Commit a67a1dd2b58c27862714a21269b2f77b69304b74
1 parent
5a17f58adf
Exists in
master
rewrite error with fmax
Showing 4 changed files with 108 additions and 128 deletions Inline Diff
lab3/Makefile
View file @
a67a1dd
GCC = gcc | 1 | 1 | GCC = gcc | |
CFLAGS = -O3 -fopenmp | 2 | 2 | CFLAGS = -O3 -fopenmp | |
3 | LDFLAG = -lm | |||
OMP_FLAG = -fopenmp | 3 | 4 | OMP_FLAG = -fopenmp | |
RM = rm -rf | 4 | 5 | RM = rm -rf | |
MPI = mpicc | 5 | 6 | MPI = mpicc | |
MPI_FLAG = -O1 -g | 6 | 7 | MPI_FLAG = -O1 -g | |
EXE = omp_heat2D ser_heat2D mpi_heat2D | 7 | 8 | EXE = omp_heat2D ser_heat2D mpi_heat2D | |
8 | 9 | |||
all : $(EXE) | 9 | 10 | all : $(EXE) | |
10 | 11 | |||
#.PHONY: all clean purge | 11 | 12 | #.PHONY: all clean purge | |
12 | 13 | |||
13 | 14 | |||
ser_heat2D: ser_heat2D.o | 14 | 15 | ser_heat2D: ser_heat2D.o | |
$(GCC) $(CFLAGS) -o $@ $^ | 15 | 16 | $(GCC) $(CFLAGS) -o $@ $^ $(LDFLAG) | |
16 | 17 | |||
omp_heat2D: omp_heat2D.o | 17 | 18 | omp_heat2D: omp_heat2D.o | |
$(GCC) $(CFLAGS) -o $@ $^ | 18 | 19 | $(GCC) $(CFLAGS) -o $@ $^ $(LDFLAG) | |
19 | 20 | |||
mpi_heat2D: | 20 | 21 | mpi_heat2D: | |
$(MPI) $(MPI_FLAG) mpi_heat2D.c -o $@ | 21 | 22 | $(MPI) $(MPI_FLAG) mpi_heat2D.c -o $@ $(LDFLAG) | |
22 | 23 | |||
23 | 24 | |||
24 | 25 | |||
%.o :%.c | 25 | 26 | %.o :%.c |
lab3/mpi_heat2D.c
View file @
a67a1dd
/**************************************************************************** | 1 | 1 | /**************************************************************************** | |
* DESCRIPTION: | 2 | 2 | * DESCRIPTION: | |
* Serial HEAT2D Example - C Version | 3 | 3 | * MPI HEAT2D Example - C Version | |
* This example is based on a simplified | 4 | 4 | * This example is based on a simplified | |
* two-dimensional heat equation domain decomposition. The initial | 5 | 5 | * two-dimensional heat equation domain decomposition. The initial | |
* temperature is computed to be high in the middle of the domain and | 6 | 6 | * temperature is computed to be high in the middle of the domain and | |
* zero at the boundaries. The boundaries are held at zero throughout | 7 | 7 | * zero at the boundaries. The boundaries are held at zero throughout | |
* the simulation. During the time-stepping, an array containing two | 8 | 8 | * the simulation. During the time-stepping, an array containing two | |
* domains is used; these domains alternate between old data and new data. | 9 | 9 | * domains is used; these domains alternate between old data and new data. | |
* | 10 | 10 | * | |
* The physical region, and the boundary conditions, are suggested | 11 | 11 | * The physical region, and the boundary conditions, are suggested | |
by this diagram; | 12 | 12 | by this diagram; | |
13 | 13 | |||
u = 0 | 14 | 14 | u = 0 | |
+------------------+ | 15 | 15 | +------------------+ | |
| | | 16 | 16 | | | | |
u = 100 | | u = 100 | 17 | 17 | u = 100 | | u = 100 | |
| | | 18 | 18 | | | | |
| | | 19 | 19 | | | | |
| | | 20 | 20 | | | | |
| | | 21 | 21 | | | | |
+------------------+ | 22 | 22 | +------------------+ | |
u = 100 | 23 | 23 | u = 100 | |
24 | 24 | |||
Interrior point : | 25 | 25 | Interrior point : | |
u[Central] = (1/4) * ( u[North] + u[South] + u[East] + u[West] ) | 26 | 26 | u[Central] = (1/4) * ( u[North] + u[South] + u[East] + u[West] ) | |
27 | 27 | |||
28 | 28 | |||
PARALLEL MPI VERSION : | 29 | 29 | PARALLEL MPI VERSION : | |
30 | 30 | |||
+-------------------+ | 31 | 31 | +-------------------+ | |
| | P0 m=(n-2)/P +2 | 32 | 32 | | | P0 m=(n-2)/P +2 | |
+-------------------+ | 33 | 33 | +-------------------+ | |
| | P1 | 34 | 34 | | | P1 | |
+-------------------+ | 35 | 35 | +-------------------+ | |
n | | .. | 36 | 36 | n | | .. | |
+-------------------+ | 37 | 37 | +-------------------+ | |
| | Pq | 38 | 38 | | | Pq | |
+-------------------+ | 39 | 39 | +-------------------+ | |
40 | 40 | |||
<-------- n --------> | 41 | 41 | <-------- n --------> | |
<-------n-2 ------> | 42 | 42 | <-------n-2 ------> | |
43 | 43 | |||
44 | 44 | |||
****************************************************************************/ | 45 | 45 | ****************************************************************************/ | |
#include <stdio.h> | 46 | 46 | #include <stdio.h> | |
#include <stdlib.h> | 47 | 47 | #include <stdlib.h> | |
#include <math.h> | 48 | 48 | #include <math.h> | |
#include <mpi.h> | 49 | 49 | #include <mpi.h> | |
#define NN 50 | 50 | 50 | #define NN 50 | |
#define MM 50 | 51 | 51 | #define MM 50 | |
52 | 52 | |||
#define RING 100 | 53 | 53 | #define RING 100 | |
#define ITER_PRINT 100 | 54 | 54 | #define ITER_PRINT 100 | |
#define PRINT_DATA 1 | 55 | 55 | #define PRINT_DATA 1 | |
#define MAX_ITER 1000 | 56 | 56 | #define _MAX_ITER 1000 | |
#define _EPSILON 0.01 | 57 | 57 | #define _EPSILON 0.01 | |
58 | 58 | |||
59 | 59 | |||
60 | ||||
61 | ||||
float update(int rank,int size, int nx,int ny, float *u, float *unew); | 60 | 62 | float update(int rank,int size, int nx,int ny, float *u, float *unew); | |
