doxygen-example/mpp__write_8h_source.html

 !***********************************************************************
 !*                   GNU Lesser General Public License
 !*
 !* This file is part of the GFDL Flexible Modeling System (FMS).
 !*
 !* FMS is free software: you can redistribute it and/or modify it under
 !* the terms of the GNU Lesser General Public License as published by
 !* the Free Software Foundation, either version 3 of the License, or (at
 !* your option) any later version.
 !*
 !* FMS is distributed in the hope that it will be useful, but WITHOUT
 !* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 !* FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 !* for more details.
 !*
 !* You should have received a copy of the GNU Lesser General Public
 !* License along with FMS.  If not, see <http://www.gnu.org/licenses/>.
 !***********************************************************************
 subroutine MPP_WRITE_( unit, field, data, tstamp)
       use mpp_parameter_mod, only : NULLUNIT
       integer, intent(in) :: unit
       type(fieldtype), intent(in) :: field
       MPP_TYPE_, intent(in) :: data MPP_RANK_
       MPP_TYPE_, intent(in), optional :: tstamp

       if (unit == NULLUNIT) return
       MPP_WRITE_RECORD_
       return
     end subroutine MPP_WRITE_
will
************************************************************************GNU Lesser General Public License **This file is part of the GFDL Flexible Modeling System(FMS). ! *! *FMS is free software without even the implied warranty of MERCHANTABILITY or *FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License *for more details **You should have received a copy of the GNU Lesser General Public *License along with FMS If see< http:! ***********************************************************************!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !! MPP_TRANSMIT !! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! subroutine MPP_TRANSMIT_(put_data, put_len, to_pe, get_data, get_len, from_pe, block, tag, recv_request, send_request)!a message-passing routine intended to be reminiscent equally of both MPI and SHMEM!put_data and get_data are contiguous MPP_TYPE_ arrays!at each call, your put_data array is put to to_pe 's get_data! your get_data array is got from from_pe 's put_data!i.e we assume that typically(e.g updating halo regions) each PE performs a put _and_ a get!special PE designations:! NULL_PE:to disable a put or a get(e.g at boundaries)! ANY_PE:if remote PE for the put or get is to be unspecific! ALL_PES:broadcast and collect operations(collect not yet implemented)!ideally we would not pass length, but this f77-style call performs better(arrays passed by address, not descriptor)!further, this permits< length > contiguous words from an array of any rank to be passed(avoiding f90 rank conformance check)!caller is responsible for completion checks(mpp_sync_self) before and after integer, intent(in) ::put_len, to_pe, get_len, from_pe MPP_TYPE_, intent(in) ::put_data(*) MPP_TYPE_, intent(out) ::get_data(*) logical, intent(in), optional ::block integer, intent(in), optional ::tag integer, intent(out), optional ::recv_request, send_request logical ::block_comm integer ::i MPP_TYPE_, allocatable, save ::local_data(:) !local copy used by non-parallel code(no SHMEM or MPI) integer ::comm_tag integer ::rsize if(.NOT.module_is_initialized) call mpp_error(FATAL, 'MPP_TRANSMIT:You must first call mpp_init.') if(to_pe.EQ.NULL_PE .AND. from_pe.EQ.NULL_PE) return block_comm=.true. if(PRESENT(block)) block_comm=block if(debug) then call SYSTEM_CLOCK(tick) write(stdout_unit,'(a, i18, a, i6, a, 2i6, 2i8)')&'T=', tick, ' PE=', pe, ' MPP_TRANSMIT begin:to_pe, from_pe, put_len, get_len=', to_pe, from_pe, put_len, get_len end if comm_tag=DEFAULT_TAG if(present(tag)) comm_tag=tag!do put first and then get if(to_pe.GE.0 .AND. to_pe.LT.npes) then!use non-blocking sends if(debug .and.(current_clock.NE.0)) call SYSTEM_CLOCK(start_tick)!z1l:truly non-blocking send.! if(request(to_pe).NE.MPI_REQUEST_NULL) then !only one message from pe-> to_pe in queue *PE waiting for to_pe ! call error else get_len so only do gets but you cannot have a pure get with MPI call a get means do a wait to ensure put on remote PE is complete error call increase mpp_nml request_multiply call MPP_TRANSMIT get_len end if return end subroutine MPP_TRANSMIT_ ! MPP_BROADCAST ! subroutine but that doesn t allow !broadcast to a subset of PEs This version will
Definition: mpp_transmit_mpi.h:166

