1 /* Copyright (c) 2013-2017. The SimGrid Team.
2 * All rights reserved. */
4 /* This program is free software; you can redistribute it and/or modify it
5 * under the terms of the license (GNU LGPL) which comes with this package. */
8 * Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
9 * University Research and Technology
10 * Corporation. All rights reserved.
11 * Copyright (c) 2004-2006 The University of Tennessee and The University
12 * of Tennessee Research Foundation. All rights
14 * Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
15 * University of Stuttgart. All rights reserved.
16 * Copyright (c) 2004-2005 The Regents of the University of California.
17 * All rights reserved.
18 * Copyright (c) 2008 Sun Microsystems, Inc. All rights reserved.
20 * Additional copyrights may follow
23 #include "../colls_private.h"
24 #include "../coll_tuned_topo.h"
28 * Barrier is ment to be a synchronous operation, as some BTLs can mark
29 * a request done before its passed to the NIC and progress might not be made
30 * elsewhere we cannot allow a process to exit the barrier until its last
31 * [round of] sends are completed.
33 * It is last round of sends rather than 'last' individual send as each pair of
34 * peers can use different channels/devices/btls and the receiver of one of
35 * these sends might be forced to wait as the sender
36 * leaves the collective and does not make progress until the next mpi call
41 * Simple double ring version of barrier
43 * synchronous gurantee made by last ring of sends are synchronous
48 int Coll_barrier_ompi_doublering::barrier(MPI_Comm comm)
57 XBT_DEBUG("ompi_coll_tuned_barrier_ompi_doublering rank %d", rank);
59 left = ((rank-1+size)%size);
60 right = ((rank+1)%size);
62 if (rank > 0) { /* receive message from the left */
63 Request::recv((void*)NULL, 0, MPI_BYTE, left,
64 COLL_TAG_BARRIER, comm,
68 /* Send message to the right */
69 Request::send((void*)NULL, 0, MPI_BYTE, right,
73 /* root needs to receive from the last node */
75 Request::recv((void*)NULL, 0, MPI_BYTE, left,
76 COLL_TAG_BARRIER, comm,
80 /* Allow nodes to exit */
81 if (rank > 0) { /* post Receive from left */
82 Request::recv((void*)NULL, 0, MPI_BYTE, left,
83 COLL_TAG_BARRIER, comm,
87 /* send message to the right one */
88 Request::send((void*)NULL, 0, MPI_BYTE, right,
92 /* rank 0 post receive from the last node */
94 Request::recv((void*)NULL, 0, MPI_BYTE, left,
95 COLL_TAG_BARRIER, comm,
105 * To make synchronous, uses sync sends and sync sendrecvs
108 int Coll_barrier_ompi_recursivedoubling::barrier(MPI_Comm comm)
110 int rank, size, adjsize;
116 "ompi_coll_tuned_barrier_ompi_recursivedoubling rank %d",
119 /* do nearest power of 2 less than size calc */
120 for( adjsize = 1; adjsize <= size; adjsize <<= 1 );
123 /* if size is not exact power of two, perform an extra step */
124 if (adjsize != size) {
125 if (rank >= adjsize) {
126 /* send message to lower ranked node */
127 remote = rank - adjsize;
128 Request::sendrecv(NULL, 0, MPI_BYTE, remote,
130 NULL, 0, MPI_BYTE, remote,
132 comm, MPI_STATUS_IGNORE);
134 } else if (rank < (size - adjsize)) {
136 /* receive message from high level rank */
137 Request::recv((void*)NULL, 0, MPI_BYTE, rank+adjsize,
138 COLL_TAG_BARRIER, comm,
144 /* exchange messages */
145 if ( rank < adjsize ) {
147 while ( mask < adjsize ) {
148 remote = rank ^ mask;
150 if (remote >= adjsize) continue;
152 /* post receive from the remote node */
153 Request::sendrecv(NULL, 0, MPI_BYTE, remote,
155 NULL, 0, MPI_BYTE, remote,
157 comm, MPI_STATUS_IGNORE);
161 /* non-power of 2 case */
