1
1
openmpi/opal/mca/btl/ugni/btl_ugni_smsg.h

143 строки
5.1 KiB
C
Исходник Обычный вид История

/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil -*- */
/*
* Copyright (c) 2011-2015 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2011 UT-Battelle, LLC. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#if !defined(MCA_BTL_UGNI_SMSG_H)
#define MCA_BTL_UGNI_SMSG_H
#include "btl_ugni.h"
#include "btl_ugni_endpoint.h"
#include "btl_ugni_frag.h"
#include "btl_ugni_rdma.h"
typedef enum {
MCA_BTL_UGNI_TAG_SEND,
MCA_BTL_UGNI_TAG_DISCONNECT,
MCA_BTL_UGNI_TAG_GET_INIT,
MCA_BTL_UGNI_TAG_RDMA_COMPLETE
} mca_btl_ugni_smsg_tag_t;
typedef struct mca_btl_ugni_smsg_mbox_t {
opal_free_list_item_t super;
mca_btl_ugni_endpoint_attr_t attr;
} mca_btl_ugni_smsg_mbox_t;
OBJ_CLASS_DECLARATION(mca_btl_ugni_smsg_mbox_t);
int mca_btl_ugni_smsg_init (mca_btl_ugni_module_t *ugni_module);
int mca_btl_ugni_smsg_process (mca_btl_base_endpoint_t *ep);
int mca_btl_ugni_progress_remote_smsg (mca_btl_ugni_module_t *btl);
static inline int mca_btl_ugni_progress_local_smsg (mca_btl_ugni_module_t *ugni_module)
{
mca_btl_ugni_base_frag_t *frag;
gni_cq_entry_t event_data;
gni_return_t grc;
/* nothing to do */
if (0 == ugni_module->active_send_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;
}
OPAL_THREAD_LOCK(&ugni_module->device->dev_lock);
grc = GNI_CqGetEvent (ugni_module->smsg_local_cq, &event_data);
OPAL_THREAD_UNLOCK(&ugni_module->device->dev_lock);
if (GNI_RC_NOT_DONE == grc) {
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;
}
if (OPAL_UNLIKELY((GNI_RC_SUCCESS != grc && !event_data) || GNI_CQ_OVERRUN(event_data))) {
/* TODO -- need to handle overrun -- how do we do this without an event?
will the event eventually come back? Ask Cray */
BTL_ERROR(("post error! cq overrun = %d", (int)GNI_CQ_OVERRUN(event_data)));
assert (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
return opal_common_rc_ugni_to_opal (grc);
}
assert (GNI_CQ_GET_TYPE(event_data) == GNI_CQ_EVENT_TYPE_SMSG);
frag = (mca_btl_ugni_base_frag_t *) opal_pointer_array_get_item (&ugni_module->pending_smsg_frags_bb,
GNI_CQ_GET_MSG_ID(event_data));
if (OPAL_UNLIKELY(NULL == frag)) {
assert (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
return OPAL_ERROR;
}
opal_atomic_add_32(&ugni_module->active_send_count,-1);
frag->flags |= MCA_BTL_UGNI_FRAG_SMSG_COMPLETE;
if (!(frag->flags & MCA_BTL_UGNI_FRAG_IGNORE)) {
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_ugni_frag_complete (frag, OPAL_SUCCESS);
}
return 1;
}
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 inline int opal_mca_btl_ugni_smsg_send (mca_btl_ugni_base_frag_t *frag,
void *hdr, size_t hdr_len,
void *payload, size_t payload_len,
mca_btl_ugni_smsg_tag_t tag)
{
gni_return_t grc;
OPAL_THREAD_LOCK(&frag->endpoint->common->dev->dev_lock);
grc = GNI_SmsgSendWTag (frag->endpoint->smsg_ep_handle, hdr, hdr_len,
payload, payload_len, frag->msg_id, tag);
OPAL_THREAD_UNLOCK(&frag->endpoint->common->dev->dev_lock);
if (OPAL_LIKELY(GNI_RC_SUCCESS == grc)) {
/* increment the active send counter */
opal_atomic_add_32(&frag->endpoint->btl->active_send_count,1);
if (mca_btl_ugni_component.progress_thread_enabled) {
if (frag->base.des_flags & MCA_BTL_DES_FLAGS_SIGNAL) {
/* errors for PostCqWrite treated as non-fatal */
(void) mca_btl_ugni_post_cqwrite (frag->endpoint, frag->endpoint->btl->rdma_local_cq,
frag->endpoint->rmt_irq_mem_hndl, 0xdead, NULL, NULL, NULL);
2014-12-05 02:18:16 +03:00
}
}
(void) mca_btl_ugni_progress_local_smsg ((mca_btl_ugni_module_t *) frag->endpoint->btl);
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;
}
if (OPAL_LIKELY(GNI_RC_NOT_DONE == grc)) {
BTL_VERBOSE(("out of credits"));
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_OUT_OF_RESOURCE;
}
BTL_ERROR(("GNI_SmsgSendWTag failed with rc = %d. handle = %lu, hdr_len = %d, payload_len = %d",
grc, (uintptr_t) frag->endpoint->smsg_ep_handle, (int) hdr_len, (int) payload_len));
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_ERROR;
}
static inline int mca_btl_ugni_send_frag (struct mca_btl_base_endpoint_t *btl_peer,
mca_btl_ugni_base_frag_t *frag) {
if (OPAL_LIKELY(!(frag->flags & MCA_BTL_UGNI_FRAG_EAGER))) {
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_mca_btl_ugni_smsg_send (frag, &frag->hdr.send, frag->hdr_size,
frag->segments[1].seg_addr.pval,
frag->segments[1].seg_len,
MCA_BTL_UGNI_TAG_SEND);
}
frag->hdr.eager.size = frag->segments[1].seg_len;
frag->hdr.eager.address = frag->segments[1].seg_addr.lval;
frag->hdr.eager.ctx = (void *) frag;
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_mca_btl_ugni_smsg_send (frag, &frag->hdr.eager, frag->hdr_size,
NULL, 0, MCA_BTL_UGNI_TAG_GET_INIT);
}
#endif /* MCA_BTL_UGNI_SMSG_H */