1
1
openmpi/ompi/tools/ompi_info/components.c

368 строки
12 KiB
C
Исходник Обычный вид История

/*
* Copyright (c) 2004-2010 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2011 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.
Per RFC, bring in the following changes: * Remove paffinity, maffinity, and carto frameworks -- they've been wholly replaced by hwloc. * Move ompi_mpi_init() affinity-setting/checking code down to ORTE. * Update sm, smcuda, wv, and openib components to no longer use carto. Instead, use hwloc data. There are still optimizations possible in the sm/smcuda BTLs (i.e., making multiple mpools). Also, the old carto-based code found out how many NUMA nodes were ''available'' -- not how many were used ''in this job''. The new hwloc-using code computes the same value -- it was not updated to calculate how many NUMA nodes are used ''by this job.'' * Note that I cannot compile the smcuda and wv BTLs -- I ''think'' they're right, but they need to be verified by their owners. * The openib component now does a bunch of stuff to figure out where "near" OpenFabrics devices are. '''THIS IS A CHANGE IN DEFAULT BEHAVIOR!!''' and still needs to be verified by OpenFabrics vendors (I do not have a NUMA machine with an OpenFabrics device that is a non-uniform distance from multiple different NUMA nodes). * Completely rewrite the OMPI_Affinity_str() routine from the "affinity" mpiext extension. This extension now understands hyperthreads; the output format of it has changed a bit to reflect this new information. * Bunches of minor changes around the code base to update names/types from maffinity/paffinity-based names to hwloc-based names. * Add some helper functions into the hwloc base, mainly having to do with the fact that we have the hwloc data reporting ''all'' topology information, but sometimes you really only want the (online | available) data. This commit was SVN r26391.
2012-05-07 14:52:54 +00:00
* Copyright (c) 2006-2012 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010-2012 Los Alamos National Security, LLC.
* All rights reserved.
* Copyright (c) 2011-2012 University of Houston. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include <stdlib.h>
#include <string.h>
#include "opal/util/argv.h"
#include "opal/runtime/opal_info_support.h"
#include "orte/runtime/runtime.h"
#include "orte/runtime/orte_info_support.h"
#include "ompi/mca/allocator/base/base.h"
#include "ompi/mca/coll/base/base.h"
#include "ompi/mca/io/io.h"
#include "ompi/mca/io/base/base.h"
#include "ompi/mca/mpool/base/base.h"
#include "ompi/mca/pml/pml.h"
#include "ompi/mca/pml/base/base.h"
#include "ompi/mca/bml/base/base.h"
#include "ompi/mca/rcache/rcache.h"
#include "ompi/mca/rcache/base/base.h"
#include "ompi/mca/btl/base/base.h"
#include "ompi/mca/mtl/mtl.h"
#include "ompi/mca/mtl/base/base.h"
#include "ompi/mca/topo/base/base.h"
#include "ompi/mca/osc/osc.h"
#include "ompi/mca/osc/base/base.h"
#include "ompi/mca/pubsub/base/base.h"
#include "ompi/mca/dpm/base/base.h"
#include "ompi/mca/op/base/base.h"
#include "ompi/mca/vprotocol/base/base.h"
#include "ompi/mca/fbtl/fbtl.h"
#include "ompi/mca/fbtl/base/base.h"
#include "ompi/mca/fs/fs.h"
#include "ompi/mca/fs/base/base.h"
#include "ompi/mca/fcoll/fcoll.h"
#include "ompi/mca/fcoll/base/base.h"
#include "ompi/mca/sharedfp/sharedfp.h"
#include "ompi/mca/sharedfp/base/base.h"
#include "ompi/runtime/params.h"
#if OPAL_ENABLE_FT_CR == 1
#include "ompi/mca/crcp/crcp.h"
#include "ompi/mca/crcp/base/base.h"
#endif
#include "ompi/tools/ompi_info/ompi_info.h"
/*
* Public variables
*/
/*
* Private variables
*/
static bool opened_components = false;
/*
* Open all MCA components so that they can register their MCA
* parameters. Take a shotgun approach here and indiscriminately open
* all components -- don't be selective. To this end, we need to clear
* out the environment of all OMPI_MCA_<type> variables to ensure
* that the open algorithms don't try to only open one component.
*/
int ompi_info_register_components(opal_pointer_array_t *mca_types,
opal_pointer_array_t *component_map)
{
int i;
char *env, *str;
char *target, *save, *type;
char **env_save=NULL;
bool need_close_components = false;
opal_info_component_map_t *map;
/* Clear out the environment. Use strdup() to orphan the resulting
* strings because items are placed in the environment by reference,
* not by value.
