2007-08-02 01:33:25 +04:00
|
|
|
/*
|
|
|
|
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
|
|
|
* University Research and Technology
|
|
|
|
* Corporation. All rights reserved.
|
2013-07-04 12:34:37 +04:00
|
|
|
* Copyright (c) 2004-2013 The University of Tennessee and The University
|
2007-08-02 01:33:25 +04:00
|
|
|
* 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.5A
|
|
|
|
* Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
|
2007-08-30 14:51:02 +04:00
|
|
|
* Copyright (c) 2007 Voltaire. All rights reserved.
|
2010-03-25 06:29:27 +03:00
|
|
|
* Copyright (c) 2010 IBM Corporation. All rights reserved.
|
2013-01-28 03:25:10 +04:00
|
|
|
* Copyright (c) 2012 Los Alamos National Security, LLC.
|
|
|
|
* All rights reserved.
|
2007-08-02 01:33:25 +04:00
|
|
|
* $COPYRIGHT$
|
|
|
|
*
|
|
|
|
* Additional copyrights may follow
|
|
|
|
*
|
|
|
|
* $HEADER$
|
|
|
|
*/
|
2006-06-08 08:29:27 +04:00
|
|
|
/**
|
2007-08-02 01:33:25 +04:00
|
|
|
* @file
|
|
|
|
* Description of the Registration Cache framework
|
|
|
|
*/
|
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
#include "opal_config.h"
|
2007-08-02 01:33:25 +04:00
|
|
|
|
2006-06-08 08:29:27 +04:00
|
|
|
#include "opal/mca/mca.h"
|
2013-02-13 01:10:11 +04:00
|
|
|
#include "opal/util/show_help.h"
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
#include "opal/util/proc.h"
|
2013-01-28 03:25:10 +04:00
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
#include "opal/class/ompi_free_list.h"
|
|
|
|
#include "opal/class/opal_rb_tree.h"
|
2006-06-08 08:29:27 +04:00
|
|
|
#include "mpool_base_tree.h"
|
|
|
|
|
|
|
|
|
2007-08-02 01:33:25 +04:00
|
|
|
static int num_leaks = 0;
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
static int max_mem_leaks = -1;
|
2007-08-02 01:33:25 +04:00
|
|
|
static char *leak_msg = NULL;
|
|
|
|
|
|
|
|
static int condition(void *value);
|
|
|
|
static void action(void *key, void *value);
|
|
|
|
|
2006-06-12 20:44:00 +04:00
|
|
|
OBJ_CLASS_INSTANCE(mca_mpool_base_tree_item_t, ompi_free_list_item_t, NULL, NULL);
|
2006-06-08 08:29:27 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* use globals for the tree and the tree_item free list..
|
|
|
|
*/
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
opal_rb_tree_t mca_mpool_base_tree;
|
2006-06-08 08:29:27 +04:00
|
|
|
ompi_free_list_t mca_mpool_base_tree_item_free_list;
|
2007-08-30 14:51:02 +04:00
|
|
|
static opal_mutex_t tree_lock;
|
2006-06-08 08:29:27 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* simple minded compare function...
|
|
|
|
*/
|
|
|
|
int mca_mpool_base_tree_node_compare(void * key1, void * key2)
|
|
|
|
{
|
|
|
|
if(key1 < key2)
|
|
|
|
{
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
else if(key1 > key2)
|
|
|
|
{
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* initialize the rb tree
|
|
|
|
*/
|
|
|
|
int mca_mpool_base_tree_init(void) {
|
|
|
|
int rc;
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
OBJ_CONSTRUCT(&mca_mpool_base_tree, opal_rb_tree_t);
|
2006-06-08 08:29:27 +04:00
|
|
|
OBJ_CONSTRUCT(&mca_mpool_base_tree_item_free_list, ompi_free_list_t);
|
2007-08-30 14:51:02 +04:00
|
|
|
OBJ_CONSTRUCT(&tree_lock, opal_mutex_t);
|
2007-11-02 02:38:50 +03:00
|
|
|
rc = ompi_free_list_init_new(&mca_mpool_base_tree_item_free_list,
|
|
|
|
sizeof(mca_mpool_base_tree_item_t),
|
2010-07-06 18:33:36 +04:00
|
|
|
opal_cache_line_size,
|
2007-11-02 02:38:50 +03:00
|
|
|
OBJ_CLASS(mca_mpool_base_tree_item_t),
|
2010-07-06 18:33:36 +04:00
|
|
|
0,opal_cache_line_size,
|
2007-11-02 02:38:50 +03:00
|
|
|
0, -1 , 4, NULL);
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
if(OPAL_SUCCESS == rc) {
|
|
|
|
rc = opal_rb_tree_init(&mca_mpool_base_tree, mca_mpool_base_tree_node_compare);
|
2006-06-08 08:29:27 +04:00
|
|
|
}
|
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
2013-12-29 15:45:46 +04:00
|
|
