1
1
openmpi/ompi/mca/io/base/base.h
Brian Barrett b57b8c5b3f Clean up request handling in the I/O framework to be more consistent with
other request-using frameworks.

 - Rather than having mpi/c/* functions allocate requests explicitly,
   pass the MPI_Request* down to the I/O component and have it 
   perform the allocation.
 - While the I/O base provides a base request which can be used,
   it is not required and all request management occurs within
   the component.
 - Push progress management into the component, rather than having it
   happen in the base.  Progress functions are now easily registered,
   and not all (ie, the one existing) components use progress functions
   in any rational way.

ROMIO switched to generalized requests instead of MPIO_Requests many
moons ago, and Open MPI now uses ROMIO's generalized requests, so there
is no reason to wrap those requests (which are OMPI requests) in another
level of request.

Now the file function passes the MPI_Request* to the ROMIO component,
which passes it to the underlying ROMIO function, which calls 
MPI_Grequest_start to create an OMPI request, which is what gets set
as the request to the user.  Much cleaner.

This patch has two motivations.  One, a whole heck of a lot of code
just got removed, and request handling is now much cleaner for I/O
components.  Two, by adding support for Argonne's proposed generalized
request extensions, we can allow ROMIO to provide async I/O through
generalized requests, which we couldn't rationally do in the old
setup due to the crazy request completion rules.

