element,   one_or_zero, Count,  Status )
#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_ti_logical 

SUBROUTINE wrf_quilt_get_dom_ti_char ( DataHandle,Element,   Data,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read time independent
! domain metadata named "Element"
! from the open dataset described by DataHandle.
! Metadata of type char are
! stored in string Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*)               :: Data
  INTEGER                     :: Status
  CALL wrf_message('wrf_quilt_get_dom_ti_char not supported yet')
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_ti_char 

SUBROUTINE wrf_quilt_put_dom_ti_char ( DataHandle, Element,  Data,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write time independent
! domain metadata named "Element"
! to the open dataset described by DataHandle.
! Metadata of type char are
! copied from string Data.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: Data
  INTEGER                     :: Status
  INTEGER i, itypesize, tasks_in_group, ierr, comm_io_group, me
  REAL dummy
!
!!JMTIMING  CALL start_timing
  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_put_dom_ti_char' ) 

  IF ( DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles ) THEN
    IF ( int_handle_in_use( DataHandle ) ) THEN
      CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )

!ARPDBG - potential bug. Have no access to what type of IO is being used for
! this data so if PNETCDF_QUILT is defined then we assume that's what's being used.
#ifdef PNETCDF_QUILT
      IF(compute_group_master(1))THEN
         CALL int_gen_ti_header_char( hdrbuf, hdrbufsize, itypesize, &
                                      DataHandle, Element, "", Data, &
                                      int_dom_ti_char )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      END IF
#else
      IF ( wrf_dm_on_monitor() ) THEN
         CALL int_gen_ti_header_char( hdrbuf, hdrbufsize, itypesize, &
                                      DataHandle, Element, "", Data, int_dom_ti_char )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#endif

      iserver = get_server_id ( DataHandle )
!  write(0,*)'wrf_quilt_put_dom_ti_char ',iserver
      CALL get_mpi_comm_io_groups( comm_io_group , iserver )
      CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )
      ! send the size of my local buffer to the i/o task (obufsize doesnt mean anything here)
!!JMTIMING!  CALL start_timing
!write(0,*)'calling MPI_Barrier'
!  CALL MPI_Barrier( mpi_comm_local, ierr )
!write(0,*)'back from MPI_Barrier'
!!JMTIMING!   CALL end_timing("MPI_Barrier in wrf_quilt_put_dom_ti_char")

!!JMTIMING      CALL start_timing
      ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
      reduced_dummy = 0 
      reduced = 0
      reduced(1) = hdrbufsize 
#ifdef PNETCDF_QUILT
      IF(compute_group_master(1))    reduced(2) = DataHandle
#else
      IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
!call mpi_comm_rank( comm_io_group , me, ierr )

      CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )

!!JMTIMING       CALL end_timing("MPI_Reduce in wrf_quilt_put_dom_ti_char")
      ! send data to the i/o processor
!!JMTIMING  CALL start_timing

      CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                            onebyte,                       &
                            hdrbuf, hdrbufsize , &
                            dummy, 0 )
!!JMTIMING   CALL end_timing("collect_on_comm in wrf_quilt_put_dom_ti_char")
    ENDIF
  ENDIF
!!JMTIMING   CALL end_timing("wrf_quilt_put_dom_ti_char")

#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_ti_char 

SUBROUTINE wrf_quilt_get_dom_td_real ( DataHandle,Element, DateStr,  Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! dependent domain metadata named "Element" valid at time DateStr
! from the open dataset described by DataHandle.
! Metadata of type real are
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  real                        :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_td_real 

SUBROUTINE wrf_quilt_put_dom_td_real ( DataHandle,Element, DateStr,  Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time dependent
! domain metadata named "Element" valid at time DateStr
! to the open dataset described by DataHandle.
! Metadata of type real are
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  real ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_td_real 

SUBROUTINE wrf_quilt_get_dom_td_double ( DataHandle,Element, DateStr,  Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! dependent domain metadata named "Element" valid at time DateStr
! from the open dataset described by DataHandle.
! Metadata of type double are
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  real*8                          :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                      :: OutCount
  INTEGER                     :: Status
#endif
  CALL wrf_error_fatal('wrf_quilt_get_dom_td_double not supported yet')
RETURN
END SUBROUTINE wrf_quilt_get_dom_td_double 

