1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
|
Copyright (c) 2001-2003 The Trustees of Indiana University.
All rights reserved.
Copyright (c) 1998-2001 University of Notre Dame.
All rights reserved.
Copyright (c) 1994-1998 The Ohio State University.
All rights reserved.
This file is part of the LAM/MPI software package. For license
information, see the LICENSE file in the top level directory of the
LAM/MPI source distribution.
$HEADER$
The ring program is a canonical example of MPI usage. It sends a
simple message around in a ring pattern:
Rank 0 sends to rank 1
Rank 1 sends to rank 2
Rank 2 sends to rank 3
...
Rank N-1 sends to rank N
Rank N sends to rank 0
Rank 0 sends to rank 1
etc.
The message that is sent around the ring is a single integer. If the
integer is zero, each rank quits after it has passed the message on.
Rank zero is special -- it will receive the message and decrement the
integer before it passes the message on. This scheme guarantees that
the application completes in an orderly fashion in finite time.
Rank zero is also special in that it both starts and finishes the
ring. Rank 0 sends the initial message (which is hardwired to "5" in
this example) to rank 1 to start the whole process. Rank 0 then also
has to consume the final "0" message from rank N so as not to leave a
message "hanging" out on the network when all the ranks quit.
Use "make" to compile this example. Make will use mpicc to compile
the program:
mpicc -o ring ring.c
This program can be run with any number of MPI processes (including
1). For example:
mpirun C ring
|