This commit was SVN r22235.
2009-11-26 05:13:43 +00:00

241 строка
9.4 KiB
C

/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2005 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2008 Cisco Systems, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
/**
* @file
*
* MCA io base framework public interface functions.
*/
#ifndef MCA_IO_BASE_H
#define MCA_IO_BASE_H
#include "ompi_config.h"
#include "mpi.h"
#include "opal/class/opal_list.h"
#include "ompi/class/ompi_free_list.h"
#include "ompi/mca/io/io.h"
/*
* Global functions for MCA overall io open and close
*/
BEGIN_C_DECLS
/**
* Initialize the io MCA framework
*
* @retval OMPI_SUCCESS Upon success
* @retval OMPI_ERROR Upon failure
*
* This must be the first function invoked in the io MCA
* framework. It initializes the io MCA framework, finds and
* opens io components, etc.
*
* This function is invoked during ompi_mpi_init() and during the
* initialization of the special case of the laminfo command.
*
* This function fills in the internal global variable
* mca_io_base_components_opened, which is a list of all io components
* that were successfully opened. This variable should \em only be
* used by other io base functions -- it is not considered a
* public interface member -- and is only mentioned here for
* completeness.
*/
OMPI_DECLSPEC int mca_io_base_open(void);
/**
* Create list of available io components.
*
* @param allow_multi_user_threads Will be set to true if any of the
* available components will allow multiple user threads
* @param have_hidden_threads Will be set to true if any of the
* available components have hidden threads.
*
* @retval OMPI_SUCCESS If one or more io components are available.
* @retval OMPI_ERROR If no io components are found to be available.
*
* This function is invoked during ompi_mpi_init() to query all
* successfully opened io components and create a list of all
* available io components.
*
* This function traverses the (internal global variable)
* mca_io_base_components_opened list and queries each component to see
* if it ever might want to run during this MPI process. It creates
* another internal global variable list named
* mca_io_base_components_available, consisting of a list of components
* that are available for selection when file handles are created.
* This variable should \em only be used by other io base
* functions -- it is not considered a public interface member --
* and is only mentioned here for completeness.
*/
OMPI_DECLSPEC int mca_io_base_find_available(bool enable_progress_threads,
bool enable_mpi_threads);
/**
* Select an available component for a new file handle.
*
* @param file File Handle that the component will be selected for.
* @param preferred The component that is preferred for this
* file handle (or NULL).
*
* @return OMPI_SUCCESS Upon success.
* @return OMPI_ERROR Upon failure.
*
* Note that the types of the parameters have "struct" in them
* (e.g., ompi_file_t" vs. a plain "ompi_file_t") to
* avoid an include file loop. All similar types (e.g., "struct
* ompi_file_t *", "ompi_file_t *", and "MPI_File")
* are all typedef'ed to be the same, so the fact that we use struct
* here in the prototype is ok.
*
* This function is invoked when a new file handle is created and a
* io component needs to be selected for it. It should be invoked
* near the end of the file handle creation process such that
* almost everything else is functional on the file handle.
*
* This function invokes the selection process for io components,
* which works as follows:
*
* - If the \em preferred argument is NULL, the selection set is
* defined to be all the components found during
* mca_io_base_find_available().
* - If \em preferred is not NULL, then the selection set is just
* that component. (However, in this mode, we may make 2 passes
* through the selection process -- more on this below).
* - All components in the selection set are queried to see if they
* want to run with that file handle. All components that want to
* run are ranked by their priority and the highest priority
* component is selected. All non-selected components have their
* "unquery" function invoked to let them know that they were not
* selected.
* - The selected module will have its "init" function
* invoked to let it know that it was selected. All unselected
* components will have their file_unselect function invoked.
* - If we fall through this entire process and no component is
* selected \em and the \em preferred argument is not NULL, then
* run the entire process again as if the \em preferred argument
* was NULL (i.e., use the entire available set of components).
*
* At the end of this process, we'll either have a single
* component/module pair that is selected and initialized for the
* file handle, or no component was selected and an error is
* returned up the stack.
*/
OMPI_DECLSPEC int mca_io_base_file_select(struct ompi_file_t *file,
mca_base_component_t *preferred);
/**
* Finalize a io component on a specific file handle.
*
* @param file The file handle that is being destroyed.
*
* @retval OMPI_SUCCESS Always.
*
* Note that the type of the parameter is only a "struct
* ompi_file_t" (vs. a plain "ompi_file_t") to avoid an include file
* loop. The types "struct ompi_file_t *", "ompi_file_t *", and
* "MPI_File" are all typedef'ed to be the same, so the fact that we
* use struct here in the prototype is ok.
*
* This function is invoked near the beginning of the destruction of
* a file handle. It finalizes the io component associated with the
* file handle (e.g., allowing the component to clean up and free any
* resources allocated for that file handle). Note that similar to
* mca_io_base_select(), as result of this function, other
* file handles may also be destroyed.
*/
OMPI_DECLSPEC int mca_io_base_file_unselect(struct ompi_file_t *file);
/**
* Invoke a back-end component to delete a file.
*
* @param filename Name of the file to be deleted
* @param info MPI_Info for additional information
*
* This function is a bit different than most other MPI_File_*
* functions -- it does not take a MPI_File handle. As such, this
* function function invokes appropriate delete handlers on all
* the available components (rather than some pre-selected
* module). See io.h for details.
*/
OMPI_DECLSPEC int mca_io_base_delete(char *filename,
struct ompi_info_t *info);
/**
* Shut down the io MCA framework.
*
* @retval OMPI_SUCCESS Always
*
* This function shuts down everything in the io MCA framework,
* and is called during ompi_mpi_finalize() and the special case of
* the laminfo fileand.
*
* It must be the last function invoked on the io MCA framework.
*/
OMPI_DECLSPEC int mca_io_base_close(void);
OMPI_DECLSPEC int mca_io_base_register_datarep(char *,
MPI_Datarep_conversion_function*,
MPI_Datarep_conversion_function*,
MPI_Datarep_extent_function*,
void*);
/*
* Globals
*/
/**
* Index number from the "io" MCA parameter, created when the io
* framework is initialized and used during scope selection.
*/
OMPI_DECLSPEC extern int mca_io_base_param;
/**
* io framework debugging stream ID used with opal_output() and
* opal_output_verbose().
*/
OMPI_DECLSPEC extern int mca_io_base_output;
/**
* Indicator as to whether the list of opened io components is valid or
* not.
*/
OMPI_DECLSPEC extern bool mca_io_base_components_opened_valid;
/**
* List of all opened components; created when the io framework is
* initialized and destroyed when we reduce the list to all available
* io components.
*/
OMPI_DECLSPEC extern opal_list_t mca_io_base_components_opened;
/**
* Indicator as to whether the list of available io components is valid
* or not.
*/
OMPI_DECLSPEC extern bool mca_io_base_components_available_valid;
/**
* List of all available components; created by reducing the list of open
* components to all those who indicate that they may run during this
* process.
*/
OMPI_DECLSPEC extern opal_list_t mca_io_base_components_available;
END_C_DECLS
#endif /* MCA_BASE_IO_H */