ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность

Merge pull request #1725 from hjelmn/request_fixes

Просмотр исходного кода 
ompi/request: fix bugs in MPI_Wait_some and MPI_Wait_any
Этот коммит содержится в:
Nathan Hjelm 2016-05-27 13:47:49 -06:00
родитель 3974987ba3 0591139f49
Коммит d8fd3a411a
1 изменённых файлов: 99 добавлений и 84 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

177
ompi/request/req_wait.c
Просмотреть файл

@ -1,3 +1,4 @@
/* -*- Mode: C; c-basic-offset:4 ; indent-tabs-mode:nil -*- */
/* /*
* Copyright (c) 2004-2010 The Trustees of Indiana University and Indiana * Copyright (c) 2004-2010 The Trustees of Indiana University and Indiana
* University Research and Technology * University Research and Technology
@ -12,6 +13,8 @@
* Copyright (c) 2006-2008 Cisco Systems, Inc. All rights reserved. * Copyright (c) 2006-2008 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2010-2012 Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2010-2012 Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved. * Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* Copyright (c) 2016 Los Alamos National Security, LLC. All rights
* reserved.
* $COPYRIGHT$ * $COPYRIGHT$
* *
* Additional copyrights may follow * Additional copyrights may follow
@ -85,7 +88,7 @@ int ompi_request_default_wait_any(size_t count,
int *index, int *index,
ompi_status_public_t * status) ompi_status_public_t * status)
{ {
size_t i = 0, completed = count, num_requests_null_inactive = 0; size_t completed = count, num_requests_null_inactive = 0;
int rc = OMPI_SUCCESS; int rc = OMPI_SUCCESS;
ompi_request_t **rptr=NULL; ompi_request_t **rptr=NULL;
ompi_request_t *request=NULL; ompi_request_t *request=NULL;
@ -95,7 +98,7 @@ int ompi_request_default_wait_any(size_t count,
rptr = requests; rptr = requests;
num_requests_null_inactive = 0; num_requests_null_inactive = 0;
for (i = 0; i < count; i++, rptr++) { for (size_t i = 0; i < count; i++, rptr++) {
request = *rptr; request = *rptr;
/* Sanity test */ /* Sanity test */
@ -111,6 +114,7 @@ int ompi_request_default_wait_any(size_t count,
num_requests_null_inactive++; num_requests_null_inactive++;
continue; continue;
} }
if(!OPAL_ATOMIC_CMPSET_PTR(&request->req_complete, REQUEST_PENDING, &sync)) { if(!OPAL_ATOMIC_CMPSET_PTR(&request->req_complete, REQUEST_PENDING, &sync)) {
assert(REQUEST_COMPLETE(request)); assert(REQUEST_COMPLETE(request));
wait_sync_update( &sync, 1, request->req_status.MPI_ERROR); wait_sync_update( &sync, 1, request->req_status.MPI_ERROR);
@ -118,11 +122,21 @@ int ompi_request_default_wait_any(size_t count,
break; break;
} }
} }
if(num_requests_null_inactive == count) {
*index = MPI_UNDEFINED;
if (MPI_STATUS_IGNORE != status) {
*status = ompi_status_empty;
}
WAIT_SYNC_RELEASE(&sync);
return rc;
}
SYNC_WAIT(&sync); SYNC_WAIT(&sync);
/* recheck the complete status and clean up the sync primitives */ /* recheck the complete status and clean up the sync primitives */
rptr = requests; rptr = requests;
for(i = 0; i < completed; i++, rptr++) { for(size_t i = 0, pending_count = completed; i < pending_count ; i++, rptr++) {
request = *rptr; request = *rptr;
if( request->req_state == OMPI_REQUEST_INACTIVE ) { if( request->req_state == OMPI_REQUEST_INACTIVE ) {
@ -134,44 +148,43 @@ int ompi_request_default_wait_any(size_t count,
* marked as REQUEST_COMPLETE. * marked as REQUEST_COMPLETE.