SUBROUTINE wrf_quilt_put_dom_td_double ( DataHandle,Element, DateStr,  Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time dependent
! domain metadata named "Element" valid at time DateStr
! to the open dataset described by DataHandle.
! Metadata of type double are
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  real*8 ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
  CALL wrf_error_fatal('wrf_quilt_put_dom_td_double not supported yet')
RETURN
END SUBROUTINE wrf_quilt_put_dom_td_double 

SUBROUTINE wrf_quilt_get_dom_td_integer ( DataHandle,Element, DateStr,  Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! dependent domain metadata named "Element" valid at time DateStr
! from the open dataset described by DataHandle.
! Metadata of type integer are
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  integer                          :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                      :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_td_integer 

SUBROUTINE wrf_quilt_put_dom_td_integer ( DataHandle,Element, DateStr,  Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time dependent
! domain metadata named "Element" valid at time DateStr
! to the open dataset described by DataHandle.
! Metadata of type integer are
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  integer ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_td_integer 

SUBROUTINE wrf_quilt_get_dom_td_logical ( DataHandle,Element, DateStr,  Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! dependent domain metadata named "Element" valid at time DateStr
! from the open dataset described by DataHandle.
! Metadata of type logical are
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  logical                          :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                      :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_td_logical 

SUBROUTINE wrf_quilt_put_dom_td_logical ( DataHandle,Element, DateStr,  Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time dependent
! domain metadata named "Element" valid at time DateStr
! to the open dataset described by DataHandle.
! Metadata of type logical are
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  logical ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_td_logical 

SUBROUTINE wrf_quilt_get_dom_td_char ( DataHandle,Element, DateStr,  Data,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read time dependent
! domain metadata named "Element" valid at time DateStr
! from the open dataset described by DataHandle.
! Metadata of type char are
! stored in string Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*)               :: Data
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_dom_td_char 

SUBROUTINE wrf_quilt_put_dom_td_char ( DataHandle,Element, DateStr,  Data,  Status )
!<DESCRIPTION>
! Instruct $he I/O quilt servers to write time dependent
! domain metadata named "Element" valid at time DateStr
! to the open dataset described by DataHandle.
! Metadata of type char are
! copied from string Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN) :: Data
  INTEGER                          :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_dom_td_char 

SUBROUTINE wrf_quilt_get_var_ti_real ( DataHandle,Element,  Varname, Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! independent attribute "Element" of variable "Varname"
! from the open dataset described by DataHandle.
! Attribute of type real is
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  real                          :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_ti_real 

SUBROUTINE wrf_quilt_put_var_ti_real ( DataHandle,Element,  Varname, Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time independent
! attribute "Element" of variable "Varname"
! to the open dataset described by DataHandle.
! Attribute of type real is
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  real ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_var_ti_real 

SUBROUTINE wrf_quilt_get_var_ti_double ( DataHandle,Element,  Varname, Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! independent attribute "Element" of variable "Varname"
! from the open dataset described by DataHandle.
! Attribute of type double is
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  real*8                      :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
#endif
  CALL wrf_error_fatal('wrf_quilt_get_var_ti_double not supported yet')
RETURN
END SUBROUTINE wrf_quilt_get_var_ti_double 

SUBROUTINE wrf_quilt_put_var_ti_double ( DataHandle,Element,  Varname, Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time independent
! attribute "Element" of variable "Varname"
! to the open dataset described by DataHandle.
! Attribute of type double is
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  real*8 ,        INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
  CALL wrf_error_fatal('wrf_quilt_put_var_ti_double not supported yet')
RETURN
END SUBROUTINE wrf_quilt_put_var_ti_double 

SUBROUTINE wrf_quilt_get_var_ti_integer ( DataHandle,Element,  Varname, Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! independent attribute "Element" of variable "Varname"
! from the open dataset described by DataHandle.
! Attribute of type integer is
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  integer                     :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_ti_integer 

SUBROUTINE wrf_quilt_put_var_ti_integer ( DataHandle,Element,  Varname, Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time independent
! attribute "Element" of variable "Varname"
! to the open dataset described by DataHandle.
! Attribute of type integer is
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  integer ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_var_ti_integer 

