e65fa4ff5c
This is a meta commit, that encapsulate all the ADAPT commits in the master
into a single PR for 4.1. The master commits included here are:
fe73586, a4be3bb, d712645, c2970a3, e59bde9, ee592f3 and c98e387.
Here is a detailed list of added capabilities:
* coll/adapt: Fix naming conventions and C11 atomic use
* coll/adapt: Remove unused component field in module
* Consistent handling of zero counts in the MPI API.
* Correctly handle non-blocking collectives tags
* As it is possible to have multiple outstanding non-blocking collectives
provided by different collective modules, we need a consistent
mechanism to allow them to select unique tags for each instance of a
collective.
* Add support for fallback to previous coll module on non-commutative operations (#30)
* Replace mutexes by atomic operations.
* Use the correct nbc request type (for both ibcast and ireduce)
* coll/base: document type casts in ompi_coll_base_retain_*
* add module-wide topology cache
* use standard instead of synchronous send and add mca parameter to control mode of initial send in ireduce/ibcast
* reduce number of memory allocations
* call the default request completion.
* Remove the requests from the Fortran lookup conversion tables before completing
and free it.
* piggybacking Bull functionalities
Signed-off-by: Xi Luo <xluo12@vols.utk.edu>
Signed-off-by: George Bosilca <bosilca@icl.utk.edu>
Signed-off-by: Marc Sergent <marc.sergent@atos.net>
Co-authored-by: Joseph Schuchart <schuchart@hlrs.de>
Co-authored-by: Lemarinier, Pierre <pierre.lemarinier@atos.net>
Co-authored-by: pierrele <31764860+pierrele@users.noreply.github.com>
106 строки
3.1 KiB
C
106 строки
3.1 KiB
C
/*
|
|
* Copyright (c) 2014-2020 The University of Tennessee and The University
|
|
* of Tennessee Research Foundation. All rights
|
|
* reserved.
|
|
* $COPYRIGHT$
|
|
*
|
|
* Additional copyrights may follow
|
|
*
|
|
* $HEADER$
|
|
*/
|
|
|
|
#include "coll_adapt.h"
|
|
#include "coll_adapt_topocache.h"
|
|
|
|
#include "ompi/communicator/communicator.h"
|
|
|
|
static void destruct_topology_cache(adapt_topology_cache_item_t *item)
|
|
{
|
|
if (NULL != item->tree) {
|
|
ompi_coll_base_topo_destroy_tree(&item->tree);
|
|
}
|
|
}
|
|
|
|
OBJ_CLASS_INSTANCE(adapt_topology_cache_item_t, opal_list_item_t,
|
|
NULL, &destruct_topology_cache);
|
|
|
|
static ompi_coll_tree_t *create_topology(
|
|
ompi_coll_adapt_algorithm_t algorithm,
|
|
int root,
|
|
struct ompi_communicator_t *comm)
|
|
{
|
|
switch(algorithm) {
|
|
case OMPI_COLL_ADAPT_ALGORITHM_TUNED:
|
|
{
|
|
return NULL;
|
|
}
|
|
case OMPI_COLL_ADAPT_ALGORITHM_BINOMIAL:
|
|
{
|
|
return ompi_coll_base_topo_build_bmtree(comm, root);
|
|
}
|
|
case OMPI_COLL_ADAPT_ALGORITHM_IN_ORDER_BINOMIAL:
|
|
{
|
|
return ompi_coll_base_topo_build_in_order_bmtree(comm, root);
|
|
}
|
|
case OMPI_COLL_ADAPT_ALGORITHM_BINARY:
|
|
{
|
|
return ompi_coll_base_topo_build_tree(2, comm, root);
|
|
}
|
|
case OMPI_COLL_ADAPT_ALGORITHM_PIPELINE:
|
|
{
|
|
return ompi_coll_base_topo_build_chain(1, comm, root);
|
|
}
|
|
case OMPI_COLL_ADAPT_ALGORITHM_CHAIN:
|
|
{
|
|
return ompi_coll_base_topo_build_chain(4, comm, root);
|
|
}
|
|
case OMPI_COLL_ADAPT_ALGORITHM_LINEAR:
|
|
{
|
|
int fanout = ompi_comm_size(comm) - 1;
|
|
ompi_coll_tree_t *tree;
|
|
if (fanout < 1) {
|
|
tree = ompi_coll_base_topo_build_chain(1, comm, root);
|
|
} else if (fanout <= MAXTREEFANOUT) {
|
|
tree = ompi_coll_base_topo_build_tree(ompi_comm_size(comm) - 1, comm, root);
|
|
} else {
|
|
tree = ompi_coll_base_topo_build_tree(MAXTREEFANOUT, comm, root);
|
|
}
|
|
return tree;
|
|
}
|
|
default:
|
|
printf("WARN: unknown topology %d\n", algorithm);
|
|
return NULL;
|
|
}
|
|
}
|
|
|
|
ompi_coll_tree_t* adapt_module_cached_topology(
|
|
mca_coll_base_module_t *module,
|
|
struct ompi_communicator_t *comm,
|
|
int root,
|
|
ompi_coll_adapt_algorithm_t algorithm)
|
|
{
|
|
mca_coll_adapt_module_t *adapt_module = (mca_coll_adapt_module_t*)module;
|
|
adapt_topology_cache_item_t *item;
|
|
ompi_coll_tree_t * tree;
|
|
if (NULL != adapt_module->topo_cache) {
|
|
OPAL_LIST_FOREACH(item, adapt_module->topo_cache, adapt_topology_cache_item_t) {
|
|
if (item->root == root && item->algorithm == algorithm) {
|
|
return item->tree;
|
|
}
|
|
}
|
|
} else {
|
|
adapt_module->topo_cache = OBJ_NEW(opal_list_t);
|
|
}
|
|
|
|
/* topology not found, create one */
|
|
tree = create_topology(algorithm, root, comm);
|
|
|
|
item = OBJ_NEW(adapt_topology_cache_item_t);
|
|
item->tree = tree;
|
|
item->root = root;
|
|
item->algorithm = algorithm;
|
|
opal_list_prepend(adapt_module->topo_cache, &item->super);
|
|
return tree;
|
|
}
|
|
|