2005-07-01 01:28:35 +04:00
|
|
|
/*
|
2005-11-05 22:57:48 +03:00
|
|
|
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
|
|
|
* University Research and Technology
|
|
|
|
* Corporation. All rights reserved.
|
2011-02-25 03:36:08 +03:00
|
|
|
* Copyright (c) 2004-2011 The University of Tennessee and The University
|
2005-11-05 22:57:48 +03:00
|
|
|
* of Tennessee Research Foundation. All rights
|
|
|
|
* reserved.
|
2008-01-21 15:11:18 +03:00
|
|
|
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
2005-07-01 01:28:35 +04:00
|
|
|
* University of Stuttgart. All rights reserved.
|
|
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
|
|
* All rights reserved.
|
2008-05-02 15:52:33 +04:00
|
|
|
* Copyright (c) 2007-2008 Cisco Systems, Inc. All rights reserved.
|
2007-09-24 14:11:52 +04:00
|
|
|
* Copyright (c) 2006-2007 Voltaire All rights reserved.
|
2005-07-01 01:28:35 +04:00
|
|
|
* $COPYRIGHT$
|
2008-01-21 15:11:18 +03:00
|
|
|
*
|
2005-07-01 01:28:35 +04:00
|
|
|
* Additional copyrights may follow
|
2008-01-21 15:11:18 +03:00
|
|
|
*
|
2005-07-01 01:28:35 +04:00
|
|
|
* $HEADER$
|
|
|
|
*/
|
|
|
|
|
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"
|
2005-07-01 01:28:35 +04:00
|
|
|
|
2008-04-18 00:43:56 +04:00
|
|
|
#include "opal/util/arch.h"
|
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
#include "btl_openib.h"
|
|
|
|
#include "btl_openib_proc.h"
|
2008-05-02 15:52:33 +04:00
|
|
|
#include "connect/base.h"
|
|
|
|
#include "connect/connect.h"
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
static void mca_btl_openib_proc_construct(mca_btl_openib_proc_t* proc);
|
|
|
|
static void mca_btl_openib_proc_destruct(mca_btl_openib_proc_t* proc);
|
|
|
|
|
2008-01-21 15:11:18 +03:00
|
|
|
OBJ_CLASS_INSTANCE(mca_btl_openib_proc_t,
|
|
|
|
opal_list_item_t, mca_btl_openib_proc_construct,
|
2005-07-01 01:28:35 +04:00
|
|
|
mca_btl_openib_proc_destruct);
|
|
|
|
|
2009-01-05 17:14:38 +03:00
|
|
|
void mca_btl_openib_proc_construct(mca_btl_openib_proc_t* ib_proc)
|
2005-07-01 01:28:35 +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
|
|
|
ib_proc->proc_opal = 0;
|
|
|
|
ib_proc->proc_ports = NULL;
|
|
|
|
ib_proc->proc_port_count = 0;
|
|
|
|
ib_proc->proc_endpoints = 0;
|
2009-01-05 17:14:38 +03:00
|
|
|
ib_proc->proc_endpoint_count = 0;
|
|
|
|
OBJ_CONSTRUCT(&ib_proc->proc_lock, opal_mutex_t);
|
2005-07-01 01:28:35 +04:00
|
|
|
/* add to list of all proc instance */
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&mca_btl_openib_component.ib_lock);
|
2009-01-05 17:14:38 +03:00
|
|
|
opal_list_append(&mca_btl_openib_component.ib_procs, &ib_proc->super);
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&mca_btl_openib_component.ib_lock);
|
2005-07-01 01:28:35 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Cleanup ib proc instance
|
|
|
|
*/
|
|
|
|
|
2009-01-05 17:14:38 +03:00
|
|
|
void mca_btl_openib_proc_destruct(mca_btl_openib_proc_t* ib_proc)
|
2005-07-01 01:28:35 +04:00
|
|
|
{
|
|
|
|
/* remove from list of all proc instances */
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&mca_btl_openib_component.ib_lock);
|
2009-01-05 17:14:38 +03:00
|
|
|
opal_list_remove_item(&mca_btl_openib_component.ib_procs, &ib_proc->super);
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&mca_btl_openib_component.ib_lock);
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
/* release resources */
|
2009-01-05 17:14:38 +03:00
|
|
|
if(NULL != ib_proc->proc_endpoints) {
|
|
|
|
free(ib_proc->proc_endpoints);
|
2005-07-01 01:28:35 +04:00
|
|
|
}
|
2009-01-05 17:14:38 +03:00
|
|
|
if (NULL != ib_proc->proc_ports) {
|
2008-05-02 15:52:33 +04:00
|
|
|
int i, j;
|
2009-01-05 17:14:38 +03:00
|
|
|
for (i = 0; i < ib_proc->proc_port_count; ++i) {
|
|
|
|
for (j = 0; j < ib_proc->proc_ports[i].pm_cpc_data_count; ++j) {
|
|
|
|
if (NULL != ib_proc->proc_ports[i].pm_cpc_data[j].cbm_modex_message) {
|
|
|
|
free(ib_proc->proc_ports[i].pm_cpc_data[j].cbm_modex_message);
|
2008-05-02 15:52:33 +04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2009-01-05 17:14:38 +03:00
|
|
|
free(ib_proc->proc_ports);
|
2008-05-02 15:52:33 +04:00
|
|
|
}
|
2009-01-05 17:14:38 +03:00
|
|
|
OBJ_DESTRUCT(&ib_proc->proc_lock);
|
2005-07-01 01:28:35 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Look for an existing IB process instances based on the associated
|
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_proc_t instance.