SUBROUTINE wrf_quilt_get_var_ti_logical ( DataHandle,Element,  Varname, Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! independent attribute "Element" of variable "Varname"
! from the open dataset described by DataHandle.
! Attribute of type logical is
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  logical                     :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_ti_logical 

SUBROUTINE wrf_quilt_put_var_ti_logical ( DataHandle,Element,  Varname, Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time independent
! attribute "Element" of variable "Varname"
! to the open dataset described by DataHandle.
! Attribute of type logical is
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  logical ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_var_ti_logical 

SUBROUTINE wrf_quilt_get_var_ti_char ( DataHandle,Element,  Varname, Data,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read time independent
! attribute "Element" of variable "Varname"
! from the open dataset described by DataHandle.
! Attribute of type char is
! stored in string Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  CHARACTER*(*)               :: Data
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_ti_char 

SUBROUTINE wrf_quilt_put_var_ti_char ( DataHandle,Element,  Varname, Data,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write time independent
! attribute "Element" of variable "Varname"
! to the open dataset described by DataHandle.
! Attribute of type char is
! copied from string Data.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>

#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_wrf_quilt
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "intio_tags.h"
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  CHARACTER*(*) , INTENT(IN)  :: Data
  INTEGER                     :: Status
  INTEGER i, itypesize, tasks_in_group, ierr, comm_io_group
  REAL dummy
!

!!JMTIMING  CALL start_timing
  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_put_var_ti_char' ) 

  IF ( DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles ) THEN
    IF ( int_handle_in_use( DataHandle ) ) THEN
      CALL MPI_TYPE_SIZE( MPI_INTEGER, itypesize, ierr )

#ifdef PNETCDF_QUILT
      IF ( compute_group_master(1) ) THEN
         CALL int_gen_ti_header_char( hdrbuf, hdrbufsize, itypesize, &
                                      DataHandle, TRIM(Element),     &
                                      TRIM(VarName), TRIM(Data), int_var_ti_char )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#else
      IF ( wrf_dm_on_monitor() ) THEN
         CALL int_gen_ti_header_char( hdrbuf, hdrbufsize, itypesize, &
                                      DataHandle, TRIM(Element),     &
                                      TRIM(VarName), TRIM(Data), int_var_ti_char )
      ELSE
         CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#endif

      iserver = get_server_id ( DataHandle )
      CALL get_mpi_comm_io_groups( comm_io_group , iserver )
      CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )

!!JMTIMING      CALL start_timing
      ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
      reduced = 0
      reduced(1) = hdrbufsize 
#ifdef PNETCDF_QUILT
      IF ( compute_group_master(1) ) reduced(2) = DataHandle
#else
      IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
      CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )
!!JMTIMING       CALL end_timing("MPI_Reduce in wrf_quilt_put_var_ti_char")
      ! send data to the i/o processor
      CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,            &
                            onebyte,                       &
                            hdrbuf, hdrbufsize , &
                            dummy, 0 )
    ENDIF
  ENDIF
!!JMTIMING   CALL end_timing("wrf_quilt_put_dom_ti_char" )

#endif
RETURN
END SUBROUTINE wrf_quilt_put_var_ti_char 

SUBROUTINE wrf_quilt_get_var_td_real ( DataHandle,Element,  DateStr,Varname, Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! dependent attribute "Element" of variable "Varname" valid at time DateStr
! from the open dataset described by DataHandle.
! Attribute of type real is
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  real                        :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_td_real 

SUBROUTINE wrf_quilt_put_var_td_real ( DataHandle,Element,  DateStr,Varname, Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time dependent
! attribute "Element" of variable "Varname" valid at time DateStr
! to the open dataset described by DataHandle.
! Attribute of type real is
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  real ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_var_td_real 

SUBROUTINE wrf_quilt_get_var_td_double ( DataHandle,Element,  DateStr,Varname, Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! dependent attribute "Element" of variable "Varname" valid at time DateStr
! from the open dataset described by DataHandle.
! Attribute of type double is
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  real*8                      :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
#endif
  CALL wrf_error_fatal('wrf_quilt_get_var_td_double not supported yet')
RETURN
END SUBROUTINE wrf_quilt_get_var_td_double 