|
/*
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
int mca_mpool_base_tree_fini(void)
|
|
|
|
{
|
|
|
|
OBJ_DESTRUCT(&mca_mpool_base_tree);
|
|
|
|
OBJ_DESTRUCT(&mca_mpool_base_tree_item_free_list);
|
|
|
|
OBJ_DESTRUCT(&tree_lock);
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
return OPAL_SUCCESS;
|
2013-12-29 15:45:46 +04:00
|
|
|
}
|
|
|
|
|
2006-06-08 08:29:27 +04:00
|
|
|
/*
|
|
|
|
* insert an item in the rb tree
|
|
|
|
*/
|
|
|
|
int mca_mpool_base_tree_insert(mca_mpool_base_tree_item_t* item) {
|
2007-08-30 14:51:02 +04:00
|
|
|
int rc;
|
|
|
|
|
|
|
|
OPAL_THREAD_LOCK(&tree_lock);
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
rc = opal_rb_tree_insert(&mca_mpool_base_tree, item->key, item);
|
2007-08-30 14:51:02 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&tree_lock);
|
|
|
|
|
|
|
|
return rc;
|
2006-06-08 08:29:27 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* remove an item from the rb tree
|
2010-03-25 06:29:27 +03:00
|
|
|
* Does not put the item back onto the free list. That
|
|
|
|
* must be done separately by calling mca_mpool_base_tree_item_put.
|
|
|
|
* This allows a caller to remove an item from the tree
|
|
|
|
* before safely cleaning up the item and only then returning it
|
|
|
|
* to the free list. If the item is returned to the free list too soon
|
|
|
|
* race conditions can occur
|
|
|
|
*
|
2006-06-08 08:29:27 +04:00
|
|
|
*/
|
|
|
|
int mca_mpool_base_tree_delete(mca_mpool_base_tree_item_t* item) {
|
2007-08-30 14:51:02 +04:00
|
|
|
int rc;
|
|
|
|
|
|
|
|
OPAL_THREAD_LOCK(&tree_lock);
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
rc = opal_rb_tree_delete(&mca_mpool_base_tree, item->key);
|
2007-08-30 14:51:02 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&tree_lock);
|
|
|
|
|
2006-06-08 08:29:27 +04:00
|
|
|
return rc;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* find the item in the rb tree
|
|
|
|
*/
|
2007-08-30 14:51:02 +04:00
|
|
|
mca_mpool_base_tree_item_t* mca_mpool_base_tree_find(void* base) {
|
|
|
|
mca_mpool_base_tree_item_t* item;
|
|
|
|
|
|
|
|
OPAL_THREAD_LOCK(&tree_lock);
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
item = (mca_mpool_base_tree_item_t*)opal_rb_tree_find(&mca_mpool_base_tree,
|
2007-08-30 14:51:02 +04:00
|
|
|
base);
|
|
|
|
OPAL_THREAD_UNLOCK(&tree_lock);
|
|
|
|
|
|
|
|
return item;
|
2006-06-08 08:29:27 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* get a tree item from the free list
|
|
|
|
*/
|
|
|
|
mca_mpool_base_tree_item_t* mca_mpool_base_tree_item_get(void) {
|
2006-06-12 20:44:00 +04:00
|
|
|
ompi_free_list_item_t* item = NULL;
|
2013-07-09 02:07:52 +04:00
|
|
|
OMPI_FREE_LIST_GET_MT(&mca_mpool_base_tree_item_free_list, item);
|
2013-07-04 12:34:37 +04:00
|
|
|
if(NULL != item) {
|
2006-06-08 08:29:27 +04:00
|
|
|
return (mca_mpool_base_tree_item_t*) item;
|
|
|
|
} else {
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
}
|
2006-06-08 18:23:20 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* put an item back into the free list
|
|
|
|
*/
|
|
|
|
void mca_mpool_base_tree_item_put(mca_mpool_base_tree_item_t* item) {
|
2013-07-09 02:07:52 +04:00
|
|
|
OMPI_FREE_LIST_RETURN_MT(&mca_mpool_base_tree_item_free_list,
|
2006-06-12 00:19:12 +04:00
|
|
|
&(item->super));
|
2006-06-08 18:23:20 +04:00
|
|
|
}
|
2007-08-02 01:33:25 +04:00
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Print a show_help kind of message for an items still left in the
|
|
|
|
* tree
|
|
|
|
*/
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
void mca_mpool_base_tree_print(int show_up_to_mem_leaks)
|
2007-08-02 01:33:25 +04:00
|
|
|
{
|
|
|
|
/* If they asked to show 0 leaks, then don't show anything. */
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
if (0 == show_up_to_mem_leaks) {
|
2007-08-02 01:33:25 +04:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