*/ */
OPAL_ATOMIC_CMPSET_PTR(&request->req_complete, &sync, REQUEST_PENDING); OPAL_ATOMIC_CMPSET_PTR(&request->req_complete, &sync, REQUEST_PENDING);
if (REQUEST_COMPLETE(request)) {
completed = i;
}
} }
if(num_requests_null_inactive == count) { rptr = requests + completed;
*index = MPI_UNDEFINED; request = *rptr;
if (MPI_STATUS_IGNORE != status) {
*status = ompi_status_empty; assert( REQUEST_COMPLETE(request) );
} /* Per note above, we have to call gen request query_fn even
} else { if STATUS_IGNORE was provided */
assert( REQUEST_COMPLETE(request) ); if (OMPI_REQUEST_GEN == request->req_type) {
/* Per note above, we have to call gen request query_fn even rc = ompi_grequest_invoke_query(request, &request->req_status);
if STATUS_IGNORE was provided */
if (OMPI_REQUEST_GEN == request->req_type) {
rc = ompi_grequest_invoke_query(request, &request->req_status);
}
if (MPI_STATUS_IGNORE != status) {
/* Do *NOT* set status->MPI_ERROR here! See MPI-1.1 doc,
sec 3.2.5, p.22 */
int old_error = status->MPI_ERROR;
*status = request->req_status;
status->MPI_ERROR = old_error;
}
rc = request->req_status.MPI_ERROR;
if( request->req_persistent ) {
request->req_state = OMPI_REQUEST_INACTIVE;
} else if (MPI_SUCCESS == rc) {
/* Only free the request if there is no error on it */
/* If there's an error while freeing the request,
assume that the request is still there. Otherwise,
Bad Things will happen later! */
rc = ompi_request_free(rptr);
}
*index = completed;
} }
if (MPI_STATUS_IGNORE != status) {
/* Do *NOT* set status->MPI_ERROR here! See MPI-1.1 doc,
sec 3.2.5, p.22 */
int old_error = status->MPI_ERROR;
*status = request->req_status;
status->MPI_ERROR = old_error;
}
rc = request->req_status.MPI_ERROR;
if( request->req_persistent ) {
request->req_state = OMPI_REQUEST_INACTIVE;
} else if (MPI_SUCCESS == rc) {
/* Only free the request if there is no error on it */
/* If there's an error while freeing the request,
assume that the request is still there. Otherwise,
Bad Things will happen later! */
rc = ompi_request_free(rptr);
}
*index = completed;
#if OPAL_ENABLE_FT_CR == 1 #if OPAL_ENABLE_FT_CR == 1
if( opal_cr_is_enabled) { if( opal_cr_is_enabled) {
rptr = requests; rptr = requests;
for (i = 0; i < count; i++, rptr++) { for (size_t i = 0; i < count; i++, rptr++) {
request = *rptr; request = *rptr;
if( REQUEST_COMPLETE(request) ) { if( REQUEST_COMPLETE(request) ) {
OMPI_CRCP_REQUEST_COMPLETE(request); OMPI_CRCP_REQUEST_COMPLETE(request);
@ -352,7 +365,7 @@ int ompi_request_default_wait_some(size_t count,
int * indices, int * indices,
ompi_status_public_t * statuses) ompi_status_public_t * statuses)
{ {
size_t i, num_requests_null_inactive=0, num_requests_done=0; size_t num_requests_null_inactive=0, num_requests_done=0;
int rc = MPI_SUCCESS; int rc = MPI_SUCCESS;
ompi_request_t **rptr = NULL; ompi_request_t **rptr = NULL;
ompi_request_t *request = NULL; ompi_request_t *request = NULL;
@ -369,7 +382,7 @@ int ompi_request_default_wait_some(size_t count,
rptr = requests; rptr = requests;
num_requests_null_inactive = 0; num_requests_null_inactive = 0;
num_requests_done = 0; num_requests_done = 0;
for (i = 0; i < count; i++, rptr++) { for (size_t i = 0; i < count; i++, rptr++) {
request = *rptr; request = *rptr;
/* /*
* Check for null or completed persistent request. * Check for null or completed persistent request.
