int ompi_coll_tuned_alltoall_intra_basic_linear(void *sbuf, int scount,
struct ompi_datatype_t *sdtype,
void* rbuf, int rcount,
struct ompi_datatype_t *rdtype,
struct ompi_communicator_t *comm)
{
int i;
int rank;
int size;
int err;
int nreqs;
char *psnd;
char *prcv;
MPI_Aint lb;
MPI_Aint sndinc;
MPI_Aint rcvinc;
ompi_request_t **req;
ompi_request_t **sreq;
ompi_request_t **rreq;
/* Initialize. */
size = ompi_comm_size(comm);
rank = ompi_comm_rank(comm);
OPAL_OUTPUT((ompi_coll_tuned_stream,"ompi_coll_tuned_alltoall_intra_basic_linear rank %d", rank));
err = ompi_ddt_get_extent(sdtype, &lb, &sndinc);
if (OMPI_SUCCESS != err) {
return err;
}
sndinc *= scount;
err = ompi_ddt_get_extent(rdtype, &lb, &rcvinc);
if (OMPI_SUCCESS != err) {
return err;
}
rcvinc *= rcount;
/* simple optimization */
psnd = ((char *) sbuf) + (rank * sndinc);
prcv = ((char *) rbuf) + (rank * rcvinc);
err = ompi_ddt_sndrcv(psnd, scount, sdtype, prcv, rcount, rdtype);
if (MPI_SUCCESS != err) {
return err;
}
/* If only one process, we're done. */
if (1 == size) {
return MPI_SUCCESS;
}
/* Initiate all send/recv to/from others. */
req = rreq = comm->c_coll_basic_data->mcct_reqs;
sreq = rreq + size - 1;
prcv = (char *) rbuf;
psnd = (char *) sbuf;
/* Post all receives first -- a simple optimization */
for (nreqs = 0, i = (rank + 1) % size; i != rank;
i = (i + 1) % size, ++rreq, ++nreqs) {
err =
MCA_PML_CALL(irecv_init
(prcv + (i * rcvinc), rcount, rdtype, i,
MCA_COLL_BASE_TAG_ALLTOALL, comm, rreq));
if (MPI_SUCCESS != err) {
ompi_coll_tuned_free_reqs(req, rreq - req);
return err;
}
}
/* Now post all sends in reverse order
- We would like to minimize the search time through message queue
when messages actually arrive in the order in which they were posted.
*/
for (nreqs = 0, i = (rank + size - 1) % size; i != rank;
i = (i + size - 1) % size, ++sreq, ++nreqs) {
err =
MCA_PML_CALL(isend_init
(psnd + (i * sndinc), scount, sdtype, i,
MCA_COLL_BASE_TAG_ALLTOALL,
MCA_PML_BASE_SEND_STANDARD, comm, sreq));
if (MPI_SUCCESS != err) {
ompi_coll_tuned_free_reqs(req, sreq - req);
return err;
}
}
nreqs = (size - 1) * 2;
/* Start your engines. This will never return an error. */
MCA_PML_CALL(start(nreqs, req));
/* Wait for them all. If there's an error, note that we don't
* care what the error was -- just that there *was* an error. The
* PML will finish all requests, even if one or more of them fail.
* i.e., by the end of this call, all the requests are free-able.
* So free them anyway -- even if there was an error, and return
* the error after we free everything. */
err = ompi_request_wait_all(nreqs, req, MPI_STATUSES_IGNORE);
/* Free the reqs */
ompi_coll_tuned_free_reqs(req, nreqs);
/* All done */
return err;
}
/*
* Linear functions are copied from the basic coll module. For
* some small number of nodes and/or small data sizes they are just as
* fast as tuned/tree based segmenting operations and as such may be
* selected by the decision functions. These are copied into this module
* due to the way we select modules in V1. i.e. in V2 we will handle this
* differently and so will not have to duplicate code.
* GEF Oct05 after asking Jeff.