void inidat(int rank, int size, int nx, int ny, float *u, float *unew); | 61 | 63 | void inidat(int rank, int size, int nx, int ny, float *u, float *unew); | |
void prtdat(int rank, int size, int nx, int ny, float *u,const char *fnam); | 62 | 64 | void prtdat(int rank, int size, int nx, int ny, float *u,const char *fnam); | |
63 | 65 | |||
64 | 66 | |||
65 | 67 | |||
66 | 68 | |||
int main(int argc, char *argv[]) | 67 | 69 | int main(int argc, char *argv[]) | |
{ | 68 | 70 | { | |
int N=NN,M=MM; | 69 | 71 | int N=NN,M=MM; | |
70 | ||||
int rank,size; | 71 | |||
72 | ||||
float EPSILON=_EPSILON; | 73 | 72 | float EPSILON=_EPSILON; | |
74 | 73 | int MAX_ITER= _MAX_ITER; | ||
75 | 74 | |||
75 | int size,rank; | |||
76 | ||||
/* INITIALIZE MPI */ | 76 | 77 | /* INITIALIZE MPI */ | |
MPI_Init(&argc, &argv); | 77 | 78 | MPI_Init(&argc, &argv); | |
78 | 79 | |||
/* GET THE PROCESSOR ID AND NUMBER OF PROCESSORS */ | 79 | 80 | /* GET THE PROCESSOR ID AND NUMBER OF PROCESSORS */ | |
MPI_Comm_rank(MPI_COMM_WORLD, &rank); | 80 | 81 | MPI_Comm_rank(MPI_COMM_WORLD, &rank); | |
MPI_Comm_size(MPI_COMM_WORLD, &size); | 81 | 82 | MPI_Comm_size(MPI_COMM_WORLD, &size); | |
82 | 83 | |||
//Only Rank 0 read application parameters | 83 | 84 | //Only Rank 0 read application parameters | |
if(rank==0) { | 84 | 85 | if(rank==0) { | |
85 | 86 | |||
if(argc !=3) | 86 | 87 | if(argc !=4) | |
{ | 87 | 88 | { | |
fprintf(stderr,"usage %s N EPSILON\n ", argv[0]); | 88 | 89 | fprintf(stderr,"usage %s N EPSILON MAX_ITER\n ", argv[0]); | |
fprintf(stderr,"\t\twhere N is GRID size, EPSILON is Tolerance\n"); | 89 | 90 | fprintf(stderr,"\t\twhere N is GRID size, EPSILON is Tolerance, MAX_ITER is max iteration\n"); | |
fprintf(stderr,"\t\texample N = 100, EPSILON = 0.1\n"); | 90 | 91 | fprintf(stderr,"\t\texample N = 100, EPSILON = 0.1, MAX_ITER=1000\n"); | |
return -1; | 91 | 92 | return -1; | |
} | 92 | 93 | } | |
93 | ||||
N = M = atoi(argv[1]); | 94 | 94 | N = M = atoi(argv[1]); | |
EPSILON = atof(argv[2]); | 95 | 95 | EPSILON = atof(argv[2]); | |
96 | 96 | MAX_ITER = atoi (argv[3]); | ||
if(N % size!=0) | 97 | 97 | if(N % size!=0) | |
{ | 98 | 98 | { | |
fprintf(stderr,"Grid Size MUST be divisible by the number of processors !"); | 99 | 99 | fprintf(stderr,"Grid Size MUST be divisible by the number of processors !"); | |
return -1; | 100 | 100 | return -1; | |
} | 101 | 101 | } | |
102 | 102 | |||
} | 103 | 103 | } | |
104 | 104 | |||
//Wait for rank 0 , all process start here | 105 | 105 | //Wait for rank 0 , all process start here | |
MPI_Barrier(MPI_COMM_WORLD); | 106 | 106 | MPI_Barrier(MPI_COMM_WORLD); | |
107 | 107 | |||
//Exchange N | 108 | 108 | //Exchange N | |
MPI_Bcast(&N , 1, MPI_FLOAT, 0 , MPI_COMM_WORLD); | 109 | 109 | MPI_Bcast(&N , 1, MPI_INT, 0 , MPI_COMM_WORLD); | |
//Exchange EPSILON | 110 | 110 | //Exchange EPSILON | |
MPI_Bcast(&EPSILON , 1, MPI_FLOAT, 0 , MPI_COMM_WORLD); | 111 | 111 | MPI_Bcast(&EPSILON , 1, MPI_FLOAT, 0 , MPI_COMM_WORLD); | |
112 | ||||
113 | //Exchange MAX_ITER | |||
114 | MPI_Bcast(&MAX_ITER , 1, MPI_INT, 0 , MPI_COMM_WORLD); | |||
112 | 115 | |||
//local size | 113 | 116 | //local size | |
M = (N-2) / size + 2 ; | 114 | 117 | M = (N-2) / size + 2 ; | |
115 | 118 | |||
float *u = (float *)malloc(N * M * sizeof(float)); | 116 | 119 | float *u = (float *)malloc(N * M * sizeof(float)); | |
float *unew = (float *)malloc(N * M * sizeof(float)); | 117 | 120 | float *unew = (float *)malloc(N * M * sizeof(float)); | |
118 | 121 | |||
if(u==0 || unew ==0) | 119 | 122 | if(u==0 || unew ==0) | |
{ | 120 | 123 | { | |
perror("Can't allocated data\n"); | 121 | 124 | perror("Can't allocated data\n"); | |
return -1; | 122 | 125 | return -1; | |
} | 123 | 126 | } | |
124 | 127 | |||
if(rank==0) { | 125 | 128 | if(rank==0) { | |
126 | 129 | |||
printf ( "\n" ); | 127 | 130 | printf ( "\n" ); | |
printf ( "HEATED_PLATE\n" ); | 128 | 131 | printf ( "HEATED_PLATE\n" ); | |
printf ( " Parallel MPI version using %d processors \n",size ); | 129 | 132 | printf ( " Parallel MPI version using %d processors \n",size ); | |
printf ( " A program to solve for the steady state temperature distribution\n" ); | 130 | 133 | printf ( " A program to solve for the steady state temperature distribution\n" ); | |
printf ( " over a rectangular plate.\n" ); | 131 | 134 | printf ( " over a rectangular plate.\n" ); | |
printf ( "\n" ); | 132 | 135 | printf ( "\n" ); | |
printf ( " Spatial grid of %d by %d points.\n", N, N ); | 133 | 136 | printf ( " Spatial grid of %d by %d points.\n", N, N ); | |
printf ( " Each processor will use grid of %d by %d points.\n", M, N ); | 134 | 137 | printf ( " Each processor will use grid of %d by %d points.\n", M, N ); | |
printf ( " The iteration will end until tolerance <= %f\n\n",EPSILON); | 135 | 138 | printf ( " The iteration will end until tolerance <= %f\n\n",EPSILON); | |
136 | 139 | |||
} | 137 | 140 | } | |
138 | 141 | |||
