1
1
openmpi/ompi/mca/coll/fca/coll_fca_module.c

435 строки
15 KiB
C
Исходник Обычный вид История

/**
Copyright (c) 2011 Mellanox Technologies. All rights reserved.
Per RFC, bring in the following changes: * Remove paffinity, maffinity, and carto frameworks -- they've been wholly replaced by hwloc. * Move ompi_mpi_init() affinity-setting/checking code down to ORTE. * Update sm, smcuda, wv, and openib components to no longer use carto. Instead, use hwloc data. There are still optimizations possible in the sm/smcuda BTLs (i.e., making multiple mpools). Also, the old carto-based code found out how many NUMA nodes were ''available'' -- not how many were used ''in this job''. The new hwloc-using code computes the same value -- it was not updated to calculate how many NUMA nodes are used ''by this job.'' * Note that I cannot compile the smcuda and wv BTLs -- I ''think'' they're right, but they need to be verified by their owners. * The openib component now does a bunch of stuff to figure out where "near" OpenFabrics devices are. '''THIS IS A CHANGE IN DEFAULT BEHAVIOR!!''' and still needs to be verified by OpenFabrics vendors (I do not have a NUMA machine with an OpenFabrics device that is a non-uniform distance from multiple different NUMA nodes). * Completely rewrite the OMPI_Affinity_str() routine from the "affinity" mpiext extension. This extension now understands hyperthreads; the output format of it has changed a bit to reflect this new information. * Bunches of minor changes around the code base to update names/types from maffinity/paffinity-based names to hwloc-based names. * Add some helper functions into the hwloc base, mainly having to do with the fact that we have the hwloc data reporting ''all'' topology information, but sometimes you really only want the (online | available) data. This commit was SVN r26391.
2012-05-07 14:52:54 +00:00
Copyright (c) 2012 Cisco Systems, Inc. All rights reserved.
$COPYRIGHT$
Additional copyrights may follow
$HEADER$
*/
#include "coll_fca.h"
/*
* Initial query function that is invoked during MPI_INIT, allowing
* this module to indicate what level of thread support it provides.
*/
int mca_coll_fca_init_query(bool enable_progress_threads,
bool enable_mpi_threads)
{
return OMPI_SUCCESS;
}
static int have_remote_peers(ompi_group_t *group, size_t size, int *local_peers)
{
ompi_proc_t *proc;
size_t i;
int ret;
*local_peers = 0;
ret = 0;
for (i = 0; i < size; ++i) {
proc = ompi_group_peer_lookup(group, i);
if (FCA_IS_LOCAL_PROCESS(proc->proc_flags)) {
++*local_peers;
} else {
ret = 1;
}
}
return ret;
}
static inline ompi_proc_t* __local_rank_lookup(ompi_communicator_t *comm, int rank)
{
return ompi_group_peer_lookup(comm->c_local_group, rank);
}
/**
* Fills local rank information in fca_module.
*/
static int __get_local_ranks(mca_coll_fca_module_t *fca_module)
{
ompi_communicator_t *comm = fca_module->comm;
ompi_proc_t* proc;
int i, rank;
/* Count the local ranks */
fca_module->num_local_procs = 0;
for (rank = 0; rank < ompi_comm_size(comm); ++rank) {
proc = __local_rank_lookup(comm, rank);
if (FCA_IS_LOCAL_PROCESS(proc->proc_flags)) {
if (rank == fca_module->rank) {
fca_module->local_proc_idx = fca_module->num_local_procs;
}
++fca_module->num_local_procs;
}
}
/* Make a list of local ranks */
fca_module->local_ranks = calloc(fca_module->num_local_procs,
sizeof *fca_module->local_ranks);
if (!fca_module->local_ranks) {
FCA_ERROR("Failed to allocate memory for %d local ranks",
fca_module->num_local_procs);
return OMPI_ERROR;
}
i = 0;
for (rank = 0; rank < ompi_comm_size(comm); ++rank) {
proc = __local_rank_lookup(comm, rank);
if (FCA_IS_LOCAL_PROCESS(proc->proc_flags)) {
fca_module->local_ranks[i++] = rank;
}
}