SUBROUTINE wrf_quilt_put_var_td_double ( DataHandle,Element,  DateStr,Varname, Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time dependent
! attribute "Element" of variable "Varname" valid at time DateStr
! to the open dataset described by DataHandle.
! Attribute of type double is
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  real*8 ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
  CALL wrf_error_fatal('wrf_quilt_put_var_td_double not supported yet')
RETURN
END SUBROUTINE wrf_quilt_put_var_td_double 

SUBROUTINE wrf_quilt_get_var_td_integer ( DataHandle,Element,  DateStr,Varname, Data, Count, Outcount,Status)
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! dependent attribute "Element" of variable "Varname" valid at time DateStr
! from the open dataset described by DataHandle.
! Attribute of type integer is
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  integer                     :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_td_integer 

SUBROUTINE wrf_quilt_put_var_td_integer ( DataHandle,Element,  DateStr,Varname, Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time dependent
! attribute "Element" of variable "Varname" valid at time DateStr
! to the open dataset described by DataHandle.
! Attribute of type integer is
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  integer ,       INTENT(IN)  :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_var_td_integer 

SUBROUTINE wrf_quilt_get_var_td_logical ( DataHandle,Element,  DateStr,Varname, Data, Count, Outcount, Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read Count words of time
! dependent attribute "Element" of variable "Varname" valid at time DateStr
! from the open dataset described by DataHandle.
! Attribute of type logical is
! stored in array Data.
! Actual number of words read is returned in OutCount.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  logical                          :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                      :: OutCount
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_td_logical 

SUBROUTINE wrf_quilt_put_var_td_logical ( DataHandle,Element,  DateStr,Varname, Data, Count,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write Count words of time dependent
! attribute "Element" of variable "Varname" valid at time DateStr
! to the open dataset described by DataHandle.
! Attribute of type logical is
! copied from array Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  logical ,            INTENT(IN) :: Data(*)
  INTEGER ,       INTENT(IN)  :: Count
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_var_td_logical 

SUBROUTINE wrf_quilt_get_var_td_char ( DataHandle,Element,  DateStr,Varname, Data,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to attempt to read time dependent
! attribute "Element" of variable "Varname" valid at time DateStr
! from the open dataset described by DataHandle.
! Attribute of type char is
! stored in string Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  CHARACTER*(*)               :: Data
  INTEGER                     :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_td_char 

SUBROUTINE wrf_quilt_put_var_td_char ( DataHandle,Element,  DateStr,Varname, Data,  Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to write time dependent
! attribute "Element" of variable "Varname" valid at time DateStr
! to the open dataset described by DataHandle.
! Attribute of type char is
! copied from string Data.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)  :: DataHandle
  CHARACTER*(*) , INTENT(IN)  :: Element
  CHARACTER*(*) , INTENT(IN)  :: DateStr
  CHARACTER*(*) , INTENT(IN)  :: VarName 
  CHARACTER*(*) , INTENT(IN) :: Data
  INTEGER                    :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_put_var_td_char 

SUBROUTINE wrf_quilt_read_field ( DataHandle , DateStr , VarName , Field , FieldType , Comm , IOComm, &
                            DomainDesc , MemoryOrder , Stagger , DimNames ,              &
                            DomainStart , DomainEnd ,                                    &
                            MemoryStart , MemoryEnd ,                                    &
                            PatchStart , PatchEnd ,                                      &
                            Status )
!<DESCRIPTION>
! Instruct the I/O quilt servers to read the variable named VarName from the
! dataset pointed to by DataHandle.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  INTEGER ,       INTENT(IN)    :: DataHandle 
  CHARACTER*(*) , INTENT(INOUT) :: DateStr
  CHARACTER*(*) , INTENT(INOUT) :: VarName
  INTEGER ,       INTENT(INOUT) :: Field(*)
  integer                       ,intent(in)    :: FieldType
  integer                       ,intent(inout) :: Comm
  integer                       ,intent(inout) :: IOComm
  integer                       ,intent(in)    :: DomainDesc
  character*(*)                 ,intent(in)    :: MemoryOrder
  character*(*)                 ,intent(in)    :: Stagger
  character*(*) , dimension (*) ,intent(in)    :: DimNames
  integer ,dimension(*)         ,intent(in)    :: DomainStart, DomainEnd
  integer ,dimension(*)         ,intent(in)    :: MemoryStart, MemoryEnd
  integer ,dimension(*)         ,intent(in)    :: PatchStart,  PatchEnd
  integer                       ,intent(out)   :: Status
  Status = 0
#endif
RETURN
END SUBROUTINE wrf_quilt_read_field

