2010-07-13 14:46:20 +04:00
|
|
|
/*
|
|
|
|
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
|
|
|
* University Research and Technology
|
|
|
|
* Corporation. All rights reserved.
|
|
|
|
* Copyright (c) 2004-2006 The University of Tennessee and The University
|
|
|
|
* of Tennessee Research Foundation. All rights
|
|
|
|
* reserved.
|
|
|
|
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
|
|
|
* University of Stuttgart. All rights reserved.
|
|
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
|
|
* All rights reserved.
|
|
|
|
* Copyright (c) 2010 Oracle and/or its affiliates. All rights reserved.
|
|
|
|
* $COPYRIGHT$
|
|
|
|
*
|
|
|
|
* Additional copyrights may follow
|
|
|
|
*
|
|
|
|
* $HEADER$
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
|
|
/*%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%*/
|
|
|
|
|
|
|
|
#include "ompi_config.h"
|
|
|
|
#include "ompi/constants.h"
|
|
|
|
#include "ompi/mca/pml/pml.h"
|
|
|
|
#include "ompi/mca/bml/bml.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/mca/mpool/mpool.h"
|
2010-07-13 14:46:20 +04:00
|
|
|
#include "pml_bfo.h"
|
|
|
|
#include "pml_bfo_rdma.h"
|
|
|
|
|
|
|
|
/* Use this registration if no registration needed for a BTL instead of NULL.
|
|
|
|
* This will help other code to distinguish case when memory is not registered
|
|
|
|
* from case when registration is not needed */
|
|
|
|
static mca_mpool_base_registration_t pml_bfo_dummy_reg;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Check to see if memory is registered or can be registered. Build a
|
|
|
|
* set of registrations on the request.
|
|
|
|
*/
|
|
|
|
|
|
|
|
size_t mca_pml_bfo_rdma_btls(
|
|
|
|
mca_bml_base_endpoint_t* bml_endpoint,
|
|
|
|
unsigned char* base,
|
|
|
|
size_t size,
|
|
|
|
mca_pml_bfo_com_btl_t* rdma_btls)
|
|
|
|
{
|
|
|
|
int num_btls = mca_bml_base_btl_array_get_size(&bml_endpoint->btl_rdma);
|
|
|
|
double weight_total = 0;
|
|
|
|
int num_btls_used = 0, n;
|
|
|
|
|
|
|
|
/* shortcut when there are no rdma capable btls */
|
|
|
|
if(num_btls == 0) {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* check to see if memory is registered */
|
|
|
|
for(n = 0; n < num_btls && num_btls_used < mca_pml_bfo.max_rdma_per_request;
|
|
|
|
n++) {
|
|
|
|
mca_bml_base_btl_t* bml_btl =
|
|
|
|
mca_bml_base_btl_array_get_index(&bml_endpoint->btl_rdma,
|
|
|
|
(bml_endpoint->btl_rdma_index + n) % num_btls);
|
|
|
|
mca_mpool_base_registration_t* reg = &pml_bfo_dummy_reg;
|
|
|
|
mca_mpool_base_module_t *btl_mpool = bml_btl->btl->btl_mpool;
|
|
|
|
|
|
|
|
if( NULL != btl_mpool ) {
|
|
|
|
if(!mca_pml_bfo.leave_pinned) {
|
|
|
|
/* look through existing registrations */
|
|
|
|
btl_mpool->mpool_find(btl_mpool, base, size, ®);
|
|
|
|
} else {
|
|
|
|
/* register the memory */
|
|
|
|
btl_mpool->mpool_register(btl_mpool, base, size, 0, ®);
|
|
|
|
}
|
|
|
|
|
|
|
|
if(NULL == reg)
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
rdma_btls[num_btls_used].bml_btl = bml_btl;
|
|
|
|
rdma_btls[num_btls_used].btl_reg = reg;
|
|
|
|
weight_total += bml_btl->btl_weight;
|
|
|
|
num_btls_used++;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* if we don't use leave_pinned and all BTLs that already have this memory
|
|
|
|
* registered amount to less then half of available bandwidth - fall back to
|
|
|
|
* pipeline protocol */
|
|
|
|
if(0 == num_btls_used || (!mca_pml_bfo.leave_pinned && weight_total < 0.5))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
mca_pml_bfo_calc_weighted_length(rdma_btls, num_btls_used, size,
|
|
|
|
weight_total);
|
|
|
|
|
|
|
|
bml_endpoint->btl_rdma_index = (bml_endpoint->btl_rdma_index + 1) % num_btls;
|
|
|
|
return num_btls_used;
|
|
|
|
}
|
|
|
|
|
|
|
|
size_t mca_pml_bfo_rdma_pipeline_btls( mca_bml_base_endpoint_t* bml_endpoint,
|
|
|
|
size_t size,
|
|
|
|
mca_pml_bfo_com_btl_t* rdma_btls )
|
|
|
|
{
|
|
|
|
int i, num_btls = mca_bml_base_btl_array_get_size(&bml_endpoint->btl_rdma);
|
|
|
|
double weight_total = 0;
|
|
|
|
|
|
|
|
for(i = 0; i < num_btls && i < mca_pml_bfo.max_rdma_per_request; i++) {
|
|
|
|
rdma_btls[i].bml_btl =
|
|
|
|
mca_bml_base_btl_array_get_next(&bml_endpoint->btl_rdma);
|
|
|
|
if(NULL != rdma_btls[i].bml_btl->btl->btl_mpool)
|
|
|
|
rdma_btls[i].btl_reg = NULL;
|
|
|
|
else
|
|
|
|
rdma_btls[i].btl_reg = &pml_bfo_dummy_reg;
|
|
|
|
|
|
|
|
weight_total += rdma_btls[i].bml_btl->btl_weight;
|
|
|
|
}
|
|
|
|
|
|
|
|
mca_pml_bfo_calc_weighted_length(rdma_btls, i, size, weight_total);
|
|
|
|
|
|
|
|
return i;
|
|
|
|
}
|