forked from johannesgerer/jburkardt-cpp
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathwave_mpi.html
334 lines (288 loc) · 10 KB
/
wave_mpi.html
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
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
<html>
<head>
<title>
WAVE_MPI - Parallel Solution of the 1D Wave Equation using MPI
</title>
</head>
<body bgcolor="#EEEEEE" link="#CC0000" alink="#FF3300" vlink="#000055">
<h1 align = "center">
WAVE_MPI <br> Parallel Solution of the 1D Wave Equation using MPI
</h1>
<hr>
<p>
<b>WAVE_MPI</b>
is a C++ program which
solves the 1D wave equation in parallel, using MPI.
</p>
<p>
This program solves the 1D wave equation of the form:
<pre>
Utt = c^2 Uxx
</pre>
over the spatial interval [X1,X2] and time interval [T1,T2],
with initial conditions:
<pre>
U(X,T1) = U_T1(X),
Ut(X,T1) = UT_T1(X),
</pre>
and boundary conditions of Dirichlet type:
<pre>
U(X1,T) = U_X1(T),<br>
U(X2,T) = U_X2(T).
</pre>
The value <b>C</b> represents the propagation speed of waves.
</p>
<p>
The program uses the finite difference method, and marches
forward in time, solving for all the values of U at the next
time step by using the values known at the previous two time steps.
</p>
<p>
Central differences may be used to approximate both the time
and space derivatives in the original differential equation.
</p>
<p>
Thus, assuming we have available the approximated values of U
at the current and previous times, we may write a discretized
version of the wave equation as follows:
<pre>
Uxx(X,T) = ( U(X+dX,T ) - 2 U(X,T) + U(X-dX,T, ) ) / dX^2
Utt(X,T) = ( U(X, T+dT) - 2 U(X,T) + U(X, T-dt) ) / dT^2
</pre>
If we multiply the first term by C^2 and solve for the single
unknown value U(X,T+dt), we have:
<pre>
U(X,T+dT) = ( C^2 * dT^2 / dX^2 ) * U(X+dX,T)
+ 2 * ( 1 - C^2 * dT^2 / dX^2 ) * U(X, T)
+ ( C^2 * dT^2 / dX^2 ) * U(X-dX,T)
- U(X, T-dT)
</pre>
(Equation to advance from time T to time T+dT, except for FIRST step!)
</p>
<p>
However, on the very first step, we only have the values of U
for the initial time, but not for the previous time step.
In that case, we use the initial condition information for dUdT
which can be approximated by a central difference that involves
U(X,T+dT) and U(X,T-dT):
<pre>
dU/dT(X,T) = ( U(X,T+dT) - U(X,T-dT) ) / ( 2 * dT )
</pre>
and so we can estimate U(X,T-dT) as
<pre>
U(X,T-dT) = U(X,T+dT) - 2 * dT * dU/dT(X,T)
</pre>
If we replace the "missing" value of U(X,T-dT) by the known values
on the right hand side, we now have U(X,T+dT) on both sides of the
equation, so we have to rearrange to get the formula we use
for just the first time step:
<pre>
U(X,T+dT) = 1/2 * ( C^2 * dT^2 / dX^2 ) * U(X+dX,T)
+ ( 1 - C^2 * dT^2 / dX^2 ) * U(X, T)
+ 1/2 * ( C^2 * dT^2 / dX^2 ) * U(X-dX,T)
+ dT * dU/dT(X, T )
</pre>
(Equation to advance from time T to time T+dT for FIRST step.)
</p>
<p>
It should be clear now that the quantity ALPHA = C * dT / dX will affect
the stability of the calculation. If it is greater than 1, then
the middle coefficient (1-C^2 dT^2 / dX^2) is negative, and the
sum of the magnitudes of the three coefficients becomes unbounded.
</p>
<p>
We wish to use MPI in order to accelerate this computation. We use the
method of domain decomposition - that is, we assume we have P MPI processes,
we divide the original interval into P subintervals, and we expect
each process to update the data associated with its subinterval.
</p>
<p>
However, to compute the estimated solution U(X,T+dT) at the next time
step requires information about U(X-dX,T) and U(X+dX,T). When process
ID tries to make these estimates, it will need one value from process
ID-1, and one from process ID+1, before it can make all the updates.
MPI allows the processes to communicate this information using messages.
</p>
<p>
At the end of the complete calculation, we wish to print a table of the
solution at the final time. To do this in an organized fashion, we want
each process to send its final result to the master process (with ID = 0).
Once all the data has been collected, the master process prints it.
</p>
<h3 align = "center">
Licensing:
</h3>
<p>
The computer code and data files described and made available on this web page
are distributed under
<a href = "../../txt/gnu_lgpl.txt">the GNU LGPL license.</a>
</p>
<h3 align = "center">
Languages:
</h3>
<p>
<b>WAVE_MPI</b> is available in
<a href = "../../c_src/wave_mpi/wave_mpi.html">a C version</a> and
<a href = "../../cpp_src/wave_mpi/wave_mpi.html">a C++ version</a> and
<a href = "../../f77_src/wave_mpi/wave_mpi.html">a FORTRAN77 version</a> and
<a href = "../../f_src/wave_mpi/wave_mpi.html">a FORTRAN90 version</a>.
