1
1
openmpi/ompi/mca/coll/basic/coll_basic_exscan.c

135 строки
4.4 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) 2004-2006 The University of Tennessee and The University
* of Tennessee Research Foundation. All rights
* reserved.
* 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) 2011 NVIDIA Corporation. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include "coll_basic.h"
#include <stdio.h>
#include "mpi.h"
#include "ompi/constants.h"
#include "ompi/op/op.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/pml/pml.h"
#include "ompi/mca/coll/coll.h"
#include "ompi/mca/coll/base/coll_tags.h"
#include "coll_basic.h"
/*
* exscan_intra
*
* Function: - basic exscan operation
* Accepts: - same arguments as MPI_Exscan()
* Returns: - MPI_SUCCESS or error code
*/
int
mca_coll_basic_exscan_intra(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 size, rank, err;
ptrdiff_t true_lb, true_extent, lb, extent;
char *free_buffer = NULL;
char *reduce_buffer = NULL;
rank = ompi_comm_rank(comm);
size = ompi_comm_size(comm);
/* For MPI_IN_PLACE, just adjust send buffer to point to
* receive buffer. */
if (MPI_IN_PLACE == sbuf) {
sbuf = rbuf;
}
/* If we're rank 0, then just send our sbuf to the next rank, and
* we are done. */
if (0 == rank) {
return MCA_PML_CALL(send(sbuf, count, dtype, rank + 1,
MCA_COLL_BASE_TAG_EXSCAN,
MCA_PML_BASE_SEND_STANDARD, comm));
}
/* If we're the last rank, then just receive the result from the
* prior rank, and we are done. */
else if ((size - 1) == rank) {
return MCA_PML_CALL(recv(rbuf, count, dtype, rank - 1,
MCA_COLL_BASE_TAG_EXSCAN, comm,
MPI_STATUS_IGNORE));
}
/* Otherwise, get the result from the prior rank, combine it with my
* data, and send it to the next rank */
/* Get a temporary buffer to perform the reduction into. Rationale
* for malloc'ing this size is provided in coll_basic_reduce.c. */
- 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);
free_buffer = (char*)malloc(true_extent + (count - 1) * extent);
if (NULL == free_buffer) {
return OMPI_ERR_OUT_OF_RESOURCE;
}
reduce_buffer = free_buffer - lb;
err = ompi_datatype_copy_content_same_ddt(dtype, count,
reduce_buffer, (char*)sbuf);
/* Receive the reduced value from the prior rank */
err = MCA_PML_CALL(recv(rbuf, count, dtype, rank - 1,
MCA_COLL_BASE_TAG_EXSCAN, comm, MPI_STATUS_IGNORE));
if (MPI_SUCCESS != err) {
goto error;
}
/* Now reduce the prior rank's result with my source buffer. The source
* buffer had been previously copied into the temporary reduce_buffer. */
ompi_op_reduce(op, rbuf, reduce_buffer, count, dtype);
/* Send my result off to the next rank */
err = MCA_PML_CALL(send(reduce_buffer, count, dtype, rank + 1,
MCA_COLL_BASE_TAG_EXSCAN,
MCA_PML_BASE_SEND_STANDARD, comm));
/* Error */
error:
free(free_buffer);
/* All done */
return err;
}
/*
* exscan_inter
*
* Function: - basic exscan operation
* Accepts: - same arguments as MPI_Exscan()
* Returns: - MPI_SUCCESS or error code
*/
int
mca_coll_basic_exscan_inter(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)
{
return OMPI_ERR_NOT_IMPLEMENTED;
}