SUBROUTINE wrf_quilt_write_field ( DataHandle , DateStr , VarName , Field , FieldType , Comm , IOComm,  &
                             DomainDesc , MemoryOrder , Stagger , DimNames ,              &
                             DomainStart , DomainEnd ,                                    &
                             MemoryStart , MemoryEnd ,                                    &
                             PatchStart , PatchEnd ,                                      &
                             Status )
!<DESCRIPTION>
! Prepare instructions for the I/O quilt servers to write the variable named
! VarName to the dataset pointed to by DataHandle.
!
! During a "training" write this routine accumulates number and sizes of
! messages that will be sent to the I/O server associated with this compute
! (client) task.
!
! During a "real" write, this routine begins by allocating
! int_local_output_buffer if it has not already been allocated.  Sizes
! accumulated during "training" are used to determine how big
! int_local_output_buffer must be.  This routine then stores "int_field"
! headers and associated field data in int_local_output_buffer.  The contents
! of int_local_output_buffer are actually sent to the I/O quilt server in
! routine wrf_quilt_iosync().  This scheme allows output of multiple variables
! to be aggregated into a single "iosync" operation.
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  USE module_state_description
  USE module_wrf_quilt
  IMPLICIT NONE
  INCLUDE 'mpif.h'
#include "wrf_io_flags.h"
  INTEGER ,       INTENT(IN)    :: DataHandle 
  CHARACTER*(*) , INTENT(IN)    :: DateStr
  CHARACTER*(*) , INTENT(IN)    :: VarName
!  INTEGER ,       INTENT(IN)    :: Field(*)
  integer                       ,intent(in)    :: FieldType
  integer                       ,intent(inout) :: Comm
  integer                       ,intent(inout) :: IOComm
  integer                       ,intent(in)    :: DomainDesc
  character*(*)                 ,intent(in)    :: MemoryOrder
  character*(*)                 ,intent(in)    :: Stagger
  character*(*) , dimension (*) ,intent(in)    :: DimNames
  integer ,dimension(*)         ,intent(in)    :: DomainStart, DomainEnd
  integer ,dimension(*)         ,intent(in)    :: MemoryStart, MemoryEnd
  integer ,dimension(*)         ,intent(in)    :: PatchStart,  PatchEnd
  integer                       ,intent(out)   :: Status

  integer ii,jj,kk,myrank

  REAL, DIMENSION( MemoryStart(1):MemoryEnd(1), &
                   MemoryStart(2):MemoryEnd(2), &
                   MemoryStart(3):MemoryEnd(3) ) :: Field
  INTEGER locsize , typesize, itypesize
  INTEGER ierr, tasks_in_group, comm_io_group, dummy, i
  INTEGER, EXTERNAL :: use_package

!!ARPTIMING  CALL start_timing
  CALL wrf_debug ( DEBUG_LVL, 'in wrf_quilt_write_field' ) 

  IF ( .NOT. (DataHandle .GE. 1 .AND. DataHandle .LE. int_num_handles) ) THEN
    CALL wrf_error_fatal("frame/module_io_quilt.F: wrf_quilt_write_field: invalid data handle" )
  ENDIF
  IF ( .NOT. int_handle_in_use( DataHandle ) ) THEN
    CALL wrf_error_fatal("frame/module_io_quilt.F: wrf_quilt_write_field: DataHandle not opened" )
  ENDIF

  locsize = (PatchEnd(1)-PatchStart(1)+1)* &
            (PatchEnd(2)-PatchStart(2)+1)* &
            (PatchEnd(3)-PatchStart(3)+1)