</p>
<h3 align = "center">
Related Data and Programs:
</h3>
<p>
<a href = "../../cpp_src/communicator_mpi/communicator_mpi.html">
COMMUNICATOR_MPI</a>,
a C++ program which
creates new communicators involving a subset of initial
set of MPI processes in the default communicator MPI_COMM_WORLD.
</p>
<p>
<a href = "../../cpp_src/heat_mpi/heat_mpi.html">
HEAT_MPI</a>,
a C++ program which
solves the 1D Time Dependent Heat Equation using MPI.
</p>
<p>
<a href = "../../cpp_src/hello_mpi/hello_mpi.html">
HELLO_MPI</a>,
a C++ program which
prints out "Hello, world!" using the MPI parallel programming environment.
</p>
<p>
<a href = "../../examples/moab/moab.html">
MOAB</a>,
examples which
illustrate the use of the MOAB job scheduler for a computer cluster.
</p>
<p>
<a href = "../../cpp_src/mpi/mpi.html">
MPI</a>,
C++ programs which
illustrate the use of the MPI application program interface
for carrying out parallel computations in a distributed memory environment.
</p>
<p>
<a href = "../../cpp_src/mpi_stubs/mpi_stubs.html">
MPI_STUBS</a>,
a C++ library which
contains "stub" MPI routines, allowing a user to compile, load, and possibly
run an MPI program on a serial machine.
</p>
<p>
<a href = "../../cpp_src/multitask_mpi/multitask_mpi.html">
MULTITASK_MPI</a>,
a C++ program which
demonstrates how to "multitask", that is, to execute several unrelated
and distinct tasks simultaneously, using MPI for parallel execution.
</p>
<p>
<a href = "../../cpp_src/prime_mpi/prime_mpi.html">
PRIME_MPI</a>,
a C++ program which
counts the number of primes between 1 and N, using MPI for parallel execution.
</p>
<p>
<a href = "../../cpp_src/quad_mpi/quad_mpi.html">
QUAD_MPI</a>,
a C++ program which
approximates an integral using a quadrature rule, and carries out the
computation in parallel using MPI.
</p>
<p>
<a href = "../../cpp_src/random_mpi/random_mpi.html">
RANDOM_MPI</a>,
a C++ program which
demonstrates one way to generate the same sequence of random numbers
for both sequential execution and parallel execution under MPI.
</p>
<p>
<a href = "../../cpp_src/ring_mpi/ring_mpi.html">
RING_MPI</a>,
a C++ program which
uses the MPI parallel programming environment, and measures the time
necessary to copy a set of data around a ring of processes.
</p>
<p>
<a href = "../../cpp_src/satisfy_mpi/satisfy_mpi.html">
SATISFY_MPI</a>,
a C++ program which
demonstrates, for a particular circuit, an exhaustive search
for solutions of the circuit satisfiability problem, using MPI to
carry out the calculation in parallel.
</p>
<p>
<a href = "../../cpp_src/search_mpi/search_mpi.html">
SEARCH_MPI</a>,
a C++ program which
searches integers between A and B for a value J such that F(J) = C,
using MPI.
</p>
<h3 align = "center">
Reference:
</h3>
<p>
<ol>
<li>
Geoffrey Fox, Mark Johnson, Gregory Lyzenga, Steve Otto, John Salmon,
David Walker,<br>
Solving problems on concurrent Processors,<br>
Volume 1: General Techniques and Regular Problems, <br>
Prentice Hall, 1988,<br>
ISBN: 0-13-8230226,<br>
LC: QA76.5.F627.
</li>
</ol>
</p>
<h3 align = "center">
Source Code:
</h3>
<p>
<ul>
<li>
<a href = "wave_mpi.cpp">wave_mpi.cpp</a>, the source code.
</li>
</ul>
</p>
<h3 align = "center">
Examples and Tests:
</h3>
<p>
<b>WAVE_FSU</b> compiles and runs the program on the FSU HPC cluster.
<ul>
<li>
<a href = "wave_fsu.sh">wave_fsu.sh</a>,
the MOAB script.
</li>
<li>
<a href = "wave_fsu_output.txt">wave_fsu_output.txt</a>,
the output file.
</li>
</ul>
</p>
<p>
<b>WAVE_LOCAL</b> compiles and runs the program on a local machine
with multiple processors.
<ul>
<li>
<a href = "wave_local.sh">wave_local.sh</a>,
the BASH script.
</li>
<li>
<a href = "wave_local_output.txt">wave_local_output.txt</a>,
the output file.
</li>
</ul>
</p>
<p>
You can go up one level to <a href = "../cpp_src.html">
the C++ source codes</a>.
</p>
<hr>
<i>
Last revised on 17 November 2013.
</i>
<!-- John Burkardt -->
</body>
<!-- Initial HTML skeleton created by HTMLINDEX. -->
</html>