1 /* Copyright (c) 2013-2023. 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. */
7 #include "../colls_private.hpp"
9 /*****************************************************************************
11 Copyright (c) 2006, Ahmad Faraj & Xin Yuan,
14 Redistribution and use in source and binary forms, with or without
15 modification, are permitted provided that the following conditions are met:
17 * Redistributions of source code must retain the above copyright notice,
18 this list of conditions and the following disclaimer.
20 * Redistributions in binary form must reproduce the above copyright notice,
21 this list of conditions and the following disclaimer in the documentation
22 and/or other materials provided with the distribution.
24 * Neither the name of the Florida State University nor the names of its
25 contributors may be used to endorse or promote products derived from this
26 software without specific prior written permission.
28 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
29 ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
30 WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
31 DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR
32 ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
33 (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
34 LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON
35 ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
36 (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
37 SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39 *************************************************************************
40 * Any results obtained from executing this software require the *
41 * acknowledgment and citation of the software and its owners. *
42 * The full citation is given below: *
44 * A. Faraj and X. Yuan. "Automatic Generation and Tuning of MPI *
45 * Collective Communication Routines." The 19th ACM International *
46 * Conference on Supercomputing (ICS), Cambridge, Massachusetts, *
48 *************************************************************************
50 *****************************************************************************/
52 /*****************************************************************************
55 * num: the number of processors in a communicator
59 * descp: takes a number and tries to find a factoring of x*y*z mesh out of it
60 ****************************************************************************/
63 static bool is_3dmesh(int num, int* i, int* j, int* k)
69 if ((num % (x * x)) == 0) {
79 /*****************************************************************************
80 * Function: allgather_3dmesh_shoot
82 * send_buff: send input buffer
83 * send_count: number of elements to send
84 * send_type: data type of elements being sent
85 * recv_buff: receive output buffer
86 * recv_count: number of elements to received
87 * recv_type: data type of elements being received
89 * Descrp: Function realizes the allgather operation using the 2dmesh
90 * algorithm. Allgather ommunication occurs first in the x dimension, y
91 * dimension, and then in the z dimension. Communication in each dimension
94 ****************************************************************************/
95 namespace simgrid::smpi {
97 int allgather__3dmesh(const void *send_buff, int send_count,
98 MPI_Datatype send_type, void *recv_buff,
99 int recv_count, MPI_Datatype recv_type,
104 int i, src, dst, rank, num_procs, block_size, my_z_base;
105 int my_z, X, Y, Z, send_offset, recv_offset;
106 int two_dsize, my_row_base, my_col_base, src_row_base, src_z_base, num_reqs;
107 int tag = COLL_TAG_ALLGATHER;
110 num_procs = comm->size();
111 extent = send_type->get_extent();
113 if (not is_3dmesh(num_procs, &X, &Y, &Z))
114 throw std::invalid_argument("allgather_3dmesh algorithm can't be used with this number of processes!");
124 my_z = rank / two_dsize;
126 my_row_base = (rank / X) * X;
127 my_col_base = (rank % Y) + (my_z * two_dsize);
128 my_z_base = my_z * two_dsize;
130 block_size = extent * send_count;
132 auto* req = new MPI_Request[num_reqs];
133 MPI_Request* req_ptr = req;
135 // do local allgather/local copy
136 recv_offset = rank * block_size;
137 Datatype::copy(send_buff, send_count, send_type, (char *)recv_buff + recv_offset,
138 recv_count, recv_type);
141 for (i = 0; i < Y; i++) {
142 src = i + my_row_base;
145 recv_offset = src * block_size;
146 *(req_ptr++) = Request::irecv((char *)recv_buff + recv_offset, send_count, recv_type, src, tag,
150 for (i = 0; i < Y; i++) {
151 dst = i + my_row_base;
154 Request::send(send_buff, send_count, send_type, dst, tag, comm);
157 Request::waitall(Y - 1, req, MPI_STATUSES_IGNORE);
160 // do colwise comm, it does not matter here if i*X or i *Y since X == Y
162 for (i = 0; i < X; i++) {
163 src = (i * Y + my_col_base);
167 src_row_base = (src / X) * X;
168 recv_offset = src_row_base * block_size;
169 *(req_ptr++) = Request::irecv((char *)recv_buff + recv_offset, recv_count * Y, recv_type, src, tag,
173 send_offset = my_row_base * block_size;
175 for (i = 0; i < X; i++) {
176 dst = (i * Y + my_col_base);
179 Request::send((char *)recv_buff + send_offset, send_count * Y, send_type, dst, tag,
183 Request::waitall(X - 1, req, MPI_STATUSES_IGNORE);
186 for (i = 1; i < Z; i++) {
187 src = (rank + i * two_dsize) % num_procs;
188 src_z_base = (src / two_dsize) * two_dsize;
190 recv_offset = (src_z_base * block_size);
192 *(req_ptr++) = Request::irecv((char *)recv_buff + recv_offset, recv_count * two_dsize, recv_type,
196 for (i = 1; i < Z; i++) {
197 dst = (rank + i * two_dsize) % num_procs;
198 send_offset = my_z_base * block_size;
199 Request::send((char *)recv_buff + send_offset, send_count * two_dsize, send_type,
202 Request::waitall(Z - 1, req, MPI_STATUSES_IGNORE);
209 } // namespace simgrid::smpi