/* Initialize grid and create input file | 139 | 142 | /* Initialize grid and create input file | |
* each process initialize its part | 140 | 143 | * each process initialize its part | |
* */ | 141 | 144 | * */ | |
142 | 145 | |||
inidat(rank,size,M,N,u,unew); | 143 | 146 | inidat(rank,size,M,N,u,unew); | |
144 | 147 | |||
prtdat(rank,size,M,N,u, "initial.dat"); | 145 | 148 | prtdat(rank,size,M,N,u, "initial.dat"); | |
146 | 149 | |||
147 | 150 | |||
float diff, globaldiff=1.0; | 148 | 151 | float diff, globaldiff=1.0; | |
int iter=0; | 149 | 152 | int iter=0; | |
150 | 153 | |||
/* | 151 | 154 | /* | |
* iterate (JACOBI ITERATION) until the new solution unew differs from the old solution u | 152 | 155 | * iterate (JACOBI ITERATION) until the new solution unew differs from the old solution u | |
* by no more than EPSILON. | 153 | 156 | * by no more than EPSILON. | |
**/ | 154 | 157 | **/ | |
155 | 158 | |||
while(globaldiff> EPSILON) { | 156 | 159 | while(globaldiff> EPSILON && iter < MAX_ITER) { | |
157 | 160 | |||
diff= update(rank,size,M,N, u, unew); | 158 | 161 | diff= update(rank,size,M,N, u, unew); | |
159 | 162 | |||
/** | 160 | 163 | /** | |
* COMPUTE GLOBAL CONVERGENCE | 161 | 164 | * COMPUTE GLOBAL CONVERGENCE | |
* */ | 162 | 165 | * */ | |
163 | 166 | |||
MPI_Allreduce(&diff, &globaldiff , 1, MPI_FLOAT, MPI_MAX, MPI_COMM_WORLD); | 164 | 167 | MPI_Allreduce(&diff, &globaldiff , 1, MPI_FLOAT, MPI_MAX, MPI_COMM_WORLD); | |
165 | 168 | |||
166 | 169 | |||
if(rank==0) | 167 | 170 | if(rank==0) | |
if(iter%ITER_PRINT==0) | 168 | 171 | if(iter%ITER_PRINT==0) | |
printf("Processor #%d, iteration %d, epsilon = %f\n ", rank,iter,globaldiff); | 169 | 172 | printf("Processor #%d, iteration %d, epsilon = %f\n ", rank,iter,globaldiff); | |
iter++; | 170 | 173 | iter++; | |
} | 171 | 174 | } | |
175 | ||||
172 | 176 | |||
prtdat(rank,size,M,N, u, "final.dat"); | 173 | 177 | prtdat(rank,size,M,N, u, "final.dat"); | |
free(u); | 174 | 178 | free(u); | |
free(unew); | 175 | 179 | free(unew); | |
MPI_Finalize(); | 176 | 180 | MPI_Finalize(); | |
} | 177 | 181 | } | |
178 | 182 | |||
179 | 183 | |||
180 | 184 | |||
/**************************************************************************** | 181 | 185 | /**************************************************************************** | |
* subroutine update | 182 | 186 | * subroutine update | |
****************************************************************************/ | 183 | 187 | ****************************************************************************/ | |
float update(int rank, int size, int nx,int ny, float *u, float *unew){ | 184 | 188 | float update(int rank, int size, int nx,int ny, float *u, float *unew){ | |
int ix, iy; | 185 | 189 | int ix, iy; | |
float diff=0.0; | 186 | 190 | float diff=0.0; | |
MPI_Status status; | 187 | 191 | MPI_Status status; | |
188 | 192 | |||
189 | 193 | |||
/* | 190 | 194 | /* | |
* EXCHANGE GHOST CELL | 191 | 195 | * EXCHANGE GHOST CELL | |
*/ | 192 | 196 | */ | |
if (rank > 0 && rank< size-1) | 193 | 197 | if (rank > 0 && rank< size-1) | |
{ | 194 | 198 | { | |
MPI_Sendrecv(&u[ny*(nx-2)], ny, MPI_FLOAT, rank+1, 0, | 195 | 199 | MPI_Sendrecv(&u[ny*(nx-2)], ny, MPI_FLOAT, rank+1, 0, | |
&u[ny*0], ny, MPI_FLOAT, rank-1, 0, MPI_COMM_WORLD, &status); | 196 | 200 | &u[ny*0], ny, MPI_FLOAT, rank-1, 0, MPI_COMM_WORLD, &status); | |
MPI_Sendrecv(&u[ny*1], ny, MPI_FLOAT, rank-1, 1, | 197 | 201 | MPI_Sendrecv(&u[ny*1], ny, MPI_FLOAT, rank-1, 1, | |
&u[ny*(nx-1)], ny, MPI_FLOAT, rank+1, 1, MPI_COMM_WORLD, &status); | 198 | 202 | &u[ny*(nx-1)], ny, MPI_FLOAT, rank+1, 1, MPI_COMM_WORLD, &status); | |
} | 199 | 203 | } | |
200 | 204 | |||
else if (rank == 0) | 201 | 205 | else if (rank == 0) | |
MPI_Sendrecv(&u[ny*(nx-2)], ny, MPI_FLOAT, rank+1, 0, | 202 | 206 | MPI_Sendrecv(&u[ny*(nx-2)], ny, MPI_FLOAT, rank+1, 0, | |
&u[ny*(nx-1)], ny, MPI_FLOAT, rank+1, 1, MPI_COMM_WORLD, &status); | 203 | 207 | &u[ny*(nx-1)], ny, MPI_FLOAT, rank+1, 1, MPI_COMM_WORLD, &status); | |
else if (rank == size-1) | 204 | 208 | else if (rank == size-1) | |
MPI_Sendrecv(&u[ny*1], ny, MPI_FLOAT, rank-1, 1, | 205 | 209 | MPI_Sendrecv(&u[ny*1], ny, MPI_FLOAT, rank-1, 1, | |
&u[ny*0], ny, MPI_FLOAT, rank-1, 0, MPI_COMM_WORLD, &status); | 206 | 210 | &u[ny*0], ny, MPI_FLOAT, rank-1, 0, MPI_COMM_WORLD, &status); | |
207 | 211 | |||
208 | 212 | |||
209 | 213 | |||
/** | 210 | 214 | /** | |
* PERFORM LOCAL COMPUTATION | 211 | 215 | * PERFORM LOCAL COMPUTATION | |
* */ | 212 | 216 | * */ | |
213 | 217 | |||
for (ix = 1; ix < nx-1; ix++) { | 214 | 218 | for (ix = 1; ix < nx-1; ix++) { | |
for (iy = 1; iy < ny-1; iy++) { | 215 | 219 | for (iy = 1; iy < ny-1; iy++) { | |
unew[ix*ny+iy] = (u[(ix+1)*ny+iy] + u[(ix-1)*ny+iy] + u[ix*ny+iy+1] + u[ix*ny+iy-1] )/4.0 | 216 | 220 | unew[ix*ny+iy] = (u[(ix+1)*ny+iy] + u[(ix-1)*ny+iy] + u[ix*ny+iy+1] + u[ix*ny+iy-1] )/4.0 | |
; | 217 | 221 | ; | |
if (diff < fabs (unew[ix*ny+iy] - u[ix*ny+iy] )) | 218 | 222 | diff = fmax( diff, fabs(unew[ix*ny+iy] - u[ix*ny+iy])); | |