time_interp_external_mod::field
type(ext_fieldtype), dimension(:), pointer, save, private field
Definition: time_interp_external.F90:137

mpp_parameter_mod
Definition: mpp_parameter.F90:19

integer
*f90 *************************************************************************GNU Lesser General Public License **This file is part of the GFDL Flexible Modeling System(FMS). ! *! *FMS is free software without even the implied warranty of MERCHANTABILITY or *FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License *for more details **You should have received a copy of the GNU Lesser General Public *License along with FMS If see< http:! ***********************************************************************subroutine MPP_UPDATE_NEST_FINE_2D_(field, nest_domain, wbuffer, ebuffer, sbuffer, nbuffer, &flags, complete, position, extra_halo, name, tile_count) MPP_TYPE_, intent(in) ::field(:,:) type(nest_domain_type), intent(inout) ::nest_domain MPP_TYPE_, intent(inout) ::wbuffer(:,:) MPP_TYPE_, intent(inout) ::ebuffer(:,:) MPP_TYPE_, intent(inout) ::sbuffer(:,:) MPP_TYPE_, intent(inout) ::nbuffer(:,:) integer, intent(in), optional ::flags logical, intent(in), optional ::complete integer, intent(in), optional ::position integer, intent(in), optional ::extra_halo character(len= *), intent(in), optional ::name integer, intent(in), optional ::tile_count MPP_TYPE_ ::field3D(size(field, 1), size(field, 2), 1) MPP_TYPE_ ::wbuffer3D(size(wbuffer, 1), size(wbuffer, 2), 1) MPP_TYPE_ ::ebuffer3D(size(ebuffer, 1), size(ebuffer, 2), 1) MPP_TYPE_ ::sbuffer3D(size(sbuffer, 1), size(sbuffer, 2), 1) MPP_TYPE_ ::nbuffer3D(size(nbuffer, 1), size(nbuffer, 2), 1) pointer(ptr, field3D) pointer(ptr_w, wbuffer3D) pointer(ptr_e, ebuffer3D) pointer(ptr_s, sbuffer3D) pointer(ptr_n, nbuffer3D) ptr=LOC(field) ptr_w=LOC(wbuffer) ptr_e=LOC(ebuffer) ptr_s=LOC(sbuffer) ptr_n=LOC(nbuffer) call mpp_update_nest_fine(field3D, nest_domain, wbuffer3D, ebuffer3D, sbuffer3D, nbuffer3D, &flags, complete, position, extra_halo, name, tile_count) returnend subroutine MPP_UPDATE_NEST_FINE_2D_subroutine MPP_UPDATE_NEST_FINE_3D_(field, nest_domain, wbuffer, sbuffer, ebuffer, nbuffer, &flags, complete, position, extra_halo, name, tile_count) MPP_TYPE_, intent(in) ::field(:,:,:) type(nest_domain_type), intent(inout) ::nest_domain MPP_TYPE_, intent(inout) ::wbuffer(:,:,:) MPP_TYPE_, intent(inout) ::ebuffer(:,:,:) MPP_TYPE_, intent(inout) ::sbuffer(:,:,:) MPP_TYPE_, intent(inout) ::nbuffer(:,:,:) integer, intent(in), optional ::flags logical, intent(in), optional ::complete integer, intent(in), optional ::position integer, intent(in), optional ::extra_halo character(len= *), intent(in), optional ::name integer, intent(in), optional ::tile_count MPP_TYPE_ ::d_type type(nestSpec), pointer ::update=> set_mismatch integer ::tile update_position nbuffersz l_size integer
Definition: mpp_update_nest_domains.h:79

