mpi: retain operation and datatype in non blocking collectives

MPI standard states a user MPI_Op and/or user MPI_Datatype can be free'd after a call to a non blocking collective and before the non-blocking collective completes. Retain user (only) MPI_Op and MPI_Datatype when the non blocking call is invoked, and set a request callback so they are free'd when the MPI_Request completes. Thanks Thomas Ponweiser for reporting this Fixes #2151 Fixes #1304 Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
open-mpi · Apr 9, 2019 · 646391b · 646391b
1 parent fd686d7
commit 646391b
Show file tree

Hide file tree

Showing 23 changed files with 306 additions and 57 deletions.
diff --git a/ompi/mca/coll/base/coll_base_util.c b/ompi/mca/coll/base/coll_base_util.c
@@ -9,8 +9,8 @@
  *                         University of Stuttgart.  All rights reserved.
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
- * Copyright (c) 2014-2017 Research Organization for Information Science
- *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2014-2019 Research Organization for Information Science
+ *                         and Technology (RIST).  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -29,6 +29,27 @@
 #include "ompi/mca/pml/pml.h"
 #include "coll_base_util.h"
 
+struct retain_op_data {
+    ompi_request_complete_fn_t req_complete_cb;
+    void *req_complete_cb_data;
+    ompi_op_t *op;
+    ompi_datatype_t *datatype;
+};
+
+struct retain_datatypes_data {
+    ompi_request_complete_fn_t req_complete_cb;
+    void *req_complete_cb_data;
+    ompi_datatype_t *stype;
+    ompi_datatype_t *rtype;
+};
+
+struct retain_datatypes_w_data {
+    ompi_request_complete_fn_t req_complete_cb;
+    void *req_complete_cb_data;
+    int count;
+    ompi_datatype_t *types[];
+};
+
 int ompi_coll_base_sendrecv_actual( const void* sendbuf, size_t scount,
                                     ompi_datatype_t* sdatatype,
                                     int dest, int stag,
@@ -103,3 +124,142 @@ int ompi_rounddown(int num, int factor)
     num /= factor;
     return num * factor;    /* floor(num / factor) * factor */
 }
+
+static int release_op_callback(struct ompi_request_t *request) {
+    struct retain_op_data * p = (struct retain_op_data *)request->req_complete_cb_data;
+    int rc = OMPI_SUCCESS;
+    assert (NULL != p);
+    if (NULL != p->req_complete_cb) {
+        request->req_complete_cb = p->req_complete_cb;
+        request->req_complete_cb_data = p->req_complete_cb_data;
+        rc = request->req_complete_cb(request);
+    }
+    if (NULL != p->op) {
+        OBJ_RELEASE(p->op);
+    }
+    if (NULL != p->datatype) {
+        OBJ_RELEASE(p->datatype);
+    }
+    free(p);
+    return rc;
+}
+
+int ompi_coll_base_retain_op( ompi_request_t *request, ompi_op_t *op,
+                              ompi_datatype_t *type) {
+    bool retain = !ompi_op_is_intrinsic(op);
+    retain |= !ompi_datatype_is_predefined(type);
+    if (OPAL_UNLIKELY(retain)) {
+        struct retain_op_data *p = (struct retain_op_data *)calloc(1, sizeof(struct retain_op_data));
+        if (OPAL_UNLIKELY(NULL == p)) {
+            return OMPI_ERR_OUT_OF_RESOURCE;
+        }
+        if (!ompi_op_is_intrinsic(op)) {
+            OBJ_RETAIN(op);
+            p->op = op;
+        }
+        if (!ompi_datatype_is_predefined(type)) {
+            OBJ_RETAIN(type);
+            p->datatype = type;
+        }
+        p->req_complete_cb = request->req_complete_cb;
+        p->req_complete_cb_data = request->req_complete_cb_data;
+        request->req_complete_cb = release_op_callback;
+        request->req_complete_cb_data = p;
+    }
+    return OMPI_SUCCESS;
+}
+
+static int release_datatypes_callback(struct ompi_request_t *request) {
+    struct retain_datatypes_data * p = (struct retain_datatypes_data *)request->req_complete_cb_data;
+    int rc = OMPI_SUCCESS;
+    assert (NULL != p);
+    if (NULL != p->req_complete_cb) {
+        request->req_complete_cb = p->req_complete_cb;
+        request->req_complete_cb_data = p->req_complete_cb_data;
+        rc = request->req_complete_cb(request);
+    }
+    if (NULL != p->stype) {
+        OBJ_RELEASE(p->stype);
