1
1
openmpi/ompi/mca/coll/base/coll_base_util.c
Gilles Gouaillardet 63d3ccde9d coll/base: only retain datatypes/op if the request has not yet completed
a non blocking collective might return ompi_request_null, so we should not
retain anything in that case.

Signed-off-by: Gilles Gouaillardet <gilles@rist.or.jp>
2019-08-09 09:57:56 +09:00

308 строки
11 KiB
C

/*
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
* University Research and Technology
* Corporation. All rights reserved.
* Copyright (c) 2004-2016 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) 2014-2019 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
*
* $HEADER$
*/
#include "ompi_config.h"
#include "mpi.h"
#include "ompi/constants.h"
#include "ompi/datatype/ompi_datatype.h"
#include "ompi/communicator/communicator.h"
#include "ompi/mca/coll/base/coll_tags.h"
#include "ompi/mca/coll/base/coll_base_functions.h"
#include "ompi/mca/topo/base/base.h"
#include "ompi/mca/pml/pml.h"
#include "coll_base_util.h"
int ompi_coll_base_sendrecv_actual( const void* sendbuf, size_t scount,
ompi_datatype_t* sdatatype,
int dest, int stag,
void* recvbuf, size_t rcount,
ompi_datatype_t* rdatatype,
int source, int rtag,
struct ompi_communicator_t* comm,
ompi_status_public_t* status )
{ /* post receive first, then send, then wait... should be fast (I hope) */
int err, line = 0;
size_t rtypesize, stypesize;
ompi_request_t *req = MPI_REQUEST_NULL;
ompi_status_public_t rstatus;
/* post new irecv */
ompi_datatype_type_size(rdatatype, &rtypesize);
err = MCA_PML_CALL(irecv( recvbuf, rcount, rdatatype, source, rtag,
comm, &req));
if (err != MPI_SUCCESS) { line = __LINE__; goto error_handler; }
/* send data to children */
ompi_datatype_type_size(sdatatype, &stypesize);
err = MCA_PML_CALL(send( sendbuf, scount, sdatatype, dest, stag,
MCA_PML_BASE_SEND_STANDARD, comm));
if (err != MPI_SUCCESS) { line = __LINE__; goto error_handler; }
err = ompi_request_wait( &req, &rstatus);
if (err != MPI_SUCCESS) { line = __LINE__; goto error_handler; }
if (MPI_STATUS_IGNORE != status) {
*status = rstatus;
}
return (MPI_SUCCESS);
error_handler:
/* Error discovered during the posting of the irecv or send,
* and no status is available.
*/
OPAL_OUTPUT ((ompi_coll_base_framework.framework_output, "%s:%d: Error %d occurred\n",
__FILE__, line, err));
(void)line; // silence compiler warning
if (MPI_STATUS_IGNORE != status) {
status->MPI_ERROR = err;
}
return (err);
}
/*
* ompi_mirror_perm: Returns mirror permutation of nbits low-order bits
* of x [*].
* [*] Warren Jr., Henry S. Hacker's Delight (2ed). 2013.
* Chapter 7. Rearranging Bits and Bytes.
*/
unsigned int ompi_mirror_perm(unsigned int x, int nbits)
{
x = (((x & 0xaaaaaaaa) >> 1) | ((x & 0x55555555) << 1));
x = (((x & 0xcccccccc) >> 2) | ((x & 0x33333333) << 2));
x = (((x & 0xf0f0f0f0) >> 4) | ((x & 0x0f0f0f0f) << 4));
x = (((x & 0xff00ff00) >> 8) | ((x & 0x00ff00ff) << 8));
x = ((x >> 16) | (x << 16));
return x >> (sizeof(x) * CHAR_BIT - nbits);
}
/*
* ompi_rounddown: Rounds a number down to nearest multiple.