unit
************************************************************************GNU Lesser General Public License **This file is part of the GFDL Flexible Modeling System(FMS). ! *! *FMS is free software without even the implied warranty of MERCHANTABILITY or *FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License *for more details **You should have received a copy of the GNU Lesser General Public *License along with FMS If see< http:! ***********************************************************************subroutine READ_RECORD_CORE_(unit, field, nwords, data, start, axsiz) integer, intent(in) ::unit type(fieldtype), intent(in) ::field integer, intent(in) ::nwords MPP_TYPE_, intent(inout) ::data(nwords) integer, intent(in) ::start(:), axsiz(:) integer(SHORT_KIND) ::i2vals(nwords)!rab used in conjunction with transfer intrinsic to determine size of a variable integer(KIND=1) ::one_byte(8) integer ::word_sz!#ifdef __sgi integer(INT_KIND) ::ivals(nwords) real(FLOAT_KIND) ::rvals(nwords)!#else! integer ::ivals(nwords)! real ::rvals(nwords)!#endif real(DOUBLE_KIND) ::r8vals(nwords) pointer(ptr1, i2vals) pointer(ptr2, ivals) pointer(ptr3, rvals) pointer(ptr4, r8vals) if(mpp_io_stack_size< nwords) call mpp_io_set_stack_size(nwords) call mpp_error(FATAL, 'MPP_READ currently requires use_netCDF option') end subroutine READ_RECORD_CORE_ subroutine READ_RECORD_(unit, field, nwords, data, time_level, domain, position, tile_count, start_in, axsiz_in)!routine that is finally called by all mpp_read routines to perform the read!a non-netCDF record contains:! field ID! a set of 4 coordinates(is:ie, js:je) giving the data subdomain! a timelevel and a timestamp(=NULLTIME if field is static)! 3D real data(stored as 1D)!if you are using direct access I/O, the RECL argument to OPEN must be large enough for the above!in a global direct access file, record position on PE is given by %record.!Treatment of timestamp:! We assume that static fields have been passed without a timestamp.! Here that is converted into a timestamp of NULLTIME.! For non-netCDF fields, field is treated no differently, but is written! with a timestamp of NULLTIME. There is no check in the code to prevent! the user from repeatedly writing a static field. integer, intent(in) ::unit, nwords type(fieldtype), intent(in) ::field MPP_TYPE_, intent(inout) ::data(nwords) integer, intent(in), optional ::time_level type(domain2D), intent(in), optional ::domain integer, intent(in), optional ::position, tile_count integer, intent(in), optional ::start_in(:), axsiz_in(:) integer, dimension(size(field%axes(:))) ::start, axsiz integer ::tlevel !, subdomain(4) integer ::i, error, is, ie, js, je, isg, ieg, jsg, jeg type(domain2d), pointer ::io_domain=> unit
Definition: mpp_read_2Ddecomp.h:167

fv3jedi_increment_mod::copy
subroutine, public copy(self, rhs)
Definition: fv3jedi_increment_mod.f90:254

end
************************************************************************GNU Lesser General Public License **This file is part of the GFDL Flexible Modeling System(FMS). ! *! *FMS is free software without even the implied warranty of MERCHANTABILITY or *FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License *for more details **You should have received a copy of the GNU Lesser General Public *License along with FMS If see< http:! ***********************************************************************!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !! MPP_TRANSMIT !! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! subroutine MPP_TRANSMIT_(put_data, put_len, to_pe, get_data, get_len, from_pe, block, tag, recv_request, send_request)!a message-passing routine intended to be reminiscent equally of both MPI and SHMEM!put_data and get_data are contiguous MPP_TYPE_ arrays!at each call, your put_data array is put to to_pe 's get_data! your get_data array is got from from_pe 's put_data!i.e we assume that typically(e.g updating halo regions) each PE performs a put _and_ a get!special PE designations:! NULL_PE:to disable a put or a get(e.g at boundaries)! ANY_PE:if remote PE for the put or get is to be unspecific! ALL_PES:broadcast and collect operations(collect not yet implemented)!ideally we would not pass length, but this f77-style call performs better(arrays passed by address, not descriptor)!further, this permits< length > contiguous words from an array of any rank to be passed(avoiding f90 rank conformance check)!caller is responsible for completion checks(mpp_sync_self) before and after integer, intent(in) ::put_len, to_pe, get_len, from_pe MPP_TYPE_, intent(in) ::put_data(*) MPP_TYPE_, intent(out) ::get_data(*) logical, intent(in), optional ::block integer, intent(in), optional ::tag integer, intent(out), optional ::recv_request, send_request logical ::block_comm integer ::i MPP_TYPE_, allocatable, save ::local_data(:) !local copy used by non-parallel code(no SHMEM or MPI) integer ::comm_tag integer ::rsize if(.NOT.module_is_initialized) call mpp_error(FATAL, 'MPP_TRANSMIT:You must first call mpp_init.') if(to_pe.EQ.NULL_PE .AND. from_pe.EQ.NULL_PE) return block_comm=.true. if(PRESENT(block)) block_comm=block if(debug) then call SYSTEM_CLOCK(tick) write(stdout_unit,'(a, i18, a, i6, a, 2i6, 2i8)')&'T=', tick, ' PE=', pe, ' MPP_TRANSMIT begin:to_pe, from_pe, put_len, get_len=', to_pe, from_pe, put_len, get_len end if comm_tag=DEFAULT_TAG if(present(tag)) comm_tag=tag!do put first and then get if(to_pe.GE.0 .AND. to_pe.LT.npes) then!use non-blocking sends if(debug .and.(current_clock.NE.0)) call SYSTEM_CLOCK(start_tick)!z1l:truly non-blocking send.! if(request(to_pe).NE.MPI_REQUEST_NULL) then !only one message from pe-> to_pe in queue *PE waiting for to_pe ! call error else get_len so only do gets but you cannot have a pure get with MPI call a get means do a wait to ensure put on remote PE is complete error call increase mpp_nml request_multiply call MPP_TRANSMIT end
Definition: mpp_transmit_mpi.h:153