+    }
+    if (NULL != p->rtype) {
+        OBJ_RELEASE(p->rtype);
+    }
+    free(p);
+    return rc;
+}
+
+int ompi_coll_base_retain_datatypes( ompi_request_t *request, ompi_datatype_t *stype,
+                                     ompi_datatype_t *rtype) {
+    bool retain = NULL != stype && !ompi_datatype_is_predefined(stype);
+    retain |= NULL != rtype && !ompi_datatype_is_predefined(rtype);
+    if (OPAL_UNLIKELY(retain)) {
+        struct retain_datatypes_data *p = (struct retain_datatypes_data *)calloc(1, sizeof(struct retain_datatypes_data));
+        if (OPAL_UNLIKELY(NULL == p)) {
+            return OMPI_ERR_OUT_OF_RESOURCE;
+        }
+        if (NULL != stype && !ompi_datatype_is_predefined(stype)) {
+            OBJ_RETAIN(stype);
+            p->stype = stype;
+        }
+        if (NULL != rtype && !ompi_datatype_is_predefined(rtype)) {
+            OBJ_RETAIN(rtype);
+            p->rtype = rtype;
+        }
+        p->req_complete_cb = request->req_complete_cb;
+        p->req_complete_cb_data = request->req_complete_cb_data;
+        request->req_complete_cb = release_datatypes_callback;
+        request->req_complete_cb_data = p;
+    }
+    return OMPI_SUCCESS;
+}
+
+static int release_datatypes_w_callback(struct ompi_request_t *request) {
+    struct retain_datatypes_w_data * p = (struct retain_datatypes_w_data *)request->req_complete_cb_data;
+    int rc = OMPI_SUCCESS;
+    assert (NULL != p);
+    if (NULL != p->req_complete_cb) {
+        request->req_complete_cb = p->req_complete_cb;
+        request->req_complete_cb_data = p->req_complete_cb_data;
+        rc = request->req_complete_cb(request);
+    }
+    for (int i=0; i<p->count; i++) {
+        OBJ_RELEASE(p->types[i]);
+    }
+    free(p);
+    return rc;
+}
+
+int ompi_coll_base_retain_datatypes_w( ompi_request_t *request, int count,
+                                       ompi_datatype_t *const stypes[], ompi_datatype_t *const rtypes[]) {
+    int datatypes = 0;
+    for (int i=0; i<count; i++) {
+        if (NULL != stypes[i] && !ompi_datatype_is_predefined(stypes[i])) {
+            datatypes++;
+        }
+        if (NULL != rtypes[i] && !ompi_datatype_is_predefined(rtypes[i])) {
+            datatypes++;
+        }
+    }
+    if (OPAL_UNLIKELY(0 < datatypes)) {
+        struct retain_datatypes_w_data *p = (struct retain_datatypes_w_data *)calloc(1, sizeof(struct retain_datatypes_data)+(datatypes-1)*sizeof(ompi_datatype_t *));
+        if (OPAL_UNLIKELY(NULL == p)) {
+            return OMPI_ERR_OUT_OF_RESOURCE;
+        }
+        datatypes = 0;
+        for (int i=0; i<count; i++) {
+            if (NULL != stypes[i] && !ompi_datatype_is_predefined(stypes[i])) {
+                p->types[datatypes++] = stypes[i];
+                OBJ_RETAIN(stypes[i]);
+            }
+            if (NULL != rtypes[i] && !ompi_datatype_is_predefined(rtypes[i])) {
+                p->types[datatypes++] = rtypes[i];
+                OBJ_RETAIN(rtypes[i]);
+            }
+        }
+        p->req_complete_cb = request->req_complete_cb;
+        p->req_complete_cb_data = request->req_complete_cb_data;
+        request->req_complete_cb = release_datatypes_w_callback;
+        request->req_complete_cb_data = p;
+    }
+    return OMPI_SUCCESS;
+}
diff --git a/ompi/mca/coll/base/coll_base_util.h b/ompi/mca/coll/base/coll_base_util.h
@@ -9,8 +9,8 @@
  *                         University of Stuttgart.  All rights reserved.
  * Copyright (c) 2004-2005 The Regents of the University of California.
  *                         All rights reserved.
- * Copyright (c) 2014-2017 Research Organization for Information Science
- *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2014-2019 Research Organization for Information Science
+ *                         and Technology (RIST).  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -27,6 +27,7 @@
 #include "ompi/mca/mca.h"
 #include "ompi/datatype/ompi_datatype.h"
 #include "ompi/request/request.h"
+#include "ompi/op/op.h"
 #include "ompi/mca/pml/pml.h"
 
 BEGIN_C_DECLS
@@ -84,5 +85,15 @@ unsigned int ompi_mirror_perm(unsigned int x, int nbits);
  */
 int ompi_rounddown(int num, int factor);
 
