1 | #!/usr/bin/env python |
---|
2 | ################################################################# |
---|
3 | # Master/Slave Parallel decomposition sample |
---|
4 | # |
---|
5 | # Run as |
---|
6 | # python demo3.py |
---|
7 | # or |
---|
8 | # mpirun -np 2 demo3.py |
---|
9 | # (perhaps try number of processors more than 2) |
---|
10 | ################################################################# |
---|
11 | # |
---|
12 | # To verify bandwidth of your architexture please |
---|
13 | # run pytiming (and ctiming) |
---|
14 | # |
---|
15 | # OMN, GPC FEB 2002 |
---|
16 | # |
---|
17 | # |
---|
18 | |
---|
19 | import sys |
---|
20 | |
---|
21 | try: |
---|
22 | import Numeric |
---|
23 | except: |
---|
24 | raise 'Module Numeric must be present to run pypar' |
---|
25 | |
---|
26 | try: |
---|
27 | import pypar |
---|
28 | except: |
---|
29 | raise 'Module pypar must be present to run parallel' |
---|
30 | |
---|
31 | sys.stderr.write("Modules Numeric, pypar imported OK\n") |
---|
32 | |
---|
33 | WORKTAG = 1 |
---|
34 | DIETAG = 2 |
---|
35 | |
---|
36 | |
---|
37 | def master(): |
---|
38 | numCompleted = 0 |
---|
39 | |
---|
40 | sys.stderr.write("[MASTER]: I am processor %d of %d on node %s\n" %(MPI_myid, MPI_numproc, MPI_node)) |
---|
41 | |
---|
42 | # start slaves distributing the first work slot |
---|
43 | for i in range(1, min(MPI_numproc, numWorks)): |
---|
44 | work = workList[i] |
---|
45 | pypar.raw_send(work, i, WORKTAG) |
---|
46 | sys.stderr.write("[MASTER]: sent work '%s' to node '%d'\n" %(work, i)) |
---|
47 | |
---|
48 | # dispach the remaining work slots on dynamic load-balancing policy |
---|
49 | # the quicker to do the job, the more jobs it takes |
---|
50 | for work in workList[MPI_numproc:]: |
---|
51 | result = ' ' |
---|
52 | err, status = pypar.raw_receive(result, pypar.any_source, pypar.any_tag, return_status=True) |
---|
53 | #sys.stderr.write( "[MASTER]: received result '%s' from node '%d'\n" %(result, err[1][0])) |
---|
54 | sys.stderr.write("[MASTER]: received result '%s' from node '%d'\n" %(result, status.source)) |
---|
55 | numCompleted += 1 |
---|
56 | pypar.raw_send(work, status.source, WORKTAG) |
---|
57 | sys.stderr.write("[MASTER]: sent work '%s' to node '%d'\n" %(work, status.source)) |
---|
58 | |
---|
59 | # all works have been dispatched out |
---|
60 | sys.stderr.write("[MASTER]: toDo : %d\n" %numWorks) |
---|
61 | sys.stderr.write("[MASTER]: done : %d\n" %numCompleted) |
---|
62 | |
---|
63 | # I've still to take into the remaining completions |
---|
64 | while(numCompleted < numWorks): |
---|
65 | result = ' ' |
---|
66 | err, status = pypar.raw_receive(result, pypar.any_source, pypar.any_tag, return_status=True) |
---|
67 | sys.stderr.write("[MASTER]: received (final) result '%s' from node '%d'\n" %(result, status.source)) |
---|
68 | numCompleted += 1 |
---|
69 | sys.stderr.write("[MASTER]: %d completed\n" %numCompleted) |
---|
70 | |
---|
71 | sys.stderr.write( "[MASTER]: about to terminate slaves\n") |
---|
72 | |
---|
73 | # say slaves to stop working |
---|
74 | for i in range(1, MPI_numproc): |
---|
75 | pypar.raw_send('#', i, DIETAG) |
---|
76 | sys.stderr.write("[MASTER]: sent (final) work '%s' to node '%d'\n" %(0, i)) |
---|
77 | |
---|
78 | return |
---|
79 | |
---|
80 | def slave(): |
---|
81 | |
---|
82 | sys.stderr.write( "[SLAVE %d]: I am processor %d of %d on node %s\n" %(MPI_myid, MPI_myid, MPI_numproc, MPI_node)) |
---|
83 | |
---|
84 | while 1: |
---|
85 | result = ' ' |
---|
86 | err, status = pypar.raw_receive(result, pypar.any_source, pypar.any_tag, return_status=True) |
---|
87 | sys.stderr.write("[SLAVE %d]: received work '%s' with tag '%d' from node '%d'\n"\ |
---|
88 | %(MPI_myid, result, status.tag, status.source)) |
---|
89 | |
---|
90 | if (status.tag == DIETAG): |
---|
91 | sys.stderr.write("[SLAVE %d]: received termination from node '%d'\n" %(MPI_myid, 0)) |
---|
92 | return |
---|
93 | else: |
---|
94 | result = 'X'+result |
---|
95 | pypar.raw_send(result, 0) |
---|
96 | sys.stderr.write("[SLAVE %d]: sent result '%s' to node '%d'\n" %(MPI_myid, result, 0)) |
---|
97 | |
---|
98 | |
---|
99 | |
---|
100 | if __name__ == '__main__': |
---|
101 | MPI_myid = pypar.rank() |
---|
102 | MPI_numproc = pypar.size() |
---|
103 | MPI_node = pypar.Get_processor_name() |
---|
104 | |
---|
105 | _workList = ('_dummy_', 'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j') |
---|
106 | workList = ('_dummy_', 'a', 'b', 'c') |
---|
107 | numWorks = len(workList) - 1 |
---|
108 | |
---|
109 | |
---|
110 | #FIXME, better control here |
---|
111 | if MPI_numproc > numWorks or MPI_numproc < 2: |
---|
112 | pypar.Finalize() |
---|
113 | if MPI_myid == 0: |
---|
114 | sys.stderr.write("ERROR: Number of processors must be in the interval [2,%d].\n" %numWorks) |
---|
115 | |
---|
116 | sys.exit(-1) |
---|
117 | |
---|
118 | if MPI_myid == 0: |
---|
119 | master() |
---|
120 | else: |
---|
121 | slave() |
---|
122 | |
---|
123 | pypar.Finalize() |
---|
124 | sys.stderr.write("MPI environment finalized.\n") |
---|
125 | |
---|