FCA_MODULE_VERBOSE(fca_module, 3, "i am %d/%d", fca_module->local_proc_idx,
fca_module->num_local_procs);
return OMPI_SUCCESS;
}
static int __fca_comm_new(mca_coll_fca_module_t *fca_module)
{
ompi_communicator_t *comm = fca_module->comm;
fca_comm_new_spec_t spec;
int info_size, all_info_size;
void *all_info, *my_info;
int *rcounts, *disps;
int i, rc, ret;
/* call fca_get_rank_info() on node managers only*/
if (fca_module->local_proc_idx == 0) {
my_info = mca_coll_fca_component.fca_ops.get_rank_info(mca_coll_fca_component.fca_context,
&info_size);
if (!my_info) {
FCA_ERROR("fca_get_rank_info returned NULL");
return OMPI_ERROR;
}
} else {
info_size = 0;
}
FCA_MODULE_VERBOSE(fca_module, 1, "Info size: %d", info_size);
/* Allocate gather buffer on the root rank */
if (fca_module->rank == 0) {
rcounts = calloc(ompi_comm_size(comm), sizeof *rcounts);
}
/* Get all rank info sizes using MPI_Gather */
rc = comm->c_coll.coll_gather(&info_size, 1, MPI_INT, rcounts, 1, MPI_INT, 0,
comm, comm->c_coll.coll_gather_module);
if (rc != OMPI_SUCCESS)
return rc;
/* Allocate buffer for gathering rank information on rank0 */
if (fca_module->rank == 0) {
all_info_size = 0;
FCA_MODULE_VERBOSE(fca_module, 1, "Total rank_info size: %d", all_info_size);
disps = calloc(ompi_comm_size(comm), sizeof *disps);
for (i = 0; i < ompi_comm_size(comm); ++i) {
disps[i] = all_info_size;
all_info_size += rcounts[i];
}
all_info = calloc(all_info_size, 1);
}
/* Send all node managers information to rank0 using MPI_Gatherv*/
rc = comm->c_coll.coll_gatherv(my_info, info_size, MPI_BYTE,
all_info, rcounts, disps, MPI_BYTE, 0,
comm, comm->c_coll.coll_gather_module);
if (rc != OMPI_SUCCESS) {
FCA_ERROR("Failed to gather rank information to rank0: %d", rc);
return rc;
}
/* Rank0 calls fca_comm_new() and fills fca_comm_spec filed */
if (fca_module->rank == 0) {
spec.rank_info = all_info;
spec.is_comm_world = comm == MPI_COMM_WORLD;
spec.rank_count = 0;
for (i = 0; i < ompi_comm_size(comm); ++i) {
FCA_MODULE_VERBOSE(fca_module, 1, "rcounts[%d]=%d disps[%d]=%d",
i, rcounts[i], i, disps[i]);
if (rcounts[i] > 0)
++spec.rank_count;
}
free(disps);
free(rcounts);
FCA_MODULE_VERBOSE(fca_module, 1, "starting fca_comm_new(), rank_count: %d",
spec.rank_count);
ret = mca_coll_fca_component.fca_ops.comm_new(mca_coll_fca_component.fca_context,
&spec, &fca_module->fca_comm_desc);
free(all_info);
}
/* Broadcast return value from rank0 to all other ranks */
rc = fca_module->previous_bcast(&ret, 1, MPI_INT, 0, comm,
fca_module->previous_bcast_module);
if (rc != OMPI_SUCCESS) {
FCA_ERROR("Failed to broadcast comm_new return value from rank0: %d", rc);
return rc;
}
/* Examine comm_new return value */
if (ret < 0) {
FCA_ERROR("COMM_NEW failed: %s", mca_coll_fca_component.fca_ops.strerror(ret));
return OMPI_ERROR;
}
/* Release allocate rank_info on node managers */
if (fca_module->local_proc_idx == 0) {
mca_coll_fca_component.fca_ops.free_rank_info(my_info);
}
/* Pass fca_comm_desc to all ranks using MPI_Bcast */
rc = fca_module->previous_bcast(&fca_module->fca_comm_desc,
sizeof(fca_module->fca_comm_desc), MPI_BYTE, 0,
comm, fca_module->previous_bcast_module);
if (rc != OMPI_SUCCESS) {
FCA_ERROR("Failed to broadcast comm_desc from rank0: %d", rc);
return rc;
}
FCA_MODULE_VERBOSE(fca_module, 1, "Received FCA communicator spec, comm_id %d",
fca_module->fca_comm_desc.comm_id);
return OMPI_SUCCESS;
}
static int __create_fca_comm(mca_coll_fca_module_t *fca_module)
{
int comm_size;
int rc, ret;
rc = __fca_comm_new(fca_module);
if (rc != OMPI_SUCCESS)
return rc;
/* allocate comm_init_spec */
FCA_MODULE_VERBOSE(fca_module, 1, "Starting COMM_INIT comm_id %d proc_idx %d num_procs %d",
fca_module->fca_comm_desc.comm_id, fca_module->local_proc_idx,