|
2005-07-01 01:28:35 +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
|
|
|
static mca_btl_openib_proc_t* mca_btl_openib_proc_lookup_proc(opal_proc_t* proc)
|
2005-07-01 01:28:35 +04:00
|
|
|
{
|
|
|
|
mca_btl_openib_proc_t* ib_proc;
|
|
|
|
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_LOCK(&mca_btl_openib_component.ib_lock);
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
for(ib_proc = (mca_btl_openib_proc_t*)
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_get_first(&mca_btl_openib_component.ib_procs);
|
2005-07-01 01:28:35 +04:00
|
|
|
ib_proc != (mca_btl_openib_proc_t*)
|
2005-07-03 20:22:16 +04:00
|
|
|
opal_list_get_end(&mca_btl_openib_component.ib_procs);
|
|
|
|
ib_proc = (mca_btl_openib_proc_t*)opal_list_get_next(ib_proc)) {
|
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(ib_proc->proc_opal == proc) {
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&mca_btl_openib_component.ib_lock);
|
2005-07-01 01:28:35 +04:00
|
|
|
return ib_proc;
|
|
|
|
}
|
|
|
|
}
|
2005-07-04 02:45:48 +04:00
|
|
|
OPAL_THREAD_UNLOCK(&mca_btl_openib_component.ib_lock);
|
2005-07-01 01:28:35 +04:00
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2008-05-02 15:52:33 +04:00
|
|
|
static void inline unpack8(char **src, uint8_t *value)
|
|
|
|
{
|
|
|
|
/* Copy one character */
|
|
|
|
*value = (uint8_t) **src;
|
|
|
|
/* Most the src ahead one */
|
|
|
|
++*src;
|
|
|
|
}
|
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
/*
|
|
|
|
* Create a IB process structure. There is a one-to-one correspondence
|
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
|
|
|
* between a opal_proc_t and a mca_btl_openib_proc_t instance. We
|
2008-05-02 15:52:33 +04:00
|
|
|
* cache additional data (specifically the list of
|
|
|
|
* mca_btl_openib_endpoint_t instances, and published addresses)
|
|
|
|
* associated w/ a given destination on this datastructure.