num_leaks = 0;
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
max_mem_leaks = show_up_to_mem_leaks;
|
|
|
|
opal_rb_tree_traverse(&mca_mpool_base_tree, condition, action);
|
2007-08-02 01:33:25 +04:00
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
if (num_leaks <= show_up_to_mem_leaks ||
|
|
|
|
show_up_to_mem_leaks < 0) {
|
2013-02-13 01:10:11 +04:00
|
|
|
opal_show_help("help-mpool-base.txt", "all mem leaks",
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
true, OPAL_NAME_PRINT(OPAL_PROC_MY_NAME),
|
|
|
|
opal_proc_local_get()->proc_hostname,
|
|
|
|
getpid(), leak_msg);
|
2007-08-02 01:33:25 +04:00
|
|
|
} else {
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
int i = num_leaks - show_up_to_mem_leaks;
|
2013-02-13 01:10:11 +04:00
|
|
|
opal_show_help("help-mpool-base.txt", "some mem leaks",
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
true, OPAL_NAME_PRINT(OPAL_PROC_MY_NAME),
|
|
|
|
opal_proc_local_get()->proc_hostname,
|
|
|
|
getpid(), leak_msg, i,
|
2012-06-27 05:28:28 +04:00
|
|
|
(i > 1) ? "s were" : " was",
|
|
|
|
(i > 1) ? "are" : "is");
|
2007-08-02 01:33:25 +04:00
|
|
|
}
|
|
|
|
free(leak_msg);
|
|
|
|
leak_msg = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* Condition function for rb traversal */
|
|
|
|
static int condition(void *value)
|
|
|
|
{
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/* Action function for rb traversal */
|
|
|
|
static void action(void *key, void *value)
|
|
|
|
{
|
|
|
|
char *tmp;
|
|
|
|
mca_mpool_base_tree_item_t *item = (mca_mpool_base_tree_item_t *) value;
|
|
|
|
|
George did the work and deserves all the credit for it. Ralph did the merge, and deserves whatever blame results from errors in it :-)
WHAT: Open our low-level communication infrastructure by moving all necessary components (btl/rcache/allocator/mpool) down in OPAL
All the components required for inter-process communications are currently deeply integrated in the OMPI layer. Several groups/institutions have express interest in having a more generic communication infrastructure, without all the OMPI layer dependencies. This communication layer should be made available at a different software level, available to all layers in the Open MPI software stack. As an example, our ORTE layer could replace the current OOB and instead use the BTL directly, gaining access to more reactive network interfaces than TCP. Similarly, external software libraries could take advantage of our highly optimized AM (active message) communication layer for their own purpose. UTK with support from Sandia, developped a version of Open MPI where the entire communication infrastucture has been moved down to OPAL (btl/rcache/allocator/mpool). Most of the moved components have been updated to match the new schema, with few exceptions (mainly BTLs where I have no way of compiling/testing them). Thus, the completion of this RFC is tied to being able to completing this move for all BTLs. For this we need help from the rest of the Open MPI community, especially those supporting some of the BTLs. A non-exhaustive list of BTLs that qualify here is: mx, portals4, scif, udapl, ugni, usnic.
This commit was SVN r32317.
2014-07-26 04:47:28 +04:00
|
|
|
if( (++num_leaks <= max_mem_leaks) || (max_mem_leaks < 0) ) {
|
2007-08-02 01:33:25 +04:00
|
|
|
|
|
|
|
/* We know that we're supposed to make the first one; check on
|
|
|
|
successive items if we're supposed to catenate more
|
|
|
|
notices. */
|
|
|
|
if (NULL == leak_msg) {
|
|
|
|
asprintf(&leak_msg, " %lu bytes at address 0x%lx",
|
|
|
|
(unsigned long) item->num_bytes,
|
|
|
|
(unsigned long) key);
|
|
|
|
} else {
|
|
|
|
asprintf(&tmp, "%s\n %lu bytes at address 0x%lx",
|
|
|
|
leak_msg, (unsigned long) item->num_bytes,
|
|
|
|
(unsigned long) key);
|
|
|
|
free(leak_msg);
|
|
|
|
leak_msg = tmp;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|