1
1
openmpi/ompi/mpi/c/reduce_scatter.c

135 строки
4.6 KiB
C
Исходник Обычный вид История

/*
* Copyright (c) 2004-2007 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2012 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* Copyright (c) 2004-2008 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 (c) 2006-2012 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Los Alamos National Security, LLC. All rights
* reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include <stdio.h>
#include "ompi/mpi/c/bindings.h"
#include "ompi/runtime/params.h"
#include "ompi/communicator/communicator.h"
#include "ompi/errhandler/errhandler.h"
- Split the datatype engine into two parts: an MPI specific part in OMPI and a language agnostic part in OPAL. The convertor is completely moved into OPAL. This offers several benefits as described in RFC http://www.open-mpi.org/community/lists/devel/2009/07/6387.php namely: - Fewer basic types (int* and float* types, boolean and wchar - Fixing naming scheme to ompi-nomenclature. - Usability outside of the ompi-layer. - Due to the fixed nature of simple opal types, their information is completely known at compile time and therefore constified - With fewer datatypes (22), the actual sizes of bit-field types may be reduced from 64 to 32 bits, allowing reorganizing the opal_datatype structure, eliminating holes and keeping data required in convertor (upon send/recv) in one cacheline... This has implications to the convertor-datastructure and other parts of the code. - Several performance tests have been run, the netpipe latency does not change with this patch on Linux/x86-64 on the smoky cluster. - Extensive tests have been done to verify correctness (no new regressions) using: 1. mpi_test_suite on linux/x86-64 using clean ompi-trunk and ompi-ddt: a. running both trunk and ompi-ddt resulted in no differences (except for MPI_SHORT_INT and MPI_TYPE_MIX_LB_UB do now run correctly). b. with --enable-memchecker and running under valgrind (one buglet when run with static found in test-suite, commited) 2. ibm testsuite on linux/x86-64 using clean ompi-trunk and ompi-ddt: all passed (except for the dynamic/ tests failed!! as trunk/MTT) 3. compilation and usage of HDF5 tests on Jaguar using PGI and PathScale compilers. 4. compilation and usage on Scicortex. - Please note, that for the heterogeneous case, (-m32 compiled binaries/ompi), neither ompi-trunk, nor ompi-ddt branch would successfully launch. This commit was SVN r21641.
2009-07-13 08:56:31 +04:00
#include "ompi/datatype/ompi_datatype.h"
#include "ompi/op/op.h"
#include "ompi/memchecker.h"
#if OPAL_HAVE_WEAK_SYMBOLS && OMPI_PROFILING_DEFINES
#pragma weak MPI_Reduce_scatter = PMPI_Reduce_scatter
#endif
#if OMPI_PROFILING_DEFINES
#include "ompi/mpi/c/profile/defines.h"
#endif
static const char FUNC_NAME[] = "MPI_Reduce_scatter";
int MPI_Reduce_scatter(void *sendbuf, void *recvbuf, int recvcounts[],
MPI_Datatype datatype, MPI_Op op, MPI_Comm comm)
{
int i, err, size, count;
MEMCHECKER(
int rank;
size = ompi_comm_size(comm);
rank = ompi_comm_rank(comm);
for (count = i = 0; i < size; ++i) {
if (0 == recvcounts[i]) {
count += recvcounts[i];
}
}
memchecker_comm(comm);
memchecker_datatype(datatype);
/* check receive buffer of current proccess, whether it's addressable. */
memchecker_call(&opal_memchecker_base_isaddressable, recvbuf,
recvcounts[rank], datatype);
/* check whether the actual send buffer is defined. */
if(MPI_IN_PLACE == sendbuf) {
memchecker_call(&opal_memchecker_base_isdefined, recvbuf, count, datatype);
} else {
memchecker_call(&opal_memchecker_base_isdefined, sendbuf, count, datatype);
}
);
if (MPI_PARAM_CHECK) {
char *msg;
err = MPI_SUCCESS;
OMPI_ERR_INIT_FINALIZE(FUNC_NAME);
if (ompi_comm_invalid(comm)) {
return OMPI_ERRHANDLER_INVOKE(MPI_COMM_WORLD, MPI_ERR_COMM,
FUNC_NAME);
}
/* Unrooted operation; same checks for all ranks on both
intracommunicators and intercommunicators */
else if (MPI_OP_NULL == op || NULL == op) {
err = MPI_ERR_OP;
} else if (!ompi_op_is_valid(op, datatype, &msg, FUNC_NAME)) {
int ret = OMPI_ERRHANDLER_INVOKE(comm, MPI_ERR_OP, msg);
free(msg);
return ret;
} else if (NULL == recvcounts) {
err = MPI_ERR_COUNT;
} else if (MPI_IN_PLACE == recvbuf) {
err = MPI_ERR_ARG;
}
OMPI_ERRHANDLER_CHECK(err, comm, err, FUNC_NAME);
/* Based on the standard each group has to provide the same total
number of elements, so the size of the recvcounts array depends
on the number of participants in the local group. */
size = ompi_comm_size(comm);
for (i = 0; i < size; ++i) {
OMPI_CHECK_DATATYPE_FOR_SEND(err, datatype, recvcounts[i]);
OMPI_ERRHANDLER_CHECK(err, comm, err, FUNC_NAME);
}
}
/* MPI-1, p114, says that each process must supply at least one
element. But at least the Pallas benchmarks call MPI_REDUCE
with a count of 0. So be sure to handle it. Grrr... */
size = ompi_comm_size(comm);
for (count = i = 0; i < size; ++i) {
if (0 == recvcounts[i]) {
++count;
}
}
if (size == count) {
return MPI_SUCCESS;
}
OPAL_CR_ENTER_LIBRARY();
/* Invoke the coll component to perform the back-end operation */
OBJ_RETAIN(op);
err = comm->c_coll.coll_reduce_scatter(sendbuf, recvbuf, recvcounts,
datatype, op, comm,
comm->c_coll.coll_reduce_scatter_module);
OBJ_RELEASE(op);
OMPI_ERRHANDLER_RETURN(err, comm, err, FUNC_NAME);
}