+int ompi_coll_base_retain_op( ompi_request_t *request, ompi_op_t *op,
+                           ompi_datatype_t *type);
+
+int ompi_coll_base_retain_datatypes( ompi_request_t *request, ompi_datatype_t *stype,
+                                     ompi_datatype_t *rtype);
+
+int ompi_coll_base_retain_datatypes_w( ompi_request_t *request, int count,
+                                       ompi_datatype_t *const stypes[],
+                                       ompi_datatype_t *const rtypes[]);
+
 END_C_DECLS
 #endif /* MCA_COLL_BASE_UTIL_EXPORT_H */
diff --git a/ompi/mpi/c/iallgather.c b/ompi/mpi/c/iallgather.c
@@ -14,8 +14,8 @@
  * Copyright (c) 2012      Oak Ridge National Laboratory. All rights reserved.
  * Copyright (c) 2013      Los Alamos National Security, LLC.  All rights
  *                         reserved.
- * Copyright (c) 2015      Research Organization for Information Science
- *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2015-2019 Research Organization for Information Science
+ *                         and Technology (RIST).  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -31,6 +31,7 @@
 #include "ompi/communicator/communicator.h"
 #include "ompi/errhandler/errhandler.h"
 #include "ompi/datatype/ompi_datatype.h"
+#include "ompi/mca/coll/base/coll_base_util.h"
 #include "ompi/memchecker.h"
 #include "ompi/runtime/ompi_spc.h"
 
@@ -102,6 +103,9 @@ int MPI_Iallgather(const void *sendbuf, int sendcount, MPI_Datatype sendtype,
     err = comm->c_coll->coll_iallgather(sendbuf, sendcount, sendtype,
                                        recvbuf, recvcount, recvtype, comm,
                                        request, comm->c_coll->coll_iallgather_module);
+    if (OPAL_LIKELY(OMPI_SUCCESS == err)) {
+        ompi_coll_base_retain_datatypes(*request, sendtype, recvtype);
+    }
 
     OMPI_ERRHANDLER_RETURN(err, comm, err, FUNC_NAME);
 }
diff --git a/ompi/mpi/c/iallgatherv.c b/ompi/mpi/c/iallgatherv.c
@@ -14,8 +14,8 @@
  * Copyright (c) 2012 Cisco Systems, Inc.  All rights reserved.
  * Copyright (c) 2012-2013 Los Alamos National Security, LLC.  All rights
  *                         reserved.
- * Copyright (c) 2015      Research Organization for Information Science
- *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2015-2019 Research Organization for Information Science
+ *                         and Technology (RIST).  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -31,6 +31,7 @@
 #include "ompi/communicator/communicator.h"
 #include "ompi/errhandler/errhandler.h"
 #include "ompi/datatype/ompi_datatype.h"
+#include "ompi/mca/coll/base/coll_base_util.h"
 #include "ompi/memchecker.h"
 #include "ompi/runtime/ompi_spc.h"
 
@@ -126,6 +127,9 @@ int MPI_Iallgatherv(const void *sendbuf, int sendcount, MPI_Datatype sendtype,
                                         recvbuf, recvcounts, displs,
                                         recvtype, comm, request,
                                         comm->c_coll->coll_iallgatherv_module);
+    if (OPAL_LIKELY(OMPI_SUCCESS == err)) {
+        ompi_coll_base_retain_datatypes(*request, sendtype, recvtype);
+    }
     OMPI_ERRHANDLER_RETURN(err, comm, err, FUNC_NAME);
 }
 
diff --git a/ompi/mpi/c/iallreduce.c b/ompi/mpi/c/iallreduce.c
@@ -12,8 +12,8 @@
  *                         All rights reserved.
  * Copyright (c) 2013      Los Alamos National Security, LLC.  All rights
  *                         reserved.
- * Copyright (c) 2015      Research Organization for Information Science
- *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2015-2019 Research Organization for Information Science
+ *                         and Technology (RIST).  All rights reserved.
  * Copyright (c) 2016      IBM Corporation.  All rights reserved.
  * $COPYRIGHT$
  *
@@ -31,6 +31,7 @@
 #include "ompi/errhandler/errhandler.h"
 #include "ompi/datatype/ompi_datatype.h"
 #include "ompi/op/op.h"
+#include "ompi/mca/coll/base/coll_base_util.h"
 #include "ompi/memchecker.h"
 #include "ompi/runtime/ompi_spc.h"
 
@@ -112,10 +113,11 @@ int MPI_Iallreduce(const void *sendbuf, void *recvbuf, int count,
 