* rounddown(10,4) = 8, rounddown(6,3) = 6, rounddown(14,3) = 12
*/
int ompi_rounddown(int num, int factor)
{
num /= factor;
return num * factor; /* floor(num / factor) * factor */
}
static void release_objs_callback(struct ompi_coll_base_nbc_request_t *request) {
if (NULL != request->data.objs.objs[0]) {
OBJ_RELEASE(request->data.objs.objs[0]);
request->data.objs.objs[0] = NULL;
}
if (NULL != request->data.objs.objs[1]) {
OBJ_RELEASE(request->data.objs.objs[1]);
request->data.objs.objs[1] = NULL;
}
}
static int complete_objs_callback(struct ompi_request_t *req) {
struct ompi_coll_base_nbc_request_t *request = (ompi_coll_base_nbc_request_t *)req;
int rc = OMPI_SUCCESS;
assert (NULL != request);
if (NULL != request->cb.req_complete_cb) {
rc = request->cb.req_complete_cb(request->req_complete_cb_data);
}
release_objs_callback(request);
return rc;
}
static int free_objs_callback(struct ompi_request_t **rptr) {
struct ompi_coll_base_nbc_request_t *request = *(ompi_coll_base_nbc_request_t **)rptr;
int rc = OMPI_SUCCESS;
if (NULL != request->cb.req_free) {
rc = request->cb.req_free(rptr);
}
release_objs_callback(request);
return rc;
}
int ompi_coll_base_retain_op( ompi_request_t *req, ompi_op_t *op,
ompi_datatype_t *type) {
ompi_coll_base_nbc_request_t *request = (ompi_coll_base_nbc_request_t *)req;
bool retain = false;
if (REQUEST_COMPLETE(req)) {
return OMPI_SUCCESS;
}
if (!ompi_op_is_intrinsic(op)) {
OBJ_RETAIN(op);
request->data.op.op = op;
retain = true;
}
if (!ompi_datatype_is_predefined(type)) {
OBJ_RETAIN(type);
request->data.op.datatype = type;
retain = true;
}
if (OPAL_UNLIKELY(retain)) {
/* We need to consider two cases :
* - non blocking collectives:
* the objects can be released when MPI_Wait() completes
* and we use the req_complete_cb callback
* - persistent non blocking collectives:
* the objects can only be released when the request is freed
* (e.g. MPI_Request_free() completes) and we use req_free callback
*/
if (req->req_persistent) {
request->cb.req_free = req->req_free;
req->req_free = free_objs_callback;
} else {
request->cb.req_complete_cb = req->req_complete_cb;
request->req_complete_cb_data = req->req_complete_cb_data;
req->req_complete_cb = complete_objs_callback;
req->req_complete_cb_data = request;
}
}
return OMPI_SUCCESS;
}
int ompi_coll_base_retain_datatypes( ompi_request_t *req, ompi_datatype_t *stype,
ompi_datatype_t *rtype) {
ompi_coll_base_nbc_request_t *request = (ompi_coll_base_nbc_request_t *)req;
bool retain = false;
if (REQUEST_COMPLETE(req)) {
return OMPI_SUCCESS;
}
if (NULL != stype && !ompi_datatype_is_predefined(stype)) {
OBJ_RETAIN(stype);
request->data.types.stype = stype;
retain = true;
}
if (NULL != rtype && !ompi_datatype_is_predefined(rtype)) {
OBJ_RETAIN(rtype);
request->data.types.rtype = rtype;
retain = true;
}
if (OPAL_UNLIKELY(retain)) {
if (req->req_persistent) {
request->cb.req_free = req->req_free;
req->req_free = free_objs_callback;
} else {
request->cb.req_complete_cb = req->req_complete_cb;
request->req_complete_cb_data = req->req_complete_cb_data;