fca_module->num_local_procs);
comm_size = ompi_comm_size(fca_module->comm);
ret = mca_coll_fca_comm_init(&mca_coll_fca_component.fca_ops,
mca_coll_fca_component.fca_context,
fca_module->rank, comm_size,
fca_module->local_proc_idx, fca_module->num_local_procs,
&fca_module->fca_comm_desc, &fca_module->fca_comm);
if (ret < 0) {
FCA_ERROR("COMM_INIT failed: %s", mca_coll_fca_component.fca_ops.strerror(ret));
return OMPI_ERROR;
}
/* get communicator capabilities */
ret = mca_coll_fca_component.fca_ops.comm_get_caps(fca_module->fca_comm,
&fca_module->fca_comm_caps);
if (ret < 0) {
FCA_ERROR("GET_COMM_CAPS failed: %s", mca_coll_fca_component.fca_ops.strerror(ret));
return OMPI_ERROR;
}
/* by this point every rank in the communicator is set up */
FCA_MODULE_VERBOSE(fca_module, 1, "Initialized FCA communicator, comm_id %d",
fca_module->fca_comm_desc.comm_id);
return OMPI_SUCCESS;
}
static void __destroy_fca_comm(mca_coll_fca_module_t *fca_module)
{
int ret;
mca_coll_fca_component.fca_ops.comm_destroy(fca_module->fca_comm);
if (fca_module->rank == 0) {
ret = mca_coll_fca_component.fca_ops.comm_end(mca_coll_fca_component.fca_context,
fca_module->fca_comm_desc.comm_id);
if (ret < 0) {
FCA_ERROR("COMM_END failed: %s", mca_coll_fca_component.fca_ops.strerror(ret));
}
}
FCA_MODULE_VERBOSE(fca_module, 1, "Destroyed FCA communicator, comm_id %d",
fca_module->fca_comm_desc.comm_id);
}
#define FCA_SAVE_PREV_COLL_API(__api) do {\
fca_module->previous_ ## __api = comm->c_coll.coll_ ## __api;\
fca_module->previous_ ## __api ## _module = comm->c_coll.coll_ ## __api ## _module;\
if (!comm->c_coll.coll_ ## __api || !comm->c_coll.coll_ ## __api ## _module) {\
FCA_VERBOSE(1, "(%d/%s): no underlying " # __api"; disqualifying myself",\
comm->c_contextid, comm->c_name);\
return OMPI_ERROR;\
}\
OBJ_RETAIN(fca_module->previous_ ## __api ## _module);\
} while(0)
static int __save_coll_handlers(mca_coll_fca_module_t *fca_module)
{
ompi_communicator_t *comm = fca_module->comm;
FCA_SAVE_PREV_COLL_API(barrier);
FCA_SAVE_PREV_COLL_API(bcast);
FCA_SAVE_PREV_COLL_API(reduce);
FCA_SAVE_PREV_COLL_API(allreduce);
FCA_SAVE_PREV_COLL_API(allgather);
FCA_SAVE_PREV_COLL_API(allgatherv);
FCA_SAVE_PREV_COLL_API(gather);
FCA_SAVE_PREV_COLL_API(gatherv);
FCA_SAVE_PREV_COLL_API(alltoall);
FCA_SAVE_PREV_COLL_API(alltoallv);
FCA_SAVE_PREV_COLL_API(alltoallw);
FCA_SAVE_PREV_COLL_API(reduce_scatter);
return OMPI_SUCCESS;
}
/*
* Initialize module on the communicator
*/
static int mca_coll_fca_module_enable(mca_coll_base_module_t *module,
struct ompi_communicator_t *comm)
{
mca_coll_fca_module_t *fca_module = (mca_coll_fca_module_t*) module;
int rc;
fca_module->comm = comm;
fca_module->rank = ompi_comm_rank(comm);
rc = mca_coll_fca_get_fca_lib(comm);
if (rc != OMPI_SUCCESS)
return rc;
rc = __save_coll_handlers(fca_module);
if (rc != OMPI_SUCCESS)
return rc;
rc = __get_local_ranks(fca_module);
if (rc != OMPI_SUCCESS)
return rc;
rc = __create_fca_comm(fca_module);
if (rc != OMPI_SUCCESS)
return rc;
FCA_MODULE_VERBOSE(fca_module, 1, "FCA Module initialized");
return OMPI_SUCCESS;
}
static int mca_coll_fca_ft_event(int state)
{
return OMPI_SUCCESS;
}
static void mca_coll_fca_module_clear(mca_coll_fca_module_t *fca_module)
{
fca_module->num_local_procs = 0;
fca_module->local_ranks = NULL;
fca_module->fca_comm = NULL;
fca_module->previous_barrier = NULL;
fca_module->previous_bcast = NULL;
fca_module->previous_reduce = NULL;
fca_module->previous_allreduce = NULL;
fca_module->previous_allgather = NULL;
fca_module->previous_allgatherv = NULL;
fca_module->previous_gather = NULL;
fca_module->previous_gatherv = NULL;
fca_module->previous_alltoall = NULL;
fca_module->previous_alltoallv = NULL;
fca_module->previous_alltoallw = NULL;