  CALL mpi_type_size( MPI_INTEGER, itypesize, ierr )
  ! Note that the WRF_DOUBLE branch of this IF statement must come first since 
  ! WRF_FLOAT is set equal to WRF_DOUBLE during autopromotion builds.  
  IF ( FieldType .EQ. WRF_DOUBLE ) THEN
    CALL mpi_type_size( MPI_DOUBLE_PRECISION, typesize, ierr )
  ELSE IF ( FieldType .EQ. WRF_FLOAT ) THEN
    CALL mpi_type_size( MPI_REAL, typesize, ierr )
  ELSE IF ( FieldType .EQ. WRF_INTEGER ) THEN
    CALL mpi_type_size( MPI_INTEGER, typesize, ierr )
  ELSE IF ( FieldType .EQ. WRF_LOGICAL ) THEN
    CALL mpi_type_size( MPI_LOGICAL, typesize, ierr )
  ENDIF

  IF ( .NOT. okay_to_write( DataHandle ) ) THEN

      ! This is a "training" write.
      ! it is not okay to actually write; what we do here is just "bookkeep": count up
      ! the number and size of messages that we will output to io server associated with
      ! this task

      CALL int_gen_write_field_header ( hdrbuf, hdrbufsize, itypesize, typesize,           &
                               DataHandle , DateStr , VarName , Field , FieldType , Comm , IOComm,  &
                               333933         , MemoryOrder , Stagger , DimNames ,              &   ! 333933 means training; magic number
                               DomainStart , DomainEnd ,                                    &
                               MemoryStart , MemoryEnd ,                                    &
                               PatchStart , PatchEnd )

      int_num_bytes_to_write(DataHandle) = int_num_bytes_to_write(DataHandle) + locsize * typesize + hdrbufsize

      ! Send the hdr for the write in case the interface is calling the I/O API in "learn" mode

      iserver = get_server_id ( DataHandle )
!JMDEBUGwrite(0,*)'wrf_quilt_write_field (dryrun) ',iserver
      CALL get_mpi_comm_io_groups( comm_io_group , iserver )
      ! send the size of my local buffer to the i/o task (obufsize doesnt mean anything here)

      CALL mpi_x_comm_size( comm_io_group , tasks_in_group , ierr )

#if 0
      IF ( .NOT. wrf_dm_on_monitor() ) THEN     ! only one task in compute grid sends this message; send noops on others
        CALL int_gen_noop_header( hdrbuf, hdrbufsize, itypesize )
      ENDIF
#endif


!!ARPTIMING      CALL start_timing
      ! send the size of my local buffer to the i/o task (reduced_dummy doesnt mean anything on client side)
      reduced = 0
      reduced(1) = hdrbufsize 
#ifdef PNETCDF_QUILT
      IF ( compute_group_master(1) ) reduced(2) = DataHandle
#else
      IF ( wrf_dm_on_monitor() )  reduced(2) = DataHandle
#endif
      CALL mpi_x_reduce( reduced, reduced_dummy, 2, MPI_INTEGER, MPI_SUM, tasks_in_group-1, comm_io_group, ierr )
!!ARPTIMING      CALL end_timing("MPI_Reduce in wrf_quilt_write_field dryrun")
      ! send data to the i/o processor

      CALL collect_on_comm_debug(__FILE__,__LINE__, comm_io_group,                   &
                            onebyte,                          &
                            hdrbuf, hdrbufsize ,                 &
                            dummy, 0 )

  ELSE

    IF ( .NOT. associated( int_local_output_buffer ) ) THEN
      ALLOCATE ( int_local_output_buffer( (int_num_bytes_to_write( DataHandle )+1)/itypesize ), Stat=ierr )
      IF(ierr /= 0)THEN
         CALL wrf_error_fatal("frame/module_io_quilt.F: wrf_quilt_write_field: allocate of int_local_output_buffer failed" )
      END IF
      int_local_output_cursor = 1
    ENDIF
      iserver = get_server_id ( DataHandle )
!JMDEBUGwrite(0,*)'wrf_quilt_write_field (writing) ',iserver

    ! This is NOT a "training" write.  It is OK to write now.
    CALL int_gen_write_field_header ( hdrbuf, hdrbufsize, itypesize, typesize,           &
                             DataHandle , DateStr , VarName , Field , FieldType , Comm , IOComm,  &
                             0          , MemoryOrder , Stagger , DimNames ,              &   ! non-333933 means okay to write; magic number
                             DomainStart , DomainEnd ,                                    &
                             MemoryStart , MemoryEnd ,                                    &
                             PatchStart , PatchEnd )