|
2005-07-01 01:28:35 +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
|
|
|
mca_btl_openib_proc_t* mca_btl_openib_proc_create(opal_proc_t* proc)
|
2005-07-01 01:28:35 +04:00
|
|
|
{
|
|
|
|
mca_btl_openib_proc_t* module_proc = NULL;
|
2008-01-15 02:22:03 +03:00
|
|
|
size_t msg_size;
|
|
|
|
uint32_t size;
|
2008-05-02 15:52:33 +04:00
|
|
|
int rc, i, j;
|
2008-01-15 02:22:03 +03:00
|
|
|
void *message;
|
|
|
|
char *offset;
|
2008-05-02 15:52:33 +04:00
|
|
|
int modex_message_size;
|
|
|
|
mca_btl_openib_modex_message_t dummy;
|
2008-01-21 15:11:18 +03:00
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
/* Check if we have already created a IB proc
|
|
|
|
* structure for this ompi process */
|
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
|
|
|
module_proc = mca_btl_openib_proc_lookup_proc(proc);
|
2008-01-15 02:22:03 +03:00
|
|
|
if (NULL != module_proc) {
|
2005-07-01 01:28:35 +04:00
|
|
|
/* Gotcha! */
|
|
|
|
return module_proc;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Oops! First time, gotta create a new IB proc
|
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
|
|
|
* out of the opal_proc ... */
|
2005-07-01 01:28:35 +04:00
|
|
|
module_proc = OBJ_NEW(mca_btl_openib_proc_t);
|
|
|
|
/* Initialize number of peer */
|
|
|
|
module_proc->proc_endpoint_count = 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
|
|
|
module_proc->proc_opal = proc;
|
2005-07-01 01:28:35 +04:00
|
|
|
|
2008-01-21 15:11:18 +03:00
|
|
|
/* query for the peer address info */
|
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_modex_recv(&mca_btl_openib_component.super.btl_version,
|
|
|
|
proc,
|
2008-01-15 02:22:03 +03:00
|
|
|
&message,
|
2008-01-21 15:11:18 +03:00
|
|
|
&msg_size);
|
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) {
|
|
|
|
BTL_ERROR(("[%s:%d] opal_modex_recv failed for peer %s",
|
2007-10-15 21:53:02 +04:00
|
|
|
__FILE__, __LINE__,
|
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_NAME_PRINT(proc->proc_name)));
|
2005-10-01 02:58:09 +04:00
|
|
|
OBJ_RELEASE(module_proc);
|
|
|
|
return NULL;
|
|
|
|
}
|
2008-01-15 02:22:03 +03:00
|
|
|
if (0 == msg_size) {
|
2005-10-01 02:58:09 +04:00
|
|
|
return NULL;
|
|
|
|
}
|
2005-07-01 01:28:35 +04:00
|
|
|
|
2008-01-15 02:22:03 +03:00
|
|
|
/* Message was packed in btl_openib_component.c; the format is
|
|
|
|
listed in a comment in that file */
|
2008-05-02 15:52:33 +04:00
|
|
|
modex_message_size = ((char *) &(dummy.end)) - ((char*) &dummy);
|
|
|
|
|
|
|
|
/* Unpack the number of modules in the message */
|
2010-07-12 20:17:56 +04:00
|
|
|
offset = (char *) message;
|
2008-05-02 15:52:33 +04:00
|
|
|
unpack8(&offset, &(module_proc->proc_port_count));
|
2008-06-26 00:16:34 +04:00
|
|
|
BTL_VERBOSE(("unpack: %d btls", module_proc->proc_port_count));
|
2008-05-02 15:52:33 +04:00
|
|
|
if (module_proc->proc_port_count > 0) {
|
|
|
|
module_proc->proc_ports = (mca_btl_openib_proc_modex_t *)
|
2011-07-04 18:00:41 +04:00
|
|
|
malloc(sizeof(mca_btl_openib_proc_modex_t) *
|
2008-05-02 15:52:33 +04:00
|
|
|
module_proc->proc_port_count);
|
|
|
|
} else {
|
|
|
|
module_proc->proc_ports = NULL;
|
|
|
|
}
|
2008-01-15 02:22:03 +03:00
|
|
|
|
|
|
|
/* Loop over unpacking all the ports */
|
|
|
|
for (i = 0; i < module_proc->proc_port_count; i++) {
|
|
|
|
|
2008-05-02 15:52:33 +04:00
|
|
|
/* Unpack the modex comment message struct */
|
|
|
|
size = modex_message_size;
|
|
|
|
memcpy(&(module_proc->proc_ports[i].pm_port_info), offset, size);
|
