1
1
openmpi/ompi/mca/btl/mx/btl_mx_proc.c
George Bosilca 3903009b8b Add a check for the unexpected handler. If enabled, allow the zero-copy
protocol over the MX BTL. Now, we have only one matching, the one in Open
MPI.

The problem is that when the unexpected handler is triggered, not all the
message is on the host memory. In the best case we get one MX fragment (internal
MX fragment), in the worst we get NULL. The only way to fit this with the
design of the PML is to force the eager protocol at the MX internal fragment
size, and to limit the send/receive protocol at the same size. Tests show
the outcome is not far from optimal (if the pipeline depth is increased
a little bit).

Set MX_PIPELINE_LOG in order to allow MX to use internal fragments of 4K.

This commit was SVN r12930.
2006-12-28 03:35:41 +00:00

222 строки
7.5 KiB
C

/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2005 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.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include "opal/class/opal_hash_table.h"
#include "ompi/mca/pml/base/pml_base_module_exchange.h"
#include "btl_mx.h"
#include "btl_mx_proc.h"
static void mca_btl_mx_proc_construct(mca_btl_mx_proc_t* proc);
static void mca_btl_mx_proc_destruct(mca_btl_mx_proc_t* proc);
OBJ_CLASS_INSTANCE(mca_btl_mx_proc_t,
opal_list_item_t, mca_btl_mx_proc_construct,
mca_btl_mx_proc_destruct);
void mca_btl_mx_proc_construct(mca_btl_mx_proc_t* proc)
{
proc->proc_ompi = 0;
proc->proc_addr_index = 0;
proc->proc_endpoints = NULL;
proc->proc_endpoint_count = 0;
proc->mx_peers_count = 0;
proc->mx_peers = NULL;
OBJ_CONSTRUCT(&proc->proc_lock, opal_mutex_t);
/* add to list of all proc instance */
OPAL_THREAD_LOCK(&mca_btl_mx_component.mx_lock);
opal_list_append(&mca_btl_mx_component.mx_procs, &proc->super);
OPAL_THREAD_UNLOCK(&mca_btl_mx_component.mx_lock);
}
/*
* Cleanup MX proc instance
*/
void mca_btl_mx_proc_destruct(mca_btl_mx_proc_t* proc)
{
/* remove from list of all proc instances */
OPAL_THREAD_LOCK(&mca_btl_mx_component.mx_lock);
opal_list_remove_item(&mca_btl_mx_component.mx_procs, &proc->super);
OPAL_THREAD_UNLOCK(&mca_btl_mx_component.mx_lock);
/* release resources */
if( NULL != proc->proc_endpoints ) {
free(proc->proc_endpoints);
proc->proc_endpoints = NULL;
}
if( NULL != proc->mx_peers ) {
free(proc->mx_peers);
proc->mx_peers = NULL;
}
}
/*
* Look for an existing MX process instances based on the associated
* ompi_proc_t instance.
*/
static mca_btl_mx_proc_t* mca_btl_mx_proc_lookup_ompi(ompi_proc_t* ompi_proc)
{
mca_btl_mx_proc_t* mx_proc;
OPAL_THREAD_LOCK(&mca_btl_mx_component.mx_lock);
for( mx_proc = (mca_btl_mx_proc_t*)opal_list_get_first(&mca_btl_mx_component.mx_procs);
mx_proc != (mca_btl_mx_proc_t*)opal_list_get_end(&mca_btl_mx_component.mx_procs);
mx_proc = (mca_btl_mx_proc_t*)opal_list_get_next(mx_proc) ) {
if(mx_proc->proc_ompi == ompi_proc) {
OPAL_THREAD_UNLOCK(&mca_btl_mx_component.mx_lock);
return mx_proc;
}
}
OPAL_THREAD_UNLOCK(&mca_btl_mx_component.mx_lock);
return NULL;
}
/*
* Create a MX process structure. There is a one-to-one correspondence
* between a ompi_proc_t and a mca_btl_mx_proc_t instance. We cache
* additional data (specifically the list of mca_btl_mx_endpoint_t instances,
* and published addresses) associated w/ a given destination on this
* datastructure.
*/
mca_btl_mx_proc_t* mca_btl_mx_proc_create(ompi_proc_t* ompi_proc)
{
mca_btl_mx_proc_t* module_proc = NULL;