    ! Pack header into int_local_output_buffer.  It will be sent to the 
    ! I/O servers during the next "iosync" operation.  
#ifdef DEREF_KLUDGE
    CALL int_pack_data ( hdrbuf , hdrbufsize , int_local_output_buffer(1), int_local_output_cursor )
#else
    CALL int_pack_data ( hdrbuf , hdrbufsize , int_local_output_buffer, int_local_output_cursor )
#endif

    ! Pack field data into int_local_output_buffer.  It will be sent to the 
    ! I/O servers during the next "iosync" operation.  
#ifdef DEREF_KLUDGE
    CALL int_pack_data ( Field(PatchStart(1):PatchEnd(1),PatchStart(2):PatchEnd(2),PatchStart(3):PatchEnd(3) ), &
                                  locsize * typesize , int_local_output_buffer(1), int_local_output_cursor )
#else
    CALL int_pack_data ( Field(PatchStart(1):PatchEnd(1),PatchStart(2):PatchEnd(2),PatchStart(3):PatchEnd(3) ), &
                                  locsize * typesize , int_local_output_buffer, int_local_output_cursor )
#endif

  ENDIF
  Status = 0
!!ARPTIMING  CALL end_timing("wrf_quilt_write_field")

#endif
  RETURN
END SUBROUTINE wrf_quilt_write_field

SUBROUTINE wrf_quilt_get_var_info ( DataHandle , VarName , NDim , MemoryOrder , Stagger , &
                              DomainStart , DomainEnd , Status )
!<DESCRIPTION>
! This routine applies only to a dataset that is open for read.  It instructs
! the I/O quilt servers to return information about variable VarName.
! This routine is called only by client (compute) tasks.  
!
! This is not yet supported.
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  IMPLICIT NONE
  integer               ,intent(in)     :: DataHandle
  character*(*)         ,intent(in)     :: VarName
  integer                               :: NDim
  character*(*)                         :: MemoryOrder
  character*(*)                         :: Stagger
  integer ,dimension(*)                 :: DomainStart, DomainEnd
  integer                               :: Status
#endif
RETURN
END SUBROUTINE wrf_quilt_get_var_info

subroutine wrf_quilt_find_server(iserver)

  ! This routine is called by the compute processes when they need an
  ! I/O server to write out a new file.  Upon return, this routine will
  ! set iserver to the next available I/O server group.  

  ! A mpi_recv to all of mpi_comm_avail is used to implement this, and
  ! that recv will not return until an I/O server group calls
  ! wrf_quilt_server_ready to signal that it is ready for a new file.

#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  use module_wrf_quilt, only : in_avail, mpi_comm_avail, mpi_comm_local

  implicit none
  INCLUDE 'mpif.h'
  integer, intent(inout) :: iserver
  integer :: ierr
  character(255) :: message

  call wrf_message('Polling I/O servers...')

  if(in_avail) then
     call mpi_recv(iserver,1,MPI_INTEGER,MPI_ANY_SOURCE,0,mpi_comm_avail,MPI_STATUS_IGNORE,ierr)
     if(ierr/=0) then
        call wrf_error_fatal('mpi_recv failed in wrf_quilt_find_server')
     endif
  endif

  call mpi_bcast(iserver,1,MPI_INTEGER,0,mpi_comm_local,ierr)
  if(ierr/=0) then
     call wrf_error_fatal('mpi_bcast failed in wrf_quilt_find_server')
  endif

  write(message,'("I/O server ",I0," is ready for operations.")') iserver
  call wrf_message(message)

#endif

end subroutine wrf_quilt_find_server
subroutine wrf_quilt_server_ready()

  ! This routine is called by the I/O server group's master process once the
  ! I/O server group is done writing its current file, and is waiting for
  ! a new one.  This information is passed to the monitor process by a 
  ! blocking send from the I/O server master process to the monitor.  

  ! All processes in an I/O group must call this routine, and this routine
  ! will not return (in any process) until the monitor process signals 
  ! that it wants the I/O server group to write a file.  That signal is
  ! sent in a call to wrf_quilt_find_server on the compute processes.