     /* Invoke the coll component to perform the back-end operation */
 
-    OBJ_RETAIN(op);
     err = comm->c_coll->coll_iallreduce(sendbuf, recvbuf, count, datatype,
                                        op, comm, request, comm->c_coll->coll_iallreduce_module);
-    OBJ_RELEASE(op);
+    if (OPAL_LIKELY(OMPI_SUCCESS == err)) {
+        ompi_coll_base_retain_op(*request, op, datatype);
+    }
     OMPI_ERRHANDLER_RETURN(err, comm, err, FUNC_NAME);
 }
 
diff --git a/ompi/mpi/c/ialltoall.c b/ompi/mpi/c/ialltoall.c
@@ -14,8 +14,8 @@
  * Copyright (c) 2012      Oak Ridge National Laboratory. All rights reserved.
  * Copyright (c) 2013      Los Alamos National Security, LLC.  All rights
  *                         reserved.
- * Copyright (c) 2014-2016 Research Organization for Information Science
- *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2014-2019 Research Organization for Information Science
+ *                         and Technology (RIST).  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -31,6 +31,7 @@
 #include "ompi/communicator/communicator.h"
 #include "ompi/errhandler/errhandler.h"
 #include "ompi/datatype/ompi_datatype.h"
+#include "ompi/mca/coll/base/coll_base_util.h"
 #include "ompi/memchecker.h"
 #include "ompi/runtime/ompi_spc.h"
 
@@ -101,5 +102,8 @@ int MPI_Ialltoall(const void *sendbuf, int sendcount, MPI_Datatype sendtype,
     err = comm->c_coll->coll_ialltoall(sendbuf, sendcount, sendtype,
                                       recvbuf, recvcount, recvtype, comm,
                                       request, comm->c_coll->coll_ialltoall_module);
+    if (OPAL_LIKELY(OMPI_SUCCESS == err)) {
+        ompi_coll_base_retain_datatypes(*request, sendtype, recvtype);
+    }
     OMPI_ERRHANDLER_RETURN(err, comm, err, FUNC_NAME);
 }
diff --git a/ompi/mpi/c/ialltoallv.c b/ompi/mpi/c/ialltoallv.c
@@ -13,8 +13,8 @@
  * Copyright (c) 2007      Cisco Systems, Inc.  All rights reserved.
  * Copyright (c) 2012-2013 Los Alamos National Security, LLC.  All rights
  *                         reserved.
- * Copyright (c) 2014-2016 Research Organization for Information Science
- *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2014-2019 Research Organization for Information Science
+ *                         and Technology (RIST).  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -30,6 +30,7 @@
 #include "ompi/communicator/communicator.h"
 #include "ompi/errhandler/errhandler.h"
 #include "ompi/datatype/ompi_datatype.h"
+#include "ompi/mca/coll/base/coll_base_util.h"
 #include "ompi/memchecker.h"
 #include "ompi/runtime/ompi_spc.h"
 
@@ -130,6 +131,9 @@ int MPI_Ialltoallv(const void *sendbuf, const int sendcounts[], const int sdispl
     err = comm->c_coll->coll_ialltoallv(sendbuf, sendcounts, sdispls,
                                        sendtype, recvbuf, recvcounts, rdispls,
                                        recvtype, comm, request, comm->c_coll->coll_ialltoallv_module);
+    if (OPAL_LIKELY(OMPI_SUCCESS == err)) {
+        ompi_coll_base_retain_datatypes(*request, sendtype, recvtype);
+    }
     OMPI_ERRHANDLER_RETURN(err, comm, err, FUNC_NAME);
 }
 
diff --git a/ompi/mpi/c/ialltoallw.c b/ompi/mpi/c/ialltoallw.c
@@ -13,8 +13,8 @@
  * Copyright (c) 2007      Cisco Systems, Inc.  All rights reserved.
  * Copyright (c) 2012-2013 Los Alamos National Security, LLC.  All rights
  *                         reserved.
- * Copyright (c) 2014-2016 Research Organization for Information Science
- *                         and Technology (RIST). All rights reserved.
+ * Copyright (c) 2014-2019 Research Organization for Information Science
+ *                         and Technology (RIST).  All rights reserved.
  * $COPYRIGHT$
  *
  * Additional copyrights may follow
@@ -30,6 +30,7 @@
 #include "ompi/communicator/communicator.h"
 #include "ompi/errhandler/errhandler.h"
 #include "ompi/datatype/ompi_datatype.h"
+#include "ompi/mca/coll/base/coll_base_util.h"
 #include "ompi/memchecker.h"
 #include "ompi/runtime/ompi_spc.h"
 
@@ -127,6 +128,12 @@ int MPI_Ialltoallw(const void *sendbuf, const int sendcounts[], const int sdispl
                                        sendtypes, recvbuf, recvcounts,
                                        rdispls, recvtypes, comm, request,
                                        comm->c_coll->coll_ialltoallw_module);
+    if (OPAL_LIKELY(OMPI_SUCCESS == err)) {
+        ompi_coll_base_retain_datatypes_w(*request,
+                                          OMPI_COMM_IS_INTER(comm)?ompi_comm_remote_size(comm):ompi_comm_size(comm),
+                                          sendtypes,
+                                          recvtypes);
+    }
     OMPI_ERRHANDLER_RETURN(err, comm, err, FUNC_NAME);
 }