162 if (adjsize != size) {
163 if (rank < (size - adjsize)) {
164 /* send enter message to higher ranked node */
165 remote = rank + adjsize;
166 Request::send((void*)NULL, 0, MPI_BYTE, remote,
179 * To make synchronous, uses sync sends and sync sendrecvs
182 int Coll_barrier_ompi_bruck::barrier(MPI_Comm comm)
185 int distance, to, from;
190 "ompi_coll_tuned_barrier_ompi_bruck rank %d", rank);
192 /* exchange data with rank-2^k and rank+2^k */
193 for (distance = 1; distance < size; distance <<= 1) {
194 from = (rank + size - distance) % size;
195 to = (rank + distance) % size;
197 /* send message to lower ranked node */
198 Request::sendrecv(NULL, 0, MPI_BYTE, to,
200 NULL, 0, MPI_BYTE, from,
202 comm, MPI_STATUS_IGNORE);
211 * To make synchronous, uses sync sends and sync sendrecvs
213 /* special case for two processes */
214 int Coll_barrier_ompi_two_procs::barrier(MPI_Comm comm)
218 remote = comm->rank();
220 "ompi_coll_tuned_barrier_ompi_two_procs rank %d", remote);
221 remote = (remote + 1) & 0x1;
223 Request::sendrecv(NULL, 0, MPI_BYTE, remote,
225 NULL, 0, MPI_BYTE, remote,
227 comm, MPI_STATUS_IGNORE);
228 return (MPI_SUCCESS);
233 * Linear functions are copied from the BASIC coll module
234 * they do not segment the message and are simple implementations
235 * but for some small number of nodes and/or small data sizes they
236 * are just as fast as tuned/tree based segmenting operations
237 * and as such may be selected by the decision functions
238 * These are copied into this module due to the way we select modules
239 * in V1. i.e. in V2 we will handle this differently and so will not
240 * have to duplicate code.
241 * GEF Oct05 after asking Jeff.
244 /* copied function (with appropriate renaming) starts here */
246 int Coll_barrier_ompi_basic_linear::barrier(MPI_Comm comm)
249 int size = comm->size();
250 int rank = comm->rank();
252 /* All non-root send & receive zero-length message. */
255 Request::send (NULL, 0, MPI_BYTE, 0,
259 Request::recv (NULL, 0, MPI_BYTE, 0,
261 comm, MPI_STATUS_IGNORE);
264 /* The root collects and broadcasts the messages. */
267 MPI_Request* requests;
269 requests = (MPI_Request*)malloc( size * sizeof(MPI_Request) );
270 for (i = 1; i < size; ++i) {
271 requests[i] = Request::irecv(NULL, 0, MPI_BYTE, MPI_ANY_SOURCE,
272 COLL_TAG_BARRIER, comm
275 Request::waitall( size-1, requests+1, MPI_STATUSES_IGNORE );
277 for (i = 1; i < size; ++i) {
278 requests[i] = Request::isend(NULL, 0, MPI_BYTE, i,
283 Request::waitall( size-1, requests+1, MPI_STATUSES_IGNORE );
292 /* copied function (with appropriate renaming) ends here */
295 * Another recursive doubling type algorithm, but in this case
296 * we go up the tree and back down the tree.
298 int Coll_barrier_ompi_tree::barrier(MPI_Comm comm)
300 int rank, size, depth;
306 "ompi_coll_tuned_barrier_ompi_tree %d",
309 /* Find the nearest power of 2 of the communicator size. */
310 for(depth = 1; depth < size; depth <<= 1 );
312 for (jump=1; jump<depth; jump<<=1) {
313 partner = rank ^ jump;
314 if (!(partner & (jump-1)) && partner < size) {
315 if (partner > rank) {
316 Request::recv (NULL, 0, MPI_BYTE, partner,
317 COLL_TAG_BARRIER, comm,
319 } else if (partner < rank) {
320 Request::send (NULL, 0, MPI_BYTE, partner,
328 for (jump = depth; jump>0; jump>>=1) {
329 partner = rank ^ jump;
330 if (!(partner & (jump-1)) && partner < size) {
331 if (partner > rank) {
332 Request::send (NULL, 0, MPI_BYTE, partner,
335 } else if (partner < rank) {
336 Request::recv (NULL, 0, MPI_BYTE, partner,
337 COLL_TAG_BARRIER, comm,