MPP_TYPE_
************************************************************************GNU Lesser General Public License **This file is part of the GFDL Flexible Modeling System(FMS). ! *! *FMS is free software without even the implied warranty of MERCHANTABILITY or *FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License *for more details **You should have received a copy of the GNU Lesser General Public *License along with FMS If see< http:! ***********************************************************************!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !! MPP_TRANSMIT !! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! subroutine MPP_TRANSMIT_(put_data, put_len, to_pe, get_data, get_len, from_pe, block, tag, recv_request, send_request)!a message-passing routine intended to be reminiscent equally of both MPI and SHMEM!put_data and get_data are contiguous MPP_TYPE_ arrays!at each call, your put_data array is put to to_pe 's get_data! your get_data array is got from from_pe 's put_data!i.e we assume that typically(e.g updating halo regions) each PE performs a put _and_ a get!special PE designations:! NULL_PE:to disable a put or a get(e.g at boundaries)! ANY_PE:if remote PE for the put or get is to be unspecific! ALL_PES:broadcast and collect operations(collect not yet implemented)!ideally we would not pass length, but this f77-style call performs better(arrays passed by address, not descriptor)!further, this permits< length > contiguous words from an array of any rank to be passed(avoiding f90 rank conformance check)!caller is responsible for completion checks(mpp_sync_self) before and after integer, intent(in) ::put_len, to_pe, get_len, from_pe MPP_TYPE_, intent(in) ::put_data(*) MPP_TYPE_, intent(out) ::get_data(*) logical, intent(in), optional ::block integer, intent(in), optional ::tag integer, intent(out), optional ::recv_request, send_request logical ::block_comm integer ::i MPP_TYPE_, allocatable, save ::local_data(:) !local copy used by non-parallel code(no SHMEM or MPI) integer ::comm_tag integer ::rsize if(.NOT.module_is_initialized) call mpp_error(FATAL, 'MPP_TRANSMIT:You must first call mpp_init.') if(to_pe.EQ.NULL_PE .AND. from_pe.EQ.NULL_PE) return block_comm=.true. if(PRESENT(block)) block_comm=block if(debug) then call SYSTEM_CLOCK(tick) write(stdout_unit,'(a, i18, a, i6, a, 2i6, 2i8)')&'T=', tick, ' PE=', pe, ' MPP_TRANSMIT begin:to_pe, from_pe, put_len, get_len=', to_pe, from_pe, put_len, get_len end if comm_tag=DEFAULT_TAG if(present(tag)) comm_tag=tag!do put first and then get if(to_pe.GE.0 .AND. to_pe.LT.npes) then!use non-blocking sends if(debug .and.(current_clock.NE.0)) call SYSTEM_CLOCK(start_tick)!z1l:truly non-blocking send.! if(request(to_pe).NE.MPI_REQUEST_NULL) then !only one message from pe-> to_pe in queue *PE waiting for to_pe ! call error else get_len so only do gets but you cannot have a pure get with MPI call a get means do a wait to ensure put on remote PE is complete error call increase mpp_nml request_multiply call MPP_TRANSMIT get_len end if return end subroutine MPP_TRANSMIT_ ! MPP_BROADCAST ! subroutine but that doesn t allow !broadcast to a subset of PEs This version and mpp_transmit will remain !backward compatible MPP_TYPE_
Definition: mpp_transmit_mpi.h:166