@ -385,6 +398,12 @@ int ompi_request_default_wait_some(size_t count,
num_requests_done++; num_requests_done++;
} }
} }
if(num_requests_null_inactive == count) {
*outcount = MPI_UNDEFINED;
WAIT_SYNC_RELEASE(&sync);
return rc;
}
if( 0 != num_requests_done ) { if( 0 != num_requests_done ) {
/* As we only expect one trigger update the sync with count 1 */ /* As we only expect one trigger update the sync with count 1 */
wait_sync_update(&sync, 1, request->req_status.MPI_ERROR); wait_sync_update(&sync, 1, request->req_status.MPI_ERROR);
@ -403,62 +422,58 @@ int ompi_request_default_wait_some(size_t count,
} }
#endif #endif
if(num_requests_null_inactive == count) { /* Do the final counting and */
*outcount = MPI_UNDEFINED; /* Clean up the synchronization primitives */
} else {
/* Do the final counting and */ rptr = requests;
/* Clean up the synchronization primitives */ num_requests_done = 0;
for (size_t i = 0; i < count; i++, rptr++) {
request = *rptr;
rptr = requests; if( request->req_state == OMPI_REQUEST_INACTIVE ) {
num_requests_done = 0; continue;
for (i = 0; i < count; i++, rptr++) {
request = *rptr;
if( request->req_state == OMPI_REQUEST_INACTIVE ) {
continue;
}
if( !OPAL_ATOMIC_CMPSET_PTR(&request->req_complete, &sync, REQUEST_PENDING) ) {
assert(REQUEST_COMPLETE(request));
indices[num_requests_done] = i;
num_requests_done++;
}
} }
WAIT_SYNC_RELEASE(&sync); if( !OPAL_ATOMIC_CMPSET_PTR(&request->req_complete, &sync, REQUEST_PENDING) ) {
assert(REQUEST_COMPLETE(request));
indices[num_requests_done] = i;
num_requests_done++;
}
}
*outcount = num_requests_done; WAIT_SYNC_RELEASE(&sync);
for (i = 0; i < num_requests_done; i++) { *outcount = num_requests_done;
request = requests[indices[i]];
assert( REQUEST_COMPLETE(request) );
/* Per note above, we have to call gen request query_fn even
if STATUS_IGNORE was provided */
if (OMPI_REQUEST_GEN == request->req_type) {
ompi_grequest_invoke_query(request, &request->req_status);
}
if (MPI_STATUSES_IGNORE != statuses) {
statuses[i] = request->req_status;
}
if (MPI_SUCCESS != request->req_status.MPI_ERROR) { for (size_t i = 0; i < num_requests_done; i++) {
rc = MPI_ERR_IN_STATUS; request = requests[indices[i]];
} assert( REQUEST_COMPLETE(request) );
/* Per note above, we have to call gen request query_fn even
if STATUS_IGNORE was provided */
if (OMPI_REQUEST_GEN == request->req_type) {
ompi_grequest_invoke_query(request, &request->req_status);
}
if (MPI_STATUSES_IGNORE != statuses) {
statuses[i] = request->req_status;
}
if( request->req_persistent ) { if (MPI_SUCCESS != request->req_status.MPI_ERROR) {
request->req_state = OMPI_REQUEST_INACTIVE; rc = MPI_ERR_IN_STATUS;
} else { }
/* Only free the request if there was no error */
if (MPI_SUCCESS == request->req_status.MPI_ERROR) { if( request->req_persistent ) {
int tmp; request->req_state = OMPI_REQUEST_INACTIVE;
tmp = ompi_request_free(&(requests[indices[i]])); } else {
if (OMPI_SUCCESS != tmp) { /* Only free the request if there was no error */
return tmp; if (MPI_SUCCESS == request->req_status.MPI_ERROR) {
} int tmp;
tmp = ompi_request_free(&(requests[indices[i]]));
if (OMPI_SUCCESS != tmp) {
return tmp;
} }
} }
} }
} }
return rc; return rc;
} }