{ | 219 | |||
diff = fabs ( unew[ix*ny+iy] - u[ix*ny+iy] ); | 220 | |||
} | 221 | |||
} | 222 | 223 | } | |
223 | 224 | |||
} | 224 | 225 | } | |
225 | 226 | |||
226 | 227 | |||
for (ix = 1; ix < nx-1; ix++) { | 227 | 228 | for (ix = 1; ix < nx-1; ix++) { | |
for (iy = 1; iy < ny-1; iy++) { | 228 | 229 | for (iy = 1; iy < ny-1; iy++) { | |
u[ix*ny+iy] = unew[ix*ny+iy]; | 229 | 230 | u[ix*ny+iy] = unew[ix*ny+iy]; | |
} | 230 | 231 | } | |
} | 231 | 232 | } | |
232 | 233 | |||
233 | 234 | |||
return diff; | 234 | 235 | return diff; | |
} | 235 | 236 | } | |
236 | 237 | |||
/***************************************************************************** | 237 | 238 | /***************************************************************************** | |
* Initialize Data | 238 | 239 | * Initialize Data | |
*****************************************************************************/ | 239 | 240 | *****************************************************************************/ | |
240 | 241 | |||
void inidat(int rank, int size,int nx, int ny, float *u, float *unew) | 241 | 242 | void inidat(int rank, int size,int nx, int ny, float *u, float *unew) | |
{ | 242 | 243 | { | |
int ix, iy; | 243 | 244 | int ix, iy; | |
244 | 245 | |||
/* | 245 | 246 | /* | |
*Set boundary data and interrior values | 246 | 247 | *Set boundary data and interrior values | |
* */ | 247 | 248 | * */ | |
248 | 249 | |||
249 | 250 | |||
// interior points | 250 | 251 | // interior points | |
for (ix = 1; ix < nx-1; ix++) | 251 | 252 | for (ix = 1; ix < nx-1; ix++) | |
for (iy = 1; iy < ny-1; iy++) { | 252 | 253 | for (iy = 1; iy < ny-1; iy++) { | |
u[ix*ny+iy]=0.0; | 253 | 254 | u[ix*ny+iy]=0.0; | |
} | 254 | 255 | } | |
255 | 256 | |||
//boundary left | 256 | 257 | //boundary left | |
for (ix = 0; ix < nx; ix++){ | 257 | 258 | for (ix = 0; ix < nx; ix++){ | |
u[ix*ny]=100.0; | 258 | 259 | u[ix*ny]=100.0; |
lab3/omp_heat2D.c
View file @
a67a1dd
/**************************************************************************** | 1 | 1 | /**************************************************************************** | |
* DESCRIPTION: | 2 | 2 | * DESCRIPTION: | |
* Serial HEAT2D Example - C Version | 3 | 3 | * OpenMP HEAT2D Example - C Version | |
* This example is based on a simplified | 4 | 4 | * This example is based on a simplified | |
* two-dimensional heat equation domain decomposition. The initial | 5 | 5 | * two-dimensional heat equation domain decomposition. The initial | |
* temperature is computed to be high in the middle of the domain and | 6 | 6 | * temperature is computed to be high in the middle of the domain and | |
* zero at the boundaries. The boundaries are held at zero throughout | 7 | 7 | * zero at the boundaries. The boundaries are held at zero throughout | |
* the simulation. During the time-stepping, an array containing two | 8 | 8 | * the simulation. During the time-stepping, an array containing two | |
* domains is used; these domains alternate between old data and new data. | 9 | 9 | * domains is used; these domains alternate between old data and new data. | |
* | 10 | 10 | * | |
* The physical region, and the boundary conditions, are suggested | 11 | 11 | * The physical region, and the boundary conditions, are suggested | |
by this diagram; | 12 | 12 | by this diagram; | |
13 | 13 | |||
u = 0 | 14 | 14 | u = 0 | |
+------------------+ | 15 | 15 | +------------------+ | |
| | | 16 | 16 | | | | |
u = 100 | | u = 100 | 17 | 17 | u = 100 | | u = 100 | |
| | | 18 | 18 | | | | |
+------------------+ | 19 | 19 | +------------------+ | |
u = 100 | 20 | 20 | u = 100 | |
21 | 21 | |||
Interrior point : | 22 | 22 | Interrior point : | |
u[Central] = (1/4) * ( u[North] + u[South] + u[East] + u[West] ) | 23 | 23 | u[Central] = (1/4) * ( u[North] + u[South] + u[East] + u[West] ) | |
24 | 24 | |||
****************************************************************************/ | 25 | 25 | ****************************************************************************/ | |
#include <stdio.h> | 26 | 26 | ||
27 | #include <stdio.h> | |||
#include <stdlib.h> | 27 | 28 | #include <stdlib.h> | |
#include <math.h> | 28 | 29 | #include <math.h> | |
#include <omp.h> | 29 | |||
30 | 30 | |||
#define NN 50 | 31 | 31 | #define NN 50 | |
#define MM 50 | 32 | 32 | #define MM 50 | |
33 | 33 | |||
#define ITER_PRINT 100 | 34 | 34 | #define ITER_PRINT 100 | |
35 | #define _MAX_ITER 400 | |||
#define PRINT_DATA 1 | 35 | 36 | #define PRINT_DATA 1 | |
37 | #define _EPSILON 0.01 | |||
36 | 38 | |||
#define _EPSILON 0.001 | 37 | |||
38 | 39 | |||
39 | ||||
float update(int nx,int ny, float *u, float *unew); | 40 | 40 | float update(int nx,int ny, float *u, float *unew); | |
void inidat(int nx, int ny, float *u, float *unew); | 41 | 41 | void inidat(int nx, int ny, float *u, float *unew); | |
void prtdat(int nx, int ny, float *u,const char *fnam); | 42 | 42 | void prtdat(int nx, int ny, float *u,const char *fnam); | |