2009-05-07 00:11:28 +04:00
|
|
|
#if !defined(WORDS_BIGENDIAN) && OPAL_ENABLE_HETEROGENEOUS_SUPPORT
|
2008-05-02 15:52:33 +04:00
|
|
|
MCA_BTL_OPENIB_MODEX_MSG_NTOH(module_proc->proc_ports[i].pm_port_info);
|
2008-01-15 02:22:03 +03:00
|
|
|
#endif
|
2008-05-02 15:52:33 +04:00
|
|
|
offset += size;
|
2008-06-26 00:16:34 +04:00
|
|
|
BTL_VERBOSE(("unpacked btl %d: modex message, offset now %d",
|
|
|
|
i, (int)(offset-((char*)message))));
|
2008-01-15 02:22:03 +03:00
|
|
|
|
2008-05-02 15:52:33 +04:00
|
|
|
/* Unpack the number of CPCs that follow */
|
|
|
|
unpack8(&offset, &(module_proc->proc_ports[i].pm_cpc_data_count));
|
2008-06-26 00:16:34 +04:00
|
|
|
BTL_VERBOSE(("unpacked btl %d: number of cpcs to follow %d (offset now %d)",
|
2011-07-04 18:00:41 +04:00
|
|
|
i, module_proc->proc_ports[i].pm_cpc_data_count,
|
2008-06-26 00:16:34 +04:00
|
|
|
(int)(offset-((char*)message))));
|
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
|
|
|
module_proc->proc_ports[i].pm_cpc_data = (opal_btl_openib_connect_base_module_data_t *)
|
2008-05-02 15:52:33 +04:00
|
|
|
calloc(module_proc->proc_ports[i].pm_cpc_data_count,
|
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
|
|
|
sizeof(opal_btl_openib_connect_base_module_data_t));
|
2008-05-02 15:52:33 +04:00
|
|
|
if (NULL == module_proc->proc_ports[i].pm_cpc_data) {
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Unpack the CPCs */
|
|
|
|
for (j = 0; j < module_proc->proc_ports[i].pm_cpc_data_count; ++j) {
|
|
|
|
uint8_t u8;
|
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_btl_openib_connect_base_module_data_t *cpcd;
|
2008-05-02 15:52:33 +04:00
|
|
|
cpcd = module_proc->proc_ports[i].pm_cpc_data + j;
|
|
|
|
unpack8(&offset, &u8);
|
2008-06-26 00:16:34 +04:00
|
|
|
BTL_VERBOSE(("unpacked btl %d: cpc %d: index %d (offset now %d)",
|
|
|
|
i, j, u8, (int)(offset-(char*)message)));
|
2011-07-04 18:00:41 +04:00
|
|
|
cpcd->cbm_component =
|
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_btl_openib_connect_base_get_cpc_byindex(u8);
|
2008-06-26 00:16:34 +04:00
|
|
|
BTL_VERBOSE(("unpacked btl %d: cpc %d: component %s",
|
|
|
|
i, j, cpcd->cbm_component->cbc_name));
|
2011-07-04 18:00:41 +04:00
|
|
|
|
2008-05-02 15:52:33 +04:00
|
|
|
unpack8(&offset, &cpcd->cbm_priority);
|
|
|
|
unpack8(&offset, &cpcd->cbm_modex_message_len);
|
2008-06-26 00:16:34 +04:00
|
|
|
BTL_VERBOSE(("unpacked btl %d: cpc %d: priority %d, msg len %d (offset now %d)",
|
2011-07-04 18:00:41 +04:00
|
|
|
i, j, cpcd->cbm_priority,
|
2008-06-26 00:16:34 +04:00
|
|
|
cpcd->cbm_modex_message_len,
|
|
|
|
(int)(offset-(char*)message)));
|
2008-05-02 15:52:33 +04:00
|
|
|
if (cpcd->cbm_modex_message_len > 0) {
|
|
|
|
cpcd->cbm_modex_message = malloc(cpcd->cbm_modex_message_len);
|
|
|
|
if (NULL == cpcd->cbm_modex_message) {
|
|
|
|
BTL_ERROR(("Failed to malloc"));
|
|
|
|
return NULL;
|
|
|
|
}
|
2011-07-04 18:00:41 +04:00
|
|
|
memcpy(cpcd->cbm_modex_message, offset,
|
2008-05-02 15:52:33 +04:00
|
|
|
cpcd->cbm_modex_message_len);
|
|
|
|
offset += cpcd->cbm_modex_message_len;
|
2008-06-26 00:16:34 +04:00
|
|
|
BTL_VERBOSE(("unpacked btl %d: cpc %d: blob unpacked %d %x (offset now %d)",
|
|
|
|
i, j,
|
|
|
|
((uint32_t*)cpcd->cbm_modex_message)[0],
|
|
|
|
((uint32_t*)cpcd->cbm_modex_message)[1],
|
|
|
|
(int)(offset-((char*)message))));
|
2008-05-02 15:52:33 +04:00
|
|
|
}
|
2008-01-15 02:22:03 +03:00
|
|
|
}
|
|
|
|
}
|
2007-01-04 01:35:41 +03:00
|
|
|
|
2005-10-01 02:58:09 +04:00
|
|
|
if (0 == module_proc->proc_port_count) {
