1
1
openmpi/ompi/mca/coll/inter/coll_inter_allreduce.c

115 строки
3.6 KiB
C
Исходник Обычный вид История

/*
* 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.
2015-06-24 06:59:57 +03:00
* 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 (c) 2006-2007 University of Houston. All rights reserved.
* Copyright (c) 2013 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2015 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
* $COPYRIGHT$
2015-06-24 06:59:57 +03:00
*
* Additional copyrights may follow
2015-06-24 06:59:57 +03:00
*
* $HEADER$
*/
#include "ompi_config.h"
#include "coll_inter.h"
#include "mpi.h"
#include "ompi/constants.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/communicator/communicator.h"
#include "ompi/request/request.h"
#include "ompi/op/op.h"
#include "ompi/mca/coll/coll.h"
#include "ompi/mca/coll/base/coll_tags.h"
#include "ompi/mca/pml/pml.h"
/*
* allreduce_inter
*
* Function: - allreduce using other MPI collectives
* Accepts: - same as MPI_Allreduce()
* Returns: - MPI_SUCCESS or error code
*/
int
mca_coll_inter_allreduce_inter(const void *sbuf, void *rbuf, int count,
struct ompi_datatype_t *dtype,
struct ompi_op_t *op,
struct ompi_communicator_t *comm,
mca_coll_base_module_t *module)
{
int err, rank, root = 0;
ptrdiff_t lb, extent;
char *tmpbuf = NULL, *pml_buffer = NULL;
ompi_request_t *req[2];
rank = ompi_comm_rank(comm);
2015-06-24 06:59:57 +03:00
/* Perform the reduction locally */
- 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
err = ompi_datatype_get_extent(dtype, &lb, &extent);
if (OMPI_SUCCESS != err) {
return OMPI_ERROR;
}
2015-06-24 06:59:57 +03:00
tmpbuf = (char *) malloc(count * extent);
if (NULL == tmpbuf) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
pml_buffer = tmpbuf - lb;
2015-06-24 06:59:57 +03:00
err = comm->c_local_comm->c_coll.coll_reduce(sbuf, pml_buffer, count,
2015-06-24 06:59:57 +03:00
dtype, op, root,
comm->c_local_comm,
comm->c_local_comm->c_coll.coll_reduce_module);
if (OMPI_SUCCESS != err) {
goto exit;
}
if (rank == root) {
/* Do a send-recv between the two root procs. to avoid deadlock */
err = MCA_PML_CALL(irecv(rbuf, count, dtype, 0,
MCA_COLL_BASE_TAG_ALLREDUCE, comm,
&(req[0])));
if (OMPI_SUCCESS != err) {
goto exit;
}
err = MCA_PML_CALL(isend(pml_buffer, count, dtype, 0,
MCA_COLL_BASE_TAG_ALLREDUCE,
MCA_PML_BASE_SEND_STANDARD,
comm, &(req[1])));
if (OMPI_SUCCESS != err) {
goto exit;
}
err = ompi_request_wait_all(2, req, MPI_STATUSES_IGNORE);
if (OMPI_SUCCESS != err) {
goto exit;
}
}
/* bcast the message to all the local processes */
2015-06-24 06:59:57 +03:00
err = comm->c_local_comm->c_coll.coll_bcast(rbuf, count, dtype,
root, comm->c_local_comm,
comm->c_local_comm->c_coll.coll_bcast_module);
if (OMPI_SUCCESS != err) {
goto exit;
}
exit:
if (NULL != tmpbuf) {
free(tmpbuf);
}
return err;
}