43 | 43 | |||
44 | 44 | |||
45 | 45 | |||
46 | 46 | |||
int main(int argc, char *argv[]) | 47 | 47 | int main(int argc, char *argv[]) | |
{ | 48 | 48 | { | |
49 | ||||
float diff=1.0; | 50 | |||
float EPSILON=_EPSILON; | 51 | |||
int N=NN,M=MM; | 52 | 49 | int N=NN,M=MM; | |
50 | float EPSILON=_EPSILON; | |||
51 | int MAX_ITER= _MAX_ITER; | |||
53 | 52 | |||
if(argc !=3) | 54 | 53 | if(argc !=4) | |
{ | 55 | 54 | { | |
fprintf(stderr,"usage %s N EPSILON\n ", argv[0]); | 56 | 55 | fprintf(stderr,"usage %s N EPSILON MAX_ITER\n ", argv[0]); | |
fprintf(stderr,"\t\twhere N is GRID size, EPSILON is Tolerance\n"); | 57 | 56 | fprintf(stderr,"\t\twhere N is GRID size, EPSILON is Tolerance, MAX_ITER is max iteration\n"); | |
fprintf(stderr,"\t\texample N = 100, EPSILON = 0.1\n"); | 58 | 57 | fprintf(stderr,"\t\texample N = 100, EPSILON = 0.1, MAX_ITER=1000\n"); | |
return -1; | 59 | 58 | return -1; | |
} | 60 | 59 | } | |
61 | 60 | |||
N = M = atoi(argv[1]); | 62 | 61 | N = M = atoi(argv[1]); | |
EPSILON = atof(argv[2]); | 63 | 62 | EPSILON = atof(argv[2]); | |
63 | MAX_ITER = atoi (argv[3]); | |||
64 | 64 | |||
65 | float diff=1.0; | |||
66 | ||||
float *u = (float *)malloc(N * M * sizeof(float)); | 65 | 67 | float *u = (float *)malloc(N * M * sizeof(float)); | |
float *unew = (float *)malloc(N * M * sizeof(float)); | 66 | 68 | float *unew = (float *)malloc(N * M * sizeof(float)); | |
67 | 69 | |||
if(u==0 || unew ==0) | 68 | 70 | if(u==0 || unew ==0) | |
{ | 69 | 71 | { | |
perror("Can't allocated data\n"); | 70 | 72 | perror("Can't allocated data\n"); | |
return -1; | 71 | 73 | return -1; | |
} | 72 | 74 | } | |
73 | 75 | |||
printf ( "\n" ); | 74 | 76 | printf ( "\n" ); | |
printf ( "HEATED_PLATE\n" ); | 75 | 77 | printf ( "HEATED_PLATE\n" ); | |
printf ( " Parallel OpenMP version, using %d Threads\n",omp_get_max_threads() ); | 76 | 78 | printf ( " OpenMP version\n" ); | |
printf ( " A program to solve for the steady state temperature distribution\n" ); | 77 | 79 | printf ( " A program to solve for the steady state temperature distribution\n" ); | |
printf ( " over a rectangular plate.\n" ); | 78 | 80 | printf ( " over a rectangular plate.\n" ); | |
printf ( " Spatial grid of %d by %d points.\n\n", M, N ); | 79 | 81 | printf ( "\n" ); | |
82 | printf ( " Spatial grid of %d by %d points.\n", M, N ); | |||
83 | printf ( " The iteration will end until tolerance <= %f\n\n",EPSILON); | |||
80 | 84 | |||
81 | ||||
/* Initialize grid and create input file */ | 82 | 85 | /* Initialize grid and create input file */ | |
printf("Initializing grid\n"); | 83 | 86 | printf("Initializing grid\n"); | |
84 | 87 | |||
inidat(N, M,u,unew); | 85 | 88 | inidat(N, M,u,unew); | |
86 | 89 | |||
prtdat(N, M,u, "initial.dat"); | 87 | 90 | prtdat(N, M,u, "initial.dat"); | |
88 | 91 | |||
printf("Start computing\n"); | 89 | 92 | printf("Start computing\n\n"); | |
90 | 93 | |||
int iter=0; | 91 | 94 | int iter=0; | |
92 | 95 | |||
/* | 93 | 96 | /* | |
* iterate until the new solution unew differs from the old solution u | 94 | 97 | * iterate until the new solution unew differs from the old solution u | |
* by no more than EPSILON. | 95 | 98 | * by no more than EPSILON. | |
* */ | 96 | 99 | * */ | |
97 | 100 | |||
while(diff> EPSILON) { | 98 | 101 | while(diff> EPSILON && iter <MAX_ITER) { | |
99 | 102 | |||
diff = update(N, M, u, unew); | 100 | 103 | diff= update(N, M, u, unew); | |
101 | 104 | |||
if(iter%ITER_PRINT==0) | 102 | 105 | if(iter%ITER_PRINT==0) | |
printf("Iteration %d, diff = %f\n ", iter,diff); | 103 | |||
104 | 106 | |||
107 | printf("Iteration %d, diff = %e\n ", iter,diff); | |||
108 | ||||
iter++; | 105 | 109 | iter++; | |
} | 106 | 110 | } | |
107 | 111 | |||
112 | if(diff>EPSILON) | |||
113 | printf("***Not converged MAX_ITERATION %d reached\n***", MAX_ITER); | |||
114 | ||||
prtdat(N, M, u, "final.dat"); | 108 | 115 | prtdat(N, M, u, "final.dat"); | |
109 | 116 | |||
free(u); | 110 | 117 | free(u); | |
free(unew); | 111 | 118 | free(unew); | |
} | 112 | 119 | } | |
113 | 120 | |||
114 | 121 | |||
115 | 122 | |||
/**************************************************************************** | 116 | 123 | /**************************************************************************** | |
* subroutine update | 117 | 124 | * subroutine update | |
****************************************************************************/ | 118 | 125 | ****************************************************************************/ | |
float update(int nx,int ny, float *u, float *unew) | 119 | 126 | float update(int nx,int ny, float *u, float *unew) | |
{ | 120 | 127 | { | |
int ix, iy; | 121 | 128 | int ix, iy; | |
float diff=0.0; | 122 | 129 | float diff=0.0; | |
123 | 130 | |||
#pragma omp parallel for shared(nx,ny,u,unew) private (ix,iy) | 124 | 131 | #pragma omp parallel for shared(nx,ny,u,unew) private (ix,iy) reduction(max:diff) | |