|
|
|
|
module_proc->proc_endpoints = NULL;
|
|
|
|
} else {
|
|
|
|
module_proc->proc_endpoints = (mca_btl_base_endpoint_t**)
|
2011-07-04 18:00:41 +04:00
|
|
|
malloc(module_proc->proc_port_count *
|
2008-05-02 15:52:33 +04:00
|
|
|
sizeof(mca_btl_base_endpoint_t*));
|
2005-10-01 02:58:09 +04:00
|
|
|
}
|
2008-01-15 02:22:03 +03:00
|
|
|
if (NULL == module_proc->proc_endpoints) {
|
2005-07-01 01:28:35 +04:00
|
|
|
OBJ_RELEASE(module_proc);
|
|
|
|
return NULL;
|
|
|
|
}
|
2008-05-02 15:52:33 +04:00
|
|
|
|
2008-06-26 00:16:34 +04:00
|
|
|
BTL_VERBOSE(("unpacking done!"));
|
2005-07-01 01:28:35 +04:00
|
|
|
return module_proc;
|
|
|
|
}
|
|
|
|
|
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 mca_btl_openib_proc_remove(opal_proc_t *proc,
|
2008-05-02 15:52:33 +04:00
|
|
|
mca_btl_base_endpoint_t *endpoint)
|
|
|
|
{
|
|
|
|
size_t i;
|
|
|
|
mca_btl_openib_proc_t* ib_proc = NULL;
|
|
|
|
|
|
|
|
/* Remove endpoint from the openib BTL version of the proc as
|
|
|
|
well */
|
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
|
|
|
ib_proc = mca_btl_openib_proc_lookup_proc(proc);
|
2008-05-02 15:52:33 +04:00
|
|
|
if (NULL != ib_proc) {
|
|
|
|
for (i = 0; i < ib_proc->proc_endpoint_count; ++i) {
|
|
|
|
if (ib_proc->proc_endpoints[i] == endpoint) {
|
|
|
|
ib_proc->proc_endpoints[i] = NULL;
|
|
|
|
if (i == ib_proc->proc_endpoint_count - 1) {
|
|
|
|
--ib_proc->proc_endpoint_count;
|
|
|
|
}
|
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;
|
2008-05-02 15:52:33 +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
|
|
|
return OPAL_ERR_NOT_FOUND;
|
2008-05-02 15:52:33 +04:00
|
|
|
}
|
2005-07-01 01:28:35 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Note that this routine must be called with the lock on the process
|
2008-01-21 15:11:18 +03:00
|
|
|
* already held. Insert a btl instance into the proc array and assign
|
2005-07-01 01:28:35 +04:00
|
|
|
* it an address.
|
|
|
|
*/
|
2008-01-21 15:11:18 +03:00
|
|
|
int mca_btl_openib_proc_insert(mca_btl_openib_proc_t* module_proc,
|
2005-07-01 01:28:35 +04:00
|
|
|
mca_btl_base_endpoint_t* module_endpoint)
|
|
|
|
{
|
|
|
|
/* insert into endpoint array */
|
2008-01-21 15:11:18 +03:00
|
|
|
|
|
|
|
|
2007-01-13 02:14:45 +03:00
|
|
|
#ifndef WORDS_BIGENDIAN
|
|
|
|
/* if we are little endian and our peer is not so lucky, then we
|
|
|
|
need to put all information sent to him in big endian (aka
|
|
|
|
Network Byte Order) and expect all information received to
|
|
|
|
be in NBO. Since big endian machines always send and receive
|
|
|
|
in NBO, we don't care so much about that case. */
|
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 (module_proc->proc_opal->proc_arch & OPAL_ARCH_ISBIGENDIAN) {
|
2007-01-13 02:14:45 +03:00
|
|
|
module_endpoint->nbo = true;
|
|
|
|
}
|
|
|
|
#endif
|
2008-01-21 15:11:18 +03:00
|
|
|
|
2007-01-13 02:14:45 +03:00
|
|
|
/* only allow eager rdma if the peers agree on the size of a long */
|
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((module_proc->proc_opal->proc_arch & OPAL_ARCH_LONGISxx) !=
|
|
|
|
(opal_proc_local_get()->proc_arch & OPAL_ARCH_LONGISxx)) {
|
2007-01-13 02:14:45 +03:00
|
|
|
module_endpoint->use_eager_rdma = false;
|
|
|
|
}
|
|
|
|
|
2005-07-01 01:28:35 +04:00
|
|
|
module_endpoint->endpoint_proc = module_proc;
|
|
|
|
module_proc->proc_endpoints[module_proc->proc_endpoint_count++] = module_endpoint;
|
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;
|
2005-07-01 01:28:35 +04:00
|
|
|
}
|