*/
int
ompi_coll_tuned_alltoallv_intra_basic_linear(void *sbuf, int *scounts, int *sdisps,
struct ompi_datatype_t *sdtype,
void *rbuf, int *rcounts, int *rdisps,
struct ompi_datatype_t *rdtype,
struct ompi_communicator_t *comm,
mca_coll_base_module_t *module)
{
int i, size, rank, err;
char *psnd, *prcv;
int nreqs;
ptrdiff_t sext, rext;
MPI_Request *preq;
mca_coll_tuned_module_t *tuned_module = (mca_coll_tuned_module_t*) module;
mca_coll_tuned_comm_t *data = tuned_module->tuned_data;
size = ompi_comm_size(comm);
rank = ompi_comm_rank(comm);
OPAL_OUTPUT((ompi_coll_tuned_stream,
"coll:tuned:alltoallv_intra_basic_linear rank %d", rank));
ompi_ddt_type_extent(sdtype, &sext);
ompi_ddt_type_extent(rdtype, &rext);
/* Simple optimization - handle send to self first */
psnd = ((char *) sbuf) + (sdisps[rank] * sext);
prcv = ((char *) rbuf) + (rdisps[rank] * rext);
if (0 != scounts[rank]) {
err = ompi_ddt_sndrcv(psnd, scounts[rank], sdtype,
prcv, rcounts[rank], rdtype);
if (MPI_SUCCESS != err) {
return err;
}
}
/* If only one process, we're done. */
if (1 == size) {
return MPI_SUCCESS;
}
/* Now, initiate all send/recv to/from others. */
nreqs = 0;
preq = data->mcct_reqs;
/* Post all receives first */
for (i = 0; i < size; ++i) {
if (i == rank || 0 == rcounts[i]) {
continue;
}
prcv = ((char *) rbuf) + (rdisps[i] * rext);
err = MCA_PML_CALL(irecv_init(prcv, rcounts[i], rdtype,
i, MCA_COLL_BASE_TAG_ALLTOALLV, comm,
preq++));
++nreqs;
if (MPI_SUCCESS != err) {
ompi_coll_tuned_free_reqs(data->mcct_reqs, nreqs);
return err;
}
}
/* Now post all sends */
for (i = 0; i < size; ++i) {
if (i == rank || 0 == scounts[i]) {
continue;
}
psnd = ((char *) sbuf) + (sdisps[i] * sext);
err = MCA_PML_CALL(isend_init(psnd, scounts[i], sdtype,
i, MCA_COLL_BASE_TAG_ALLTOALLV,
MCA_PML_BASE_SEND_STANDARD, comm,
preq++));
++nreqs;
if (MPI_SUCCESS != err) {
ompi_coll_tuned_free_reqs(data->mcct_reqs, nreqs);
return err;
}
}
/* Start your engines. This will never return an error. */
MCA_PML_CALL(start(nreqs, data->mcct_reqs));
/* Wait for them all. If there's an error, note that we don't care
* what the error was -- just that there *was* an error. The PML
* will finish all requests, even if one or more of them fail.
* i.e., by the end of this call, all the requests are free-able.
* So free them anyway -- even if there was an error, and return the
* error after we free everything. */
err = ompi_request_wait_all(nreqs, data->mcct_reqs,
MPI_STATUSES_IGNORE);
/* Free the requests. */
ompi_coll_tuned_free_reqs(data->mcct_reqs, nreqs);
return err;
}
/*
* bcast_lin_intra
*
* Function: - broadcast using O(N) algorithm
* Accepts: - same arguments as MPI_Bcast()
* Returns: - MPI_SUCCESS or error code
*/
int
ompi_coll_tuned_bcast_intra_basic_linear (void *buff, int count,
struct ompi_datatype_t *datatype, int root,
struct ompi_communicator_t *comm,
mca_coll_base_module_t *module)
{
int i, size, rank, err;
mca_coll_tuned_module_t *tuned_module = (mca_coll_tuned_module_t*) module;
mca_coll_tuned_comm_t *data = tuned_module->tuned_data;
ompi_request_t **preq, **reqs = data->mcct_reqs;
size = ompi_comm_size(comm);
rank = ompi_comm_rank(comm);
OPAL_OUTPUT((ompi_coll_tuned_stream,"ompi_coll_tuned_bcast_intra_basic_linear rank %d root %d", rank, root));
/* Non-root receive the data. */
if (rank != root) {
return MCA_PML_CALL(recv(buff, count, datatype, root,
MCA_COLL_BASE_TAG_BCAST, comm,
MPI_STATUS_IGNORE));
}
/* Root sends data to all others. */
for (i = 0, preq = reqs; i < size; ++i) {
if (i == rank) {
continue;
}
err = MCA_PML_CALL(isend_init(buff, count, datatype, i,
MCA_COLL_BASE_TAG_BCAST,
MCA_PML_BASE_SEND_STANDARD,
comm, preq++));
if (MPI_SUCCESS != err) {
return err;
}
}
--i;
/* Start your engines. This will never return an error. */
MCA_PML_CALL(start(i, reqs));
/* Wait for them all. If there's an error, note that we don't
* care what the error was -- just that there *was* an error. The
* PML will finish all requests, even if one or more of them fail.
* i.e., by the end of this call, all the requests are free-able.
* So free them anyway -- even if there was an error, and return
* the error after we free everything. */
err = ompi_request_wait_all(i, reqs, MPI_STATUSES_IGNORE);
/* Free the reqs */
ompi_coll_tuned_free_reqs(reqs, i);
/* All done */
return err;
}