for (ix = 1; ix < nx-1; ix++) { | 125 | 132 | for (ix = 1; ix < nx-1; ix++) { | |
for (iy = 1; iy < ny-1; iy++) { | 126 | 133 | for (iy = 1; iy < ny-1; iy++) { | |
unew[ix*ny+iy] = | 127 | 134 | unew[ix*ny+iy] = | |
(u[(ix+1)*ny+iy] + u[(ix-1)*ny+iy] + | 128 | 135 | (u[(ix+1)*ny+iy] + u[(ix-1)*ny+iy] + | |
u[ix*ny+iy+1] + u[ix*ny+iy-1] )/4.0; | 129 | 136 | u[ix*ny+iy+1] + u[ix*ny+iy-1] )*0.25; | |
130 | 137 | diff = fmax( diff, fabs(unew[ix*ny+iy] - u[ix*ny+iy])); | ||
} | 131 | 138 | } | |
} | 132 | 139 | } | |
140 | ||||
133 | 141 | |||
//compute reduction | 134 | |||
135 | ||||
136 | ||||
float mydiff; | 137 | |||
138 | ||||
/** | 139 | |||
* IMPLEMENT OMP REDUCE MAX | 140 | |||
*/ | 141 | |||
142 | ||||
#pragma omp parallel shared(nx,ny,u,unew, diff) private (ix,iy,mydiff) | 143 | |||
{ | 144 | |||
mydiff=0.0; | 145 | |||
#pragma omp for | 146 | |||
for (ix = 1; ix < nx-1; ix++) { | 147 | |||
for (iy = 1; iy < ny-1; iy++) { | 148 | |||
if (mydiff < fabs (unew[ix*ny+iy] - u[ix*ny+iy] )) | 149 | |||
{ | 150 | |||
mydiff = fabs ( unew[ix*ny+iy] - u[ix*ny+iy] ); | 151 | |||
} | 152 | |||
} | 153 | |||
} | 154 | |||
155 | ||||
156 | ||||
#pragma omp critical | 157 | |||
{ | 158 | |||
if (diff < mydiff ) | 159 | |||
{ | 160 | |||
diff = mydiff; | 161 | |||
} | 162 | |||
} | 163 | |||
164 | ||||
/* | 165 | 142 | /* | |
* COPY OLD DATA | 166 | 143 | * COPY OLD DATA | |
*/ | 167 | 144 | */ | |
#pragma omp for | 168 | 145 | #pragma omp parallel for | |
for (ix = 1; ix < nx-1; ix++) { | 169 | 146 | for (ix = 1; ix < nx-1; ix++) { | |
for (iy = 1; iy < ny-1; iy++) { | 170 | 147 | for (iy = 1; iy < ny-1; iy++) { | |
u[ix*ny+iy] = unew[ix*ny+iy]; | 171 | 148 | u[ix*ny+iy] = unew[ix*ny+iy]; | |
} | 172 | 149 | } | |
} | 173 | 150 | } | |
} | 174 | 151 | ||
175 | 152 | |||
return diff; | 176 | 153 | return diff; | |
} | 177 | 154 | } | |
178 | 155 | |||
/***************************************************************************** | 179 | 156 | /***************************************************************************** | |
* Initialize Data | 180 | 157 | * Initialize Data | |
*****************************************************************************/ | 181 | 158 | *****************************************************************************/ | |
182 | 159 | |||
void inidat(int nx, int ny, float *u, float *unew) | 183 | 160 | void inidat(int nx, int ny, float *u, float *unew) | |
{ | 184 | 161 | { | |
int ix, iy; | 185 | 162 | int ix, iy; | |
186 | 163 | |||
/* | 187 | 164 | /* | |
*Set boundary data and interrior values | 188 | 165 | *Set boundary data and interrior values | |
* */ | 189 | 166 | * */ | |
190 | 167 | |||
#pragma omp parallel private (ix,iy) | 191 | 168 | #pragma omp parallel private (ix,iy) | |
{ | 192 | 169 | { | |
// interior points | 193 | 170 | // interior points | |
#pragma omp for | 194 | 171 | #pragma omp for | |
for (ix = 1; ix < nx-1; ix++) | 195 | 172 | for (ix = 1; ix < nx-1; ix++) | |
for (iy = 1; iy < ny-1; iy++) { | 196 | 173 | for (iy = 1; iy < ny-1; iy++) { | |
u[ix*ny+iy]=5.0; | 197 | 174 | u[ix*ny+iy]=0.0; | |
} | 198 | 175 | } | |
199 | 176 | |||
//boundary left | 200 | 177 | //boundary left | |
#pragma omp for | 201 | 178 | #pragma omp for | |
for (ix = 1; ix < nx-1; ix++){ | 202 | 179 | for (ix = 1; ix < nx-1; ix++){ | |
u[ix*ny]=100.0; | 203 | 180 | u[ix*ny]=100.0; | |
204 | 181 | |||
} | 205 | 182 | } | |
206 | 183 | |||
//boundary right | 207 | 184 | //boundary right | |
#pragma omp for | 208 | 185 | #pragma omp for | |
for (ix = 1; ix < nx-1; ix++){ | 209 | 186 | for (ix = 1; ix < nx-1; ix++){ | |
u[ix*ny+ (ny-1)]=100.0; | 210 | 187 | u[ix*ny+ (ny-1)]=100.0; | |
211 | 188 | |||
} | 212 | 189 | } |
lab3/ser_heat2D.c
View file @
a67a1dd
/**************************************************************************** | 1 | 1 | /**************************************************************************** | |
* DESCRIPTION: | 2 | 2 | * DESCRIPTION: | |
* Serial HEAT2D Example - C Version | 3 | 3 | * Serial HEAT2D Example - C Version | |
* This example is based on a simplified | 4 | 4 | * This example is based on a simplified | |
* two-dimensional heat equation domain decomposition. The initial | 5 | 5 | * two-dimensional heat equation domain decomposition. The initial | |
* temperature is computed to be high in the middle of the domain and | 6 | 6 | * temperature is computed to be high in the middle of the domain and | |
* zero at the boundaries. The boundaries are held at zero throughout | 7 | 7 | * zero at the boundaries. The boundaries are held at zero throughout | |
* the simulation. During the time-stepping, an array containing two | 8 | 8 | * the simulation. During the time-stepping, an array containing two | |
* domains is used; these domains alternate between old data and new data. | 9 | 9 | * domains is used; these domains alternate between old data and new data. | |
* | 10 | 10 | * | |
* The physical region, and the boundary conditions, are suggested | 11 | 11 | * The physical region, and the boundary conditions, are suggested | |