req->req_complete_cb = complete_objs_callback;
req->req_complete_cb_data = request;
}
}
return OMPI_SUCCESS;
}
static void release_vecs_callback(ompi_coll_base_nbc_request_t *request) {
ompi_communicator_t *comm = request->super.req_mpi_object.comm;
int scount, rcount;
if (OMPI_COMM_IS_TOPO(comm)) {
(void)mca_topo_base_neighbor_count (comm, &rcount, &scount);
} else {
scount = rcount = OMPI_COMM_IS_INTER(comm)?ompi_comm_remote_size(comm):ompi_comm_size(comm);
}
if (NULL != request->data.vecs.stypes) {
for (int i=0; i<scount; i++) {
if (NULL != request->data.vecs.stypes[i]) {
OMPI_DATATYPE_RELEASE(request->data.vecs.stypes[i]);
}
}
request->data.vecs.stypes = NULL;
}
if (NULL != request->data.vecs.rtypes) {
for (int i=0; i<rcount; i++) {
if (NULL != request->data.vecs.rtypes[i]) {
OMPI_DATATYPE_RELEASE(request->data.vecs.rtypes[i]);
}
}
request->data.vecs.rtypes = NULL;
}
}
static int complete_vecs_callback(struct ompi_request_t *req) {
ompi_coll_base_nbc_request_t *request = (ompi_coll_base_nbc_request_t *)req;
int rc = OMPI_SUCCESS;
assert (NULL != request);
if (NULL != request->cb.req_complete_cb) {
rc = request->cb.req_complete_cb(request->req_complete_cb_data);
}
release_vecs_callback(request);
return rc;
}
static int free_vecs_callback(struct ompi_request_t **rptr) {
struct ompi_coll_base_nbc_request_t *request = *(ompi_coll_base_nbc_request_t **)rptr;
int rc = OMPI_SUCCESS;
if (NULL != request->cb.req_free) {
rc = request->cb.req_free(rptr);
}
release_vecs_callback(request);
return rc;
}
int ompi_coll_base_retain_datatypes_w( ompi_request_t *req,
ompi_datatype_t *stypes[], ompi_datatype_t *rtypes[]) {
ompi_coll_base_nbc_request_t *request = (ompi_coll_base_nbc_request_t *)req;
bool retain = false;
ompi_communicator_t *comm = request->super.req_mpi_object.comm;
int scount, rcount;
if (REQUEST_COMPLETE(req)) {
return OMPI_SUCCESS;
}
if (OMPI_COMM_IS_TOPO(comm)) {
(void)mca_topo_base_neighbor_count (comm, &rcount, &scount);
} else {
scount = rcount = OMPI_COMM_IS_INTER(comm)?ompi_comm_remote_size(comm):ompi_comm_size(comm);
}
for (int i=0; i<scount; i++) {
if (NULL != stypes && NULL != stypes[i] && !ompi_datatype_is_predefined(stypes[i])) {
OBJ_RETAIN(stypes[i]);
retain = true;
}
}
for (int i=0; i<rcount; i++) {
if (NULL != rtypes && NULL != rtypes[i] && !ompi_datatype_is_predefined(rtypes[i])) {
OBJ_RETAIN(rtypes[i]);
retain = true;
}
}
if (OPAL_UNLIKELY(retain)) {
request->data.vecs.stypes = stypes;
request->data.vecs.rtypes = rtypes;
if (req->req_persistent) {
request->cb.req_free = req->req_free;
req->req_free = free_vecs_callback;
} else {
request->cb.req_complete_cb = req->req_complete_cb;
request->req_complete_cb_data = req->req_complete_cb_data;
req->req_complete_cb = complete_vecs_callback;
req->req_complete_cb_data = request;
}
}
return OMPI_SUCCESS;
}
static void nbc_req_cons(ompi_coll_base_nbc_request_t *req) {
req->cb.req_complete_cb = NULL;
req->req_complete_cb_data = NULL;
req->data.objs.objs[0] = NULL;
req->data.objs.objs[1] = NULL;
}
OBJ_CLASS_INSTANCE(ompi_coll_base_nbc_request_t, ompi_request_t, nbc_req_cons, NULL);