if
l_size ! loop over number of fields ke do je do ie to je n if(.NOT. d_comm%R_do_buf(list)) cycle from_pe

diag_integral_mod::version
character(len=128) version
Definition: diag_integral.F90:121

is
l_size ! loop over number of fields ke do je do ie to is
Definition: mpp_do_redistribute.h:72

c2f.type
type
Definition: c2f.py:15

type_distribution::received
logical function received(this, seqno)
Definition: m_distribution.f90:72

a
************************************************************************GNU Lesser General Public License **This file is part of the GFDL Flexible Modeling System(FMS). ! *! *FMS is free software without even the implied warranty of MERCHANTABILITY or *FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License *for more details **You should have received a copy of the GNU Lesser General Public *License along with FMS If see< http:! ***********************************************************************!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! !! MPP_TRANSMIT !! !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! subroutine MPP_TRANSMIT_(put_data, put_len, to_pe, get_data, get_len, from_pe, block, tag, recv_request, send_request)!a message-passing routine intended to be reminiscent equally of both MPI and SHMEM!put_data and get_data are contiguous MPP_TYPE_ arrays!at each call, your put_data array is put to to_pe 's get_data! your get_data array is got from from_pe 's put_data!i.e we assume that typically(e.g updating halo regions) each PE performs a put _and_ a get!special PE designations:! NULL_PE:to disable a put or a get(e.g at boundaries)! ANY_PE:if remote PE for the put or get is to be unspecific! ALL_PES:broadcast and collect operations(collect not yet implemented)!ideally we would not pass length, but this f77-style call performs better(arrays passed by address, not descriptor)!further, this permits< length > contiguous words from an array of any rank to be passed(avoiding f90 rank conformance check)!caller is responsible for completion checks(mpp_sync_self) before and after integer, intent(in) ::put_len, to_pe, get_len, from_pe MPP_TYPE_, intent(in) ::put_data(*) MPP_TYPE_, intent(out) ::get_data(*) logical, intent(in), optional ::block integer, intent(in), optional ::tag integer, intent(out), optional ::recv_request, send_request logical ::block_comm integer ::i MPP_TYPE_, allocatable, save ::local_data(:) !local copy used by non-parallel code(no SHMEM or MPI) integer ::comm_tag integer ::rsize if(.NOT.module_is_initialized) call mpp_error(FATAL, 'MPP_TRANSMIT:You must first call mpp_init.') if(to_pe.EQ.NULL_PE .AND. from_pe.EQ.NULL_PE) return block_comm=.true. if(PRESENT(block)) block_comm=block if(debug) then call SYSTEM_CLOCK(tick) write(stdout_unit,'(a, i18, a, i6, a, 2i6, 2i8)')&'T=', tick, ' PE=', pe, ' MPP_TRANSMIT begin:to_pe, from_pe, put_len, get_len=', to_pe, from_pe, put_len, get_len end if comm_tag=DEFAULT_TAG if(present(tag)) comm_tag=tag!do put first and then get if(to_pe.GE.0 .AND. to_pe.LT.npes) then!use non-blocking sends if(debug .and.(current_clock.NE.0)) call SYSTEM_CLOCK(start_tick)!z1l:truly non-blocking send.! if(request(to_pe).NE.MPI_REQUEST_NULL) then !only one message from pe-> to_pe in queue *PE waiting for to_pe ! call error else get_len so only do gets but you cannot have a pure get with MPI call a get means do a wait to ensure put on remote PE is complete error call increase mpp_nml request_multiply call MPP_TRANSMIT get_len end if return end subroutine MPP_TRANSMIT_ ! MPP_BROADCAST ! subroutine but that doesn t allow !broadcast to a subset of PEs This version and mpp_transmit will remain !backward compatible intent(inout) a
Definition: mpp_transmit_mpi.h:178

not
************************************************************************GNU Lesser General Public License **This file is part of the GFDL Flexible Modeling System(FMS). ! *! *FMS is free software without even the implied warranty of MERCHANTABILITY or *FITNESS FOR A PARTICULAR PURPOSE See the GNU General Public License *for more details **You should have received a copy of the GNU Lesser General Public *License along with FMS If not
Definition: mpp_write.h:4