by this diagram; | 12 | 12 | by this diagram; | |
13 | 13 | |||
u = 0 | 14 | 14 | u = 0 | |
+------------------+ | 15 | 15 | +------------------+ | |
| | | 16 | 16 | | | | |
u = 100 | | u = 100 | 17 | 17 | u = 100 | | u = 100 | |
| | | 18 | 18 | | | | |
+------------------+ | 19 | 19 | +------------------+ | |
u = 100 | 20 | 20 | u = 100 | |
21 | 21 | |||
Interrior point : | 22 | 22 | Interrior point : | |
u[Central] = (1/4) * ( u[North] + u[South] + u[East] + u[West] ) | 23 | 23 | u[Central] = (1/4) * ( u[North] + u[South] + u[East] + u[West] ) | |
24 | 24 | |||
****************************************************************************/ | 25 | 25 | ****************************************************************************/ | |
#include <stdio.h> | 26 | 26 | #include <stdio.h> | |
#include <stdlib.h> | 27 | 27 | #include <stdlib.h> | |
#include <math.h> | 28 | 28 | #include <math.h> | |
29 | 29 | |||
#define NN 50 | 30 | 30 | #define NN 50 | |
#define MM 50 | 31 | 31 | #define MM 50 | |
32 | 32 | |||
#define ITER_PRINT 100 | 33 | 33 | #define ITER_PRINT 100 | |
34 | #define _MAX_ITER 400 | |||
#define PRINT_DATA 1 | 34 | 35 | #define PRINT_DATA 1 | |
35 | ||||
#define _EPSILON 0.01 | 36 | 36 | #define _EPSILON 0.01 | |
37 | 37 | |||
38 | 38 | |||
float update(int nx,int ny, float *u, float *unew); | 39 | 39 | float update(int nx,int ny, float *u, float *unew); | |
void inidat(int nx, int ny, float *u, float *unew); | 40 | 40 | void inidat(int nx, int ny, float *u, float *unew); | |
void prtdat(int nx, int ny, float *u,const char *fnam); | 41 | 41 | void prtdat(int nx, int ny, float *u,const char *fnam); | |
42 | 42 | |||
43 | 43 | |||
44 | 44 | |||
45 | 45 | |||
int main(int argc, char *argv[]) | 46 | 46 | int main(int argc, char *argv[]) | |
{ | 47 | 47 | { | |
int N=NN,M=MM; | 48 | 48 | int N=NN,M=MM; | |
49 | ||||
float EPSILON=_EPSILON; | 50 | 49 | float EPSILON=_EPSILON; | |
50 | int MAX_ITER= _MAX_ITER; | |||
51 | 51 | |||
if(argc !=3) | 52 | 52 | if(argc !=4) | |
{ | 53 | 53 | { | |
fprintf(stderr,"usage %s N EPSILON\n ", argv[0]); | 54 | 54 | fprintf(stderr,"usage %s N EPSILON MAX_ITER\n ", argv[0]); | |
fprintf(stderr,"\t\twhere N is GRID size, EPSILON is Tolerance\n"); | 55 | 55 | fprintf(stderr,"\t\twhere N is GRID size, EPSILON is Tolerance, MAX_ITER is max iteration\n"); | |
fprintf(stderr,"\t\texample N = 100, EPSILON = 0.1\n"); | 56 | 56 | fprintf(stderr,"\t\texample N = 100, EPSILON = 0.1, MAX_ITER=1000\n"); | |
return -1; | 57 | 57 | return -1; | |
} | 58 | 58 | } | |
59 | 59 | |||
N = M = atoi(argv[1]); | 60 | 60 | N = M = atoi(argv[1]); | |
EPSILON = atof(argv[2]); | 61 | 61 | EPSILON = atof(argv[2]); | |
62 | MAX_ITER = atoi (argv[3]); | |||
62 | 63 | |||
float diff=1.0; | 63 | 64 | float diff=1.0; | |
64 | 65 | |||
float *u = (float *)malloc(N * M * sizeof(float)); | 65 | 66 | float *u = (float *)malloc(N * M * sizeof(float)); | |
float *unew = (float *)malloc(N * M * sizeof(float)); | 66 | 67 | float *unew = (float *)malloc(N * M * sizeof(float)); | |
67 | 68 | |||
if(u==0 || unew ==0) | 68 | 69 | if(u==0 || unew ==0) | |
{ | 69 | 70 | { | |
perror("Can't allocated data\n"); | 70 | 71 | perror("Can't allocated data\n"); | |
return -1; | 71 | 72 | return -1; | |
} | 72 | 73 | } | |
73 | 74 | |||
printf ( "\n" ); | 74 | 75 | printf ( "\n" ); | |
printf ( "HEATED_PLATE\n" ); | 75 | 76 | printf ( "HEATED_PLATE\n" ); | |
printf ( " Serial version\n" ); | 76 | 77 | printf ( " Serial version\n" ); | |
printf ( " A program to solve for the steady state temperature distribution\n" ); | 77 | 78 | printf ( " A program to solve for the steady state temperature distribution\n" ); | |
printf ( " over a rectangular plate.\n" ); | 78 | 79 | printf ( " over a rectangular plate.\n" ); | |
printf ( "\n" ); | 79 | 80 | printf ( "\n" ); | |
printf ( " Spatial grid of %d by %d points.\n", M, N ); | 80 | 81 | printf ( " Spatial grid of %d by %d points.\n", M, N ); | |
printf ( " The iteration will end until tolerance <= %f\n\n",EPSILON); | 81 | 82 | printf ( " The iteration will end until tolerance <= %f\n\n",EPSILON); | |
82 | 83 | |||
/* Initialize grid and create input file */ | 83 | 84 | /* Initialize grid and create input file */ | |
printf("Initializing grid\n"); | 84 | 85 | printf("Initializing grid\n"); | |
85 | 86 | |||
inidat(N, M,u,unew); | 86 | 87 | inidat(N, M,u,unew); | |
87 | 88 | |||
prtdat(N, M,u, "initial.dat"); | 88 | 89 | prtdat(N, M,u, "initial.dat"); | |
89 | 90 | |||
printf("Start computing\n\n"); | 90 | 91 | printf("Start computing\n\n"); | |
91 | 92 | |||
int iter=0; | 92 | 93 | int iter=0; | |
93 | 94 | |||
/* | 94 | 95 | /* | |
* iterate until the new solution unew differs from the old solution u | 95 | 96 | * iterate until the new solution unew differs from the old solution u | |
* by no more than EPSILON. | 96 | 97 | * by no more than EPSILON. | |
* */ | 97 | 98 | * */ | |
98 | ||||
