2004-12-27 11:56:33 +03:00
|
|
|
/*
|
2005-11-05 22:57:48 +03:00
|
|
|
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
|
|
|
* University Research and Technology
|
|
|
|
* Corporation. All rights reserved.
|
2006-08-24 20:38:08 +04:00
|
|
|
* Copyright (c) 2004-2006 The University of Tennessee and The University
|
2005-11-05 22:57:48 +03:00
|
|
|
* of Tennessee Research Foundation. All rights
|
|
|
|
* reserved.
|
2004-12-27 11:56:33 +03:00
|
|
|
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
|
|
|
* University of Stuttgart. All rights reserved.
|
2005-03-24 15:43:37 +03:00
|
|
|
* Copyright (c) 2004-2005 The Regents of the University of California.
|
|
|
|
* All rights reserved.
|
2007-08-24 00:41:29 +04:00
|
|
|
* Copyright (c) 2007 University of Houston. All rights reserved.
|
2004-12-27 11:56:33 +03:00
|
|
|
* $COPYRIGHT$
|
|
|
|
*
|
|
|
|
* Additional copyrights may follow
|
|
|
|
*
|
|
|
|
* $HEADER$
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "ompi_config.h"
|
|
|
|
#include "coll_hierarch.h"
|
|
|
|
|
|
|
|
#include <stdio.h>
|
|
|
|
|
|
|
|
#include "mpi.h"
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "ompi/constants.h"
|
|
|
|
#include "ompi/communicator/communicator.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"
|
2006-02-12 04:33:29 +03:00
|
|
|
#include "ompi/mca/coll/coll.h"
|
2004-12-27 11:56:33 +03:00
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* reduce_intra
|
|
|
|
*
|
2005-10-13 21:58:13 +04:00
|
|
|
* Function: - reduction using two level hierarchy algorithm
|
2004-12-27 11:56:33 +03:00
|
|
|
* Accepts: - same as MPI_Reduce()
|
|
|
|
* Returns: - MPI_SUCCESS or error code
|
|
|
|
*/
|
2005-04-13 22:35:07 +04:00
|
|
|
int mca_coll_hierarch_reduce_intra(void *sbuf, void *rbuf, int count,
|
|
|
|
struct ompi_datatype_t *dtype,
|
|
|
|
struct ompi_op_t *op,
|
2007-08-24 00:41:29 +04:00
|
|
|
int root, struct ompi_communicator_t *comm,
|
2008-07-29 02:40:57 +04:00
|
|
|
mca_coll_base_module_t *module)
|
2005-04-13 22:35:07 +04:00
|
|
|
{
|
2005-10-13 21:58:13 +04:00
|
|
|
struct ompi_communicator_t *llcomm=NULL;
|
|
|
|
struct ompi_communicator_t *lcomm=NULL;
|
2007-08-24 00:41:29 +04:00
|
|
|
mca_coll_hierarch_module_t *hierarch_module = (mca_coll_hierarch_module_t *) module;
|
2005-10-26 16:56:56 +04:00
|
|
|
int rank;
|
2005-10-15 21:04:01 +04:00
|
|
|
int lroot, llroot;
|
2006-10-18 00:20:58 +04:00
|
|
|
ptrdiff_t extent, true_extent, lb, true_lb;
|
2005-10-13 21:58:13 +04:00
|
|
|
char *tmpbuf=NULL, *tbuf=NULL;
|
2005-10-26 16:56:56 +04:00
|
|
|
int ret=OMPI_SUCCESS;
|
2005-10-13 21:58:13 +04:00
|
|
|
|
|
|
|
rank = ompi_comm_rank ( comm );
|
2007-08-24 00:41:29 +04:00
|
|
|
lcomm = hierarch_module->hier_lcomm;
|
2005-10-13 21:58:13 +04:00
|
|
|
|
2005-10-25 22:34:47 +04:00
|
|
|
if ( mca_coll_hierarch_verbose_param ) {
|
2005-10-15 23:36:54 +04:00
|
|
|
printf("%s:%d: executing hierarchical reduce with cnt=%d and root=%d\n",
|
|
|
|
comm->c_name, rank, count, root );
|
|
|
|
}
|
|
|
|
|
2007-08-24 00:41:29 +04:00
|
|
|
llcomm = mca_coll_hierarch_get_llcomm ( root, hierarch_module, &llroot, &lroot);
|
2005-10-14 22:11:21 +04:00
|
|
|
|
2005-10-13 21:58:13 +04:00
|
|
|
if ( MPI_COMM_NULL != lcomm ) {
|
- 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
|
|
|
ompi_datatype_get_extent(dtype, &lb, &extent);
|
|
|
|
ompi_datatype_get_true_extent(dtype, &true_lb, &true_extent);
|
2005-10-18 23:20:48 +04:00
|
|
|
|
2006-08-24 20:38:08 +04:00
|
|
|
tbuf = (char*)malloc(true_extent + (count - 1) * extent);
|
2005-10-18 23:20:48 +04:00
|
|
|
if (NULL == tbuf) {
|
|
|
|
return OMPI_ERR_OUT_OF_RESOURCE;
|
|
|
|
}
|
|
|
|
tmpbuf = tbuf - lb;
|
|
|
|
|
2005-10-13 21:58:13 +04:00
|
|
|
|
2005-10-18 23:20:48 +04:00
|
|
|
if ( MPI_IN_PLACE != sbuf ) {
|
|
|
|
ret = lcomm->c_coll.coll_reduce (sbuf, tmpbuf, count, dtype,
|
2007-08-24 00:41:29 +04:00
|
|
|
op, lroot, lcomm,
|
|
|
|
lcomm->c_coll.coll_reduce_module);
|
2005-10-18 23:20:48 +04:00
|
|
|
}
|
|
|
|
else {
|
|
|
|
ret = lcomm->c_coll.coll_reduce (rbuf, tmpbuf, count, dtype,
|
2007-08-24 00:41:29 +04:00
|
|
|
op, lroot, lcomm,
|
2008-04-23 21:32:05 +04:00
|
|
|
lcomm->c_coll.coll_reduce_module);
|
2005-10-18 23:20:48 +04:00
|
|
|
}
|
|
|
|
if ( OMPI_SUCCESS != ret ) {
|
|
|
|
goto exit;
|
|
|
|
}
|
2005-10-13 21:58:13 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
if ( MPI_UNDEFINED != llroot ) {
|
2005-10-18 23:20:48 +04:00
|
|
|
if ( MPI_COMM_NULL != lcomm ) {
|
2005-10-13 21:58:13 +04:00
|
|
|
ret = llcomm->c_coll.coll_reduce (tmpbuf, rbuf, count, dtype,
|
2007-08-24 00:41:29 +04:00
|
|
|
op, llroot, llcomm,
|
2008-04-23 21:23:30 +04:00
|
|
|
llcomm->c_coll.coll_reduce_module);
|
2005-10-18 23:20:48 +04:00
|
|
|
}
|
|
|
|
else {
|
|
|
|
ret = llcomm->c_coll.coll_reduce (sbuf, rbuf, count, dtype,
|
2007-08-24 00:41:29 +04:00
|
|
|
op, llroot, llcomm,
|
2008-04-23 21:23:30 +04:00
|
|
|
llcomm->c_coll.coll_reduce_module);
|
2005-10-18 23:20:48 +04:00
|
|
|
}
|
2005-10-13 21:58:13 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
exit:
|
|
|
|
if ( NULL != tmpbuf ) {
|
|
|
|
free ( tmpbuf );
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
2004-12-27 11:56:33 +03:00
|
|
|
}
|