fca_module->previous_reduce_scatter = NULL;
}
static void mca_coll_fca_module_construct(mca_coll_fca_module_t *fca_module)
{
FCA_VERBOSE(5, "==>");
mca_coll_fca_module_clear(fca_module);
}
static void mca_coll_fca_module_destruct(mca_coll_fca_module_t *fca_module)
{
FCA_VERBOSE(5, "==>");
OBJ_RELEASE(fca_module->previous_barrier_module);
OBJ_RELEASE(fca_module->previous_bcast_module);
OBJ_RELEASE(fca_module->previous_reduce_module);
OBJ_RELEASE(fca_module->previous_allreduce_module);
OBJ_RELEASE(fca_module->previous_allgather_module);
OBJ_RELEASE(fca_module->previous_allgatherv_module);
OBJ_RELEASE(fca_module->previous_gather_module);
OBJ_RELEASE(fca_module->previous_gatherv_module);
OBJ_RELEASE(fca_module->previous_alltoall_module);
OBJ_RELEASE(fca_module->previous_alltoallv_module);
OBJ_RELEASE(fca_module->previous_alltoallw_module);
OBJ_RELEASE(fca_module->previous_reduce_scatter_module);
if (fca_module->fca_comm)
__destroy_fca_comm(fca_module);
free(fca_module->local_ranks);
mca_coll_fca_module_clear(fca_module);
}
/*
* Invoked when there's a new communicator that has been created.
* Look at the communicator and decide which set of functions and
* priority we want to return.
*/
mca_coll_base_module_t *
mca_coll_fca_comm_query(struct ompi_communicator_t *comm, int *priority)
{
mca_coll_base_module_t *module;
int size = ompi_comm_size(comm);
int local_peers;
mca_coll_fca_module_t *fca_module;
*priority = 0;
module = NULL;
if (!mca_coll_fca_component.fca_enable)
goto exit;
if (size < mca_coll_fca_component.fca_np)
goto exit;
if (!have_remote_peers(comm->c_local_group, size, &local_peers) || OMPI_COMM_IS_INTER(comm))
goto exit;
fca_module = OBJ_NEW(mca_coll_fca_module_t);
if (!fca_module)
goto exit;
fca_module->super.coll_module_enable = mca_coll_fca_module_enable;
fca_module->super.ft_event = mca_coll_fca_ft_event;
fca_module->super.coll_allgather = mca_coll_fca_component.fca_enable_allgather? mca_coll_fca_allgather : NULL;
fca_module->super.coll_allgatherv = mca_coll_fca_component.fca_enable_allgatherv? mca_coll_fca_allgatherv : NULL;
fca_module->super.coll_allreduce = mca_coll_fca_component.fca_enable_allreduce? mca_coll_fca_allreduce : NULL;
fca_module->super.coll_alltoall = mca_coll_fca_component.fca_enable_alltoall? mca_coll_fca_alltoall : NULL;
fca_module->super.coll_alltoallv = mca_coll_fca_component.fca_enable_alltoallv? mca_coll_fca_alltoallv : NULL;
fca_module->super.coll_alltoallw = mca_coll_fca_component.fca_enable_alltoallw? mca_coll_fca_alltoallw : NULL;
fca_module->super.coll_barrier = mca_coll_fca_component.fca_enable_barrier? mca_coll_fca_barrier : NULL;
fca_module->super.coll_bcast = mca_coll_fca_component.fca_enable_bcast? mca_coll_fca_bcast : NULL;
fca_module->super.coll_exscan = NULL;
fca_module->super.coll_gather = mca_coll_fca_component.fca_enable_gather? mca_coll_fca_gather : NULL;
fca_module->super.coll_gatherv = mca_coll_fca_component.fca_enable_gatherv? mca_coll_fca_gatherv : NULL;
fca_module->super.coll_reduce = mca_coll_fca_component.fca_enable_reduce? mca_coll_fca_reduce : NULL;
fca_module->super.coll_reduce_scatter = mca_coll_fca_component.fca_enable_reduce_scatter? mca_coll_fca_reduce_scatter : NULL;
fca_module->super.coll_scan = NULL;
fca_module->super.coll_scatter = NULL;
fca_module->super.coll_scatterv = NULL;
*priority = mca_coll_fca_component.fca_priority;
module = &fca_module->super;
exit:
FCA_VERBOSE(4, "Query FCA module for comm %p size %d rank %d local_peers=%d: priority=%d %s",
(void *)comm, size, ompi_comm_rank(comm), local_peers,
*priority, module ? "enabled" : "disabled");
return module;
}
OBJ_CLASS_INSTANCE(mca_coll_fca_module_t,
mca_coll_base_module_t,
mca_coll_fca_module_construct,
mca_coll_fca_module_destruct);