while(diff> EPSILON) { | 99 | |||
100 | 99 | |||
diff= update(N, M, u, unew); | 101 | 100 | while(diff> EPSILON && iter <MAX_ITER) { | |
102 | 101 | |||
102 | diff= update(N, M, u, unew); | |||
103 | ||||
if(iter%ITER_PRINT==0) | 103 | 104 | if(iter%ITER_PRINT==0) | |
104 | 105 | |||
printf("Iteration %d, diff = %e\n ", iter,diff); | 105 | 106 | printf("Iteration %d, diff = %e\n ", iter,diff); | |
106 | 107 | |||
iter++; | 107 | 108 | iter++; | |
} | 108 | 109 | } | |
109 | 110 | |||
111 | if(diff>EPSILON) | |||
112 | printf("***Not converged MAX_ITERATION %d reached\n***", MAX_ITER); | |||
113 | ||||
prtdat(N, M, u, "final.dat"); | 110 | 114 | prtdat(N, M, u, "final.dat"); | |
111 | 115 | |||
free(u); | 112 | 116 | free(u); | |
free(unew); | 113 | 117 | free(unew); | |
} | 114 | 118 | } | |
115 | 119 | |||
116 | 120 | |||
117 | 121 | |||
/**************************************************************************** | 118 | 122 | /**************************************************************************** | |
* subroutine update | 119 | 123 | * subroutine update | |
****************************************************************************/ | 120 | 124 | ****************************************************************************/ | |
float update(int nx,int ny, float *u, float *unew) | 121 | 125 | float update(int nx,int ny, float *u, float *unew) | |
{ | 122 | 126 | { | |
int ix, iy; | 123 | 127 | int ix, iy; | |
float diff=0.0; | 124 | 128 | float diff=0.0; | |
125 | 129 | |||
for (ix = 1; ix < nx-1; ix++) { | 126 | 130 | for (ix = 1; ix < nx-1; ix++) { | |
for (iy = 1; iy < ny-1; iy++) { | 127 | 131 | for (iy = 1; iy < ny-1; iy++) { | |
unew[ix*ny+iy] = (u[(ix+1)*ny+iy] + u[(ix-1)*ny+iy] + u[ix*ny+iy+1] + u[ix*ny+iy-1] )/4.0 | 128 | 132 | ||
; | 129 | 133 | unew[ix*ny+iy] = (u[(ix+1)*ny+iy] + u[(ix-1)*ny+iy] + u[ix*ny+iy+1] + u[ix*ny+iy-1] )*0.25; | |
if (diff < fabs (unew[ix*ny+iy] - u[ix*ny+iy] )) | 130 | 134 | diff = fmax( diff, fabs(unew[ix*ny+iy] - u[ix*ny+iy])); | |
{ | 131 | |||
diff = fabs ( unew[ix*ny+iy] - u[ix*ny+iy] ); | 132 | |||
} | 133 | |||
} | 134 | 135 | } | |
135 | 136 | |||
} | 136 | 137 | } | |
137 | 138 | |||
/** | 138 | 139 | /** | |
* COPY OLD DATA | 139 | 140 | * COPY OLD DATA | |
*/ | 140 | 141 | */ | |
for (ix = 1; ix < nx-1; ix++) { | 141 | 142 | for (ix = 1; ix < nx-1; ix++) { | |
for (iy = 1; iy < ny-1; iy++) { | 142 | 143 | for (iy = 1; iy < ny-1; iy++) { | |
u[ix*ny+iy] = unew[ix*ny+iy]; | 143 | 144 | u[ix*ny+iy] = unew[ix*ny+iy]; | |
} | 144 | 145 | } | |
} | 145 | 146 | } | |
146 | 147 | |||
return diff; | 147 | 148 | return diff; | |
} | 148 | 149 | } | |
149 | 150 | |||
/***************************************************************************** | 150 | 151 | /***************************************************************************** | |
* Initialize Data | 151 | 152 | * Initialize Data | |
*****************************************************************************/ | 152 | 153 | *****************************************************************************/ | |
void inidat(int nx, int ny, float *u, float *unew) | 153 | 154 | void inidat(int nx, int ny, float *u, float *unew) | |
{ | 154 | 155 | { | |
int ix, iy; | 155 | 156 | int ix, iy; | |
156 | 157 | |||
/* | 157 | 158 | /* | |
*Set boundary data and interrior values | 158 | 159 | *Set boundary data and interrior values | |
* */ | 159 | 160 | * */ | |
160 | 161 | |||
161 | 162 | |||
// interior points | 162 | 163 | // interior points | |
for (ix = 1; ix < nx-1; ix++) | 163 | 164 | for (ix = 1; ix < nx-1; ix++) | |
for (iy = 1; iy < ny-1; iy++) { | 164 | 165 | for (iy = 1; iy < ny-1; iy++) { | |
u[ix*ny+iy]=5.0; | 165 | 166 | u[ix*ny+iy]=0.0; | |
} | 166 | 167 | } | |
167 | 168 | |||
//boundary left | 168 | 169 | //boundary left | |
for (ix = 1; ix < nx-1; ix++){ | 169 | 170 | for (ix = 1; ix < nx-1; ix++){ | |
u[ix*ny]=100.0; | 170 | 171 | u[ix*ny]=100.0; | |
171 | 172 | |||
} | 172 | 173 | } | |
173 | 174 | |||
//boundary right | 174 | 175 | //boundary right | |
for (ix = 1; ix < nx-1; ix++){ | 175 | 176 | for (ix = 1; ix < nx-1; ix++){ | |
u[ix*ny+ (ny-1)]=100.0; | 176 | 177 | u[ix*ny+ (ny-1)]=100.0; | |
177 | 178 | |||
} | 178 | 179 | } | |
179 | 180 | |||
//boundary down | 180 | 181 | //boundary down | |
for (iy = 0; iy < ny; iy++){ | 181 | 182 | for (iy = 0; iy < ny; iy++){ | |
u[(nx-1)*(ny)+iy]=100.0; | 182 | 183 | u[(nx-1)*(ny)+iy]=100.0; | |
183 | 184 | |||
} | 184 | 185 | } | |
185 | 186 | |||
//boundary top | 186 | 187 | //boundary top | |
for (iy = 0; iy < ny; iy++){ | 187 | 188 | for (iy = 0; iy < ny; iy++){ | |
u[iy]=100.0; | 188 | 189 | u[iy]=100.0; | |
189 | 190 | |||
} | 190 | 191 | } | |
191 | 192 | |||
192 | 193 | |||
} | 193 | 194 | } | |
194 | 195 | |||
195 | 196 | |||
196 | 197 | |||
/************************************************************************** | 197 | 198 | /************************************************************************** |