/* Check if we have already created a MX proc
* structure for this ompi process */
module_proc = mca_btl_mx_proc_lookup_ompi(ompi_proc);
if( module_proc != NULL ) {
/* Gotcha! */
return module_proc;
}
/* Oops! First time, gotta create a new MX proc
* out of the ompi_proc ... */
module_proc = OBJ_NEW(mca_btl_mx_proc_t);
module_proc->proc_ompi = ompi_proc;
return module_proc;
}
/*
* Note that this routine must be called with the lock on the process
* already held. Insert a btl instance into the proc array and assign
* it an address.
*/
int mca_btl_mx_proc_insert( mca_btl_mx_proc_t* module_proc,
mca_btl_mx_endpoint_t* module_endpoint )
{
mca_btl_mx_addr_t *mx_peers;
int rc;
size_t size;
/* query for the peer address info */
rc = mca_pml_base_modex_recv( &mca_btl_mx_component.super.btl_version,
module_proc->proc_ompi, (void*)&mx_peers, &size );
if( OMPI_SUCCESS != rc ) {
opal_output( 0, "mca_pml_base_modex_recv failed for peer [%d,%d,%d]",
ORTE_NAME_ARGS(&module_proc->proc_ompi->proc_name) );
OBJ_RELEASE(module_proc);
return OMPI_ERR_OUT_OF_RESOURCE;
}
if( (size % sizeof(mca_btl_mx_addr_t)) != 0 ) {
opal_output( 0, "invalid mx address for peer [%d,%d,%d]",
ORTE_NAME_ARGS(&module_proc->proc_ompi->proc_name) );
OBJ_RELEASE(module_proc);
return OMPI_ERROR;
}
module_proc->mx_peers_count = size / sizeof(mca_btl_mx_addr_t);
if( 0 == module_proc->mx_peers_count ) { /* no available connection */
return OMPI_ERROR;
}
module_proc->mx_peers = mx_peers;
if( NULL == module_proc->proc_endpoints ) {
module_proc->proc_endpoints = (mca_btl_base_endpoint_t**)
malloc(module_proc->mx_peers_count * sizeof(mca_btl_base_endpoint_t*));
if( NULL == module_proc->proc_endpoints ) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
}
module_endpoint->endpoint_proc = module_proc;
return OMPI_SUCCESS;
}
int mca_btl_mx_proc_connect( mca_btl_mx_endpoint_t* module_endpoint )
{
int num_retry = 0, i;
mx_return_t mx_status;
mx_endpoint_addr_t mx_remote_addr;
mca_btl_mx_proc_t* module_proc = module_endpoint->endpoint_proc;
module_endpoint->status = MCA_BTL_MX_CONNECTION_PENDING;
for( i = module_proc->proc_addr_index; i < module_proc->mx_peers_count; i++ ) {
retry_connect:
mx_status = mx_connect( module_endpoint->endpoint_btl->mx_endpoint,
module_proc->mx_peers[i].nic_id, module_proc->mx_peers[i].endpoint_id,
mca_btl_mx_component.mx_filter, mca_btl_mx_component.mx_timeout, &mx_remote_addr );
if( MX_SUCCESS != mx_status ) {
if( MX_TIMEOUT == mx_status )
if( num_retry++ < mca_btl_mx_component.mx_connection_retries )
goto retry_connect;
{
char peer_name[MX_MAX_HOSTNAME_LEN];
if( MX_SUCCESS != mx_nic_id_to_hostname( module_proc->mx_peers[i].nic_id, peer_name ) )
sprintf( peer_name, "unknown %lx nic_id", (long)module_proc->mx_peers[i].nic_id );
opal_output( 0, "mx_connect fail for %s(%dth remote address) with key %x (error %s)\n",
peer_name, i, mca_btl_mx_component.mx_filter, mx_strerror(mx_status) );
}
continue;
}
module_endpoint->mx_peer.nic_id = module_proc->mx_peers[i].nic_id;
module_endpoint->mx_peer.endpoint_id = module_proc->mx_peers[i].endpoint_id;
module_endpoint->mx_peer_addr = mx_remote_addr;
module_endpoint->status = MCA_BTL_MX_CONNECTED;
module_proc->proc_addr_index = i;
break;
}
if( i == module_proc->mx_peers_count ) { /* no available connection */
module_endpoint->status = MCA_BTL_MX_NOT_REACHEABLE;
return OMPI_ERROR;
}
module_proc->proc_endpoints[module_proc->proc_endpoint_count++] = module_endpoint;
return OMPI_SUCCESS;
}