Merge pull request #5044 from hzhou/2101_coll_large

coll: internally use MPI_Aint for count parameters

Approved-by: Ken Raffenetti <raffenet@mcs.anl.gov>

maint/local_python/binding_c.py

@@ -885,10 +885,20 @@ def push_impl_decl(func, impl_name=None):
         impl_name = re.sub(r'^MPIX?_', 'MPIR_', func['name']) + "_impl"
     if func['impl_param_list']:
         params = ', '.join(func['impl_param_list'])
-        if func['dir'] == 'coll' and not RE.match(r'MPI_I', func['name']):
-            params = params + ", MPIR_Errflag_t *errflag"
+        if func['dir'] == 'coll':
+            # All collective impl function use MPI_Aint counts
+            params = re.sub(r' int (count|sendcount|recvcount),', r' MPI_Aint \1,', params)
+            # block collective use an extra errflag
+            if not RE.match(r'MPI_(I|Neighbor)', func['name']):
+                params = params + ", MPIR_Errflag_t *errflag"
     else:
         params="void"
+
+    if func['dir'] == 'coll':
+        # collective also dump MPIR_Xxx(...)
+        mpir_name = re.sub(r'^MPIX?_', 'MPIR_', func['name'])
+        G.impl_declares.append("int %s(%s);" % (mpir_name, params))
+    # dump MPIR_Xxx_impl(...)
     G.impl_declares.append("int %s(%s);" % (impl_name, params))
 
 def dump_CHECKENUM(var, errname, t, type="ENUM"):
@@ -907,23 +917,24 @@ def dump_CHECKENUM(var, errname, t, type="ENUM"):
 
 def dump_body_coll(func):
     # collectives call MPIR_Xxx
-    RE.match(r'MPI_(\w+)', func['name'])
-    name = RE.m.group(1)
+    mpir_name = re.sub(r'^MPIX?_', 'MPIR_', func['name'])
 
     args = ", ".join(func['impl_arg_list'])
-    if name.startswith('I'):
+    if RE.match(r'mpi_i', func['name'], re.IGNORECASE):
         # non-blocking collectives
         G.out.append("MPIR_Request *request_ptr = NULL;")
-        dump_line_with_break("mpi_errno = MPIR_%s(%s);" % (name, args))
+        dump_line_with_break("mpi_errno = %s(%s);" % (mpir_name, args))
         dump_error_check("")
         G.out.append("if (!request_ptr) {")
         G.out.append("    request_ptr = MPIR_Request_create_complete(MPIR_REQUEST_KIND__COLL);")
         G.out.append("}")
         G.out.append("*request = request_ptr->handle;")
+    elif RE.match(r'mpi_neighbor_', func['name'], re.IGNORECASE):
+        dump_line_with_break("mpi_errno = %s(%s);" % (mpir_name, args))
     else:
         # blocking collectives
         G.out.append("MPIR_Errflag_t errflag = MPIR_ERR_NONE;")
-        dump_line_with_break("mpi_errno = MPIR_%s(%s, &errflag);" % (name, args))
+        dump_line_with_break("mpi_errno = %s(%s, &errflag);" % (mpir_name, args))
 
 def dump_body_topo_fns(func, method):
     comm_ptr = func['_has_comm'] + "_ptr"

src/binding/c/coll_api.txt

@@ -169,55 +169,27 @@ MPI_Iscatterv:
     .desc: Scatters a buffer in parts to all processes in a communicator in a nonblocking way
 
 MPI_Neighbor_allgather:
-    .desc: In this function, each process i gathers data items from each process j if an edge (j,i) exists in the topology graph, and each process i sends the same data items to all processes j where an edge (i,j) exists. The send buffer is sent to each neighboring process and the l-th block in the receive buffer is received from the l-th neighbor.
-{
-    mpi_errno = MPIR_Neighbor_allgather(sendbuf, sendcount, sendtype, recvbuf,
-                                        recvcount, recvtype, comm_ptr);
-    if (mpi_errno) {
-        goto fn_fail;
-    }
-}
+    .desc: Gathers data from all neighboring processes and distribute the combined data to all neighboring processes
+/*
+    Notes:
+    In this function, each process i gathers data items from each process j if an edge (j,i) exists in the topology graph, and each process i sends the same data items to all processes j where an edge (i,j) exists. The send buffer is sent to each neighboring process and the l-th block in the receive buffer is received from the l-th neighbor.
+*/
 
 MPI_Neighbor_allgatherv:
     .desc: The vector variant of MPI_Neighbor_allgather.
-{
-    mpi_errno = MPIR_Neighbor_allgatherv(sendbuf, sendcount, sendtype, recvbuf,
-                                         recvcounts, displs, recvtype, comm_ptr);
-    if (mpi_errno) {
-        goto fn_fail;
-    }
-}
 
 MPI_Neighbor_alltoall:
-    .desc: In this function, each process i receives data items from each process j if an edge (j,i) exists in the topology graph or Cartesian topology.  Similarly, each process i sends data items to all processes j where an edge (i,j) exists. This call is more general than MPI_NEIGHBOR_ALLGATHER in that different data items can be sent to each neighbor. The k-th block in send buffer is sent to the k-th neighboring process and the l-th block in the receive buffer is received from the l-th neighbor.
-{
-    mpi_errno = MPIR_Neighbor_alltoall(sendbuf, sendcount, sendtype, recvbuf,
-                                       recvcount, recvtype, comm_ptr);
-    if (mpi_errno) {
-        goto fn_fail;
-    }
-}
+    .desc: Sends and Receivs data from all neighboring processes
+/*
+    Notes:
+    In this function, each process i receives data items from each process j if an edge (j,i) exists in the topology graph or Cartesian topology.  Similarly, each process i sends data items to all processes j where an edge (i,j) exists. This call is more general than MPI_NEIGHBOR_ALLGATHER in that different data items can be sent to each neighbor. The k-th block in send buffer is sent to the k-th neighboring process and the l-th block in the receive buffer is received from the l-th neighbor.
+*/
 
 MPI_Neighbor_alltoallv:
     .desc: The vector variant of MPI_Neighbor_alltoall allows sending/receiving different numbers of elements to and from each neighbor.
-{
-    mpi_errno = MPIR_Neighbor_alltoallv(sendbuf, sendcounts, sdispls, sendtype,
-                                        recvbuf, recvcounts, rdispls, recvtype, comm_ptr);
-    if (mpi_errno) {
-        goto fn_fail;
-    }
-}
 
 MPI_Neighbor_alltoallw:
     .desc: Like MPI_Neighbor_alltoallv but it allows one to send and receive with different types to and from each neighbor.
-{
-    mpi_errno = MPIR_Neighbor_alltoallw_impl(sendbuf, sendcounts, sdispls,
-                                             sendtypes, recvbuf, recvcounts,
-                                             rdispls, recvtypes, comm_ptr);
-    if (mpi_errno) {
-        goto fn_fail;
-    }
-}
 
 MPI_Reduce:
     .desc: Reduces values on all processes to a single value