*/
for (i = 0; i < mca_types->size; ++i) {
if (NULL == (type = (char*)opal_pointer_array_get_item(mca_types, i))) {
continue;
}
asprintf(&env, "OMPI_MCA_%s", type);
if (NULL != (save = getenv(env))) {
/* save this param so it can later be restored */
asprintf(&str, "%s=%s", env, save);
opal_argv_append_nosize(&env_save, str);
free(str);
/* can't manipulate it directly, so make a copy first */
asprintf(&target, "%s=", env);
putenv(target);
free(target);
}
free(env);
}
/* Register the MPI layer's MCA parameters */
if (OMPI_SUCCESS != ompi_mpi_register_params()) {
str = "ompi_mpi_Register_params failed";
goto error;
}
/* Find / open all components */
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("base");
opal_pointer_array_add(component_map, map);
/* set default error message from here forward */
str = "A component framework failed to open properly.";
/* MPI frameworks */
if (OMPI_SUCCESS != mca_allocator_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("allocator");
map->components = &mca_allocator_base_components;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_btl_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("btl");
map->components = &mca_btl_base_components_opened;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_coll_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("coll");
map->components = &mca_coll_base_components_opened;
opal_pointer_array_add(component_map, map);
#if OPAL_ENABLE_FT_CR == 1
if (OMPI_SUCCESS != ompi_crcp_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("crcp");
map->components = &ompi_crcp_base_components_available;
opal_pointer_array_add(component_map, map);
#endif
if (OMPI_SUCCESS != ompi_dpm_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("dpm");
map->components = &ompi_dpm_base_components_available;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_fbtl_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("fbtl");
map->components = &mca_fbtl_base_components_opened;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_fcoll_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("fcoll");
map->components = &mca_fcoll_base_components_opened;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_fs_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("fs");
map->components = &mca_fs_base_components_opened;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_io_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("io");
map->components = &mca_io_base_components_opened;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_mpool_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("mpool");
map->components = &mca_mpool_base_components;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != ompi_mtl_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("mtl");
map->components = &ompi_mtl_base_components_opened;
opal_pointer_array_add(component_map, map);
ompi_op_base_open();
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("op");
map->components = &ompi_op_base_components_opened;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != ompi_osc_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("osc");
map->components = &ompi_osc_base_open_components;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_pml_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("pml");
map->components = &mca_pml_base_components_available;
opal_pointer_array_add(component_map, map);
/* No need to call the bml_base_open() because the ob1 pml calls it.
* mca_bml_base_open();
*/
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("bml");
map->components = &mca_bml_base_components_available;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != ompi_pubsub_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("pubsub");
map->components = &ompi_pubsub_base_components_available;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_rcache_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("rcache");
map->components = &mca_rcache_base_components;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_sharedfp_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("sharedfp");
map->components = &mca_sharedfp_base_components_opened;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_topo_base_open()) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("topo");
map->components = &mca_topo_base_components_opened;
opal_pointer_array_add(component_map, map);
if (OMPI_SUCCESS != mca_vprotocol_base_open(NULL)) {
goto error;
}
map = OBJ_NEW(opal_info_component_map_t);
map->type = strdup("vprotocol");
map->components = &mca_vprotocol_base_components_available;
opal_pointer_array_add(component_map, map);
/* flag that we need to close components */
need_close_components = true;
/* Restore the environment to what it was before we started so that
* if users setenv OMPI_MCA_<framework name> to some value, they'll
* see that value when it is shown via --param output.
*/
if (NULL != env_save) {
for (i = 0; i < opal_argv_count(env_save); ++i) {
putenv(env_save[i]);
}
}
/* All done */
opened_components = true;
return OMPI_SUCCESS;
error:
fprintf(stderr, "%s\n", str);
fprintf(stderr, "ompi_info will likely not display all configuration information\n");
if (need_close_components) {
opened_components = true;
ompi_info_close_components();
}
return OMPI_ERROR;
}
void ompi_info_close_components()
{
/* Note that the order of shutdown here doesn't matter because
* we aren't *using* any components -- none were selected, so
* there are no dependencies between the frameworks. We list
* them generally "in order", but it doesn't really matter.
* We also explicitly ignore the return values from the
* close() functions -- what would we do if there was an
* error?
*/
/* close the OPAL components */
(void) opal_info_close_components();
/* close the ORTE components */
(void) orte_info_close_components();
#if OPAL_ENABLE_FT_CR == 1
(void) ompi_crcp_base_close();
#endif
(void) ompi_op_base_close();
(void) ompi_dpm_base_close();
(void) ompi_pubsub_base_close();
(void) mca_topo_base_close();
(void) mca_btl_base_close();
(void) ompi_mtl_base_close();
(void) mca_pml_base_close();
(void) mca_mpool_base_close();
(void) mca_rcache_base_close();
(void) mca_io_base_close();
(void) mca_fbtl_base_close();
(void) mca_fcoll_base_close();
(void) mca_fs_base_close();
(void) mca_sharedfp_base_close();
(void) mca_coll_base_close();
(void) mca_allocator_base_close();
(void) ompi_osc_base_close();
}