#if defined( DM_PARALLEL ) && !defined( STUBMPI )
  use module_wrf_quilt, only : mpi_comm_local, in_avail, availrank, mpi_comm_avail

  implicit none
  INCLUDE 'mpif.h'
  integer :: ierr

  write(0,*) 'Entering wrf_quilt_server_ready.'

  call mpi_barrier(mpi_comm_local,ierr)
  if(ierr/=0) then
     call wrf_error_fatal('mpi_barrier failed in wrf_quilt_server_ready')
  endif

  if(in_avail) then
     write(0,'("mpi_ssend ioserver=",I0," in wrf_quilt_server_ready")') availrank
     call mpi_ssend(availrank,1,MPI_INTEGER,0,0,mpi_comm_avail,ierr)
     if(ierr/=0) then
        call wrf_error_fatal('mpi_ssend failed in wrf_quilt_server_ready')
     endif
  endif

  call mpi_barrier(mpi_comm_local,ierr)
  if(ierr/=0) then
     call wrf_error_fatal('mpi_barrier failed in wrf_quilt_server_ready')
  endif

  write(0,*) 'Leaving wrf_quilt_server_ready.'
#endif

end subroutine wrf_quilt_server_ready

SUBROUTINE get_mpi_comm_io_groups( retval, isrvr )
!<DESCRIPTION>
! This routine returns the compute+io communicator to which this
! compute task belongs for I/O server group "isrvr".
! This routine is called only by client (compute) tasks.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
      USE module_wrf_quilt
      IMPLICIT NONE
      INTEGER, INTENT(IN ) :: isrvr
      INTEGER, INTENT(OUT) :: retval
      retval = mpi_comm_io_groups(isrvr)
#endif
      RETURN
END SUBROUTINE get_mpi_comm_io_groups

SUBROUTINE get_nio_tasks_in_group( retval )
!<DESCRIPTION>
! This routine returns the number of I/O server tasks in each 
! I/O server group.  It can be called by both clients and 
! servers.  
!</DESCRIPTION>
#if defined( DM_PARALLEL ) && !defined( STUBMPI )
      USE module_wrf_quilt
      IMPLICIT NONE
      INTEGER, INTENT(OUT) :: retval
      retval = nio_tasks_in_group
#endif
      RETURN
END SUBROUTINE get_nio_tasks_in_group

SUBROUTINE collect_on_comm_debug(file,line, comm_io_group,   &
                        sze,                                 &
                        hdrbuf, hdrbufsize ,                 &
                        outbuf, outbufsize                   )
  IMPLICIT NONE
  CHARACTER*(*) file
  INTEGER line
  INTEGER comm_io_group
  INTEGER sze
  INTEGER hdrbuf(*), outbuf(*)
  INTEGER hdrbufsize, outbufsize 

  !write(0,*)'collect_on_comm_debug ',trim(file),line,sze,hdrbufsize,outbufsize
  CALL collect_on_comm( comm_io_group,                       &
                        sze,                                 &
                        hdrbuf, hdrbufsize ,                 &
                        outbuf, outbufsize                   )
  !write(0,*)trim(file),line,'returning'
  RETURN
END


SUBROUTINE collect_on_comm_debug2(file,line,var,tag,sz,hdr_rec_size, &
                        comm_io_group,                       &
                        sze,                                 &
                        hdrbuf, hdrbufsize ,                 &
                        outbuf, outbufsize                   )
  IMPLICIT NONE
  CHARACTER*(*) file,var
  INTEGER line,tag,sz,hdr_rec_size
  INTEGER comm_io_group
  INTEGER sze
  INTEGER hdrbuf(*), outbuf(*)
  INTEGER hdrbufsize, outbufsize

!  write(0,*)'collect_on_comm_debug2 ',trim(file),line,trim(var),tag,sz,hdr_rec_size,sze,hdrbufsize,outbufsize
  CALL collect_on_comm( comm_io_group,                       &
                        sze,                                 &
                        hdrbuf, hdrbufsize ,                 &
                        outbuf, outbufsize                   )
!  write(0,*)'collect_on_comm_debug2 ',trim(file),line,'returning for ',trim(var)
  RETURN
END