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

141 строка
4.2 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) 2006-2010 University of Houston. All rights reserved.
* Copyright (c) 2015-2016 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/mca/coll/coll.h"
#include "ompi/mca/coll/base/coll_tags.h"
#include "ompi/mca/pml/pml.h"
/*
* scatterv_inter
*
* Function: - scatterv operation
* Accepts: - same arguments as MPI_Scatterv()
* Returns: - MPI_SUCCESS or error code
*/
int
mca_coll_inter_scatterv_inter(const void *sbuf, const int *scounts,
const int *disps, struct ompi_datatype_t *sdtype,
void *rbuf, int rcount,
struct ompi_datatype_t *rdtype, int root,
struct ompi_communicator_t *comm,
mca_coll_base_module_t *module)
{
int i, rank, size, err, total=0, size_local;
int *counts=NULL,*displace=NULL;
char *ptmp_free=NULL, *ptmp=NULL;
ompi_datatype_t *ndtype;
/* Initialize */
rank = ompi_comm_rank(comm);
size = ompi_comm_remote_size(comm);
size_local = ompi_comm_size(comm);
if (MPI_PROC_NULL == root) {
/* do nothing */
err = OMPI_SUCCESS;
} else if (MPI_ROOT != root) {
if(0 == rank) {
/* local root recieves the counts from the root */
counts = (int *)malloc(sizeof(int) * size_local);
err = MCA_PML_CALL(recv(counts, size_local, MPI_INT,
root, MCA_COLL_BASE_TAG_SCATTERV,
comm, MPI_STATUS_IGNORE));
if (OMPI_SUCCESS != err) {
return err;
}
/* calculate the whole buffer size and receive it from root */
for (i = 0; i < size_local; i++) {
total = total + counts[i];
}
if ( total > 0 ) {
ptrdiff_t gap, span;
span = opal_datatype_span(&rdtype->super, total, &gap);
ptmp_free = (char*)malloc(span);
if (NULL == ptmp_free) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
ptmp = ptmp_free - gap;
}
err = MCA_PML_CALL(recv(ptmp, total, rdtype,
root, MCA_COLL_BASE_TAG_SCATTERV,
comm, MPI_STATUS_IGNORE));
if (OMPI_SUCCESS != err) {
return err;
}
/* set the local displacement i.e. no displacements here */
displace = (int *)malloc(sizeof(int) * size_local);
displace[0] = 0;
for (i = 1; i < size_local; i++) {
displace[i] = displace[i-1] + counts[i-1];
}
}
/* perform the scatterv locally */
2015-06-24 06:59:57 +03:00
err = comm->c_local_comm->c_coll.coll_scatterv(ptmp, counts, displace,
rdtype, rbuf, rcount,
rdtype, 0, comm->c_local_comm,
comm->c_local_comm->c_coll.coll_scatterv_module);
if (OMPI_SUCCESS != err) {
return err;
}
if (NULL != ptmp_free) {
free(ptmp_free);
}
if (NULL != displace) {
free(displace);
}
if (NULL != counts) {
free(counts);
}
} else {
err = MCA_PML_CALL(send(scounts, size, MPI_INT, 0,
MCA_COLL_BASE_TAG_SCATTERV,
MCA_PML_BASE_SEND_STANDARD, comm));
if (OMPI_SUCCESS != err) {
return err;
}
- 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_create_indexed(size,scounts,disps,sdtype,&ndtype);
ompi_datatype_commit(&ndtype);
2015-06-24 06:59:57 +03:00
err = MCA_PML_CALL(send(sbuf, 1, ndtype, 0,
MCA_COLL_BASE_TAG_SCATTERV,
MCA_PML_BASE_SEND_STANDARD, comm));
if (OMPI_SUCCESS != err) {
return err;
}
- 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_destroy(&ndtype);
}
/* All done */
return err;
}