-
Notifications
You must be signed in to change notification settings - Fork 0
/
full_matrix.cu
316 lines (268 loc) · 11.7 KB
/
full_matrix.cu
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
#include <math.h>
#include <stdio.h>
#include <omp.h>
#include <cuda.h>
struct full_data
{
int sizex;
int sizey;
int Nmats;
double * __restrict__ rho;
double * __restrict__ rho_mat_ave;
double * __restrict__ p;
double * __restrict__ Vf;
double * __restrict__ t;
double * __restrict__ V;
double * __restrict__ x;
double * __restrict__ y;
double * __restrict__ n;
double * __restrict__ rho_ave;
};
__global__ void cc_loop1(const double * __restrict rho, const double * __restrict Vf, const double * __restrict V, double * __restrict rho_ave, int sizex, int sizey, int Nmats) {
int i = threadIdx.x + blockIdx.x * blockDim.x;
int j = threadIdx.y + blockIdx.y * blockDim.y;
if (i >= sizex || j >= sizey) return;
double ave = 0.0;
for (int mat = 0; mat < Nmats; mat++) {
// Optimisation:
if (Vf[(i+sizex*j)*Nmats+mat] > 0.0)
ave += rho[(i+sizex*j)*Nmats+mat]*Vf[(i+sizex*j)*Nmats+mat];
}
rho_ave[i+sizex*j] = ave/V[i+sizex*j];
}
__global__ void mc_loop1(const double * __restrict rho, const double * __restrict Vf, const double * __restrict V, double * __restrict rho_ave, int sizex, int sizey, int Nmats) {
int i = threadIdx.x + blockIdx.x * blockDim.x;
int j = threadIdx.y + blockIdx.y * blockDim.y;
if (i >= sizex || j >= sizey) return;
int ncells = sizex*sizey;
for (int mat = 0; mat < Nmats; mat++) {
if (Vf[ncells*mat + i+sizex*j] > 0.0)
rho_ave[i+sizex*j] += rho[ncells*mat + i+sizex*j] * Vf[ncells*mat + i+sizex*j];
}
rho_ave[i+sizex*j] /= V[i+sizex*j];
}
__global__ void cc_loop2(const double * __restrict rho, const double * __restrict Vf, const double * __restrict t, const double * __restrict n, double * __restrict p, int sizex, int sizey, int Nmats) {
int i = threadIdx.x + blockIdx.x * blockDim.x;
int j = threadIdx.y + blockIdx.y * blockDim.y;
if (i >= sizex || j >= sizey) return;
for (int mat = 0; mat < Nmats; mat++) {
if (Vf[(i+sizex*j)*Nmats+mat] > 0.0) {
double nm = n[mat];
p[(i+sizex*j)*Nmats+mat] = (nm * rho[(i+sizex*j)*Nmats+mat] * t[(i+sizex*j)*Nmats+mat]) / Vf[(i+sizex*j)*Nmats+mat];
}
else {
p[(i+sizex*j)*Nmats+mat] = 0.0;
}
}
}
__global__ void mc_loop2(const double * __restrict rho, const double * __restrict Vf, const double * __restrict t, const double * __restrict n, double * __restrict p, int sizex, int sizey, int Nmats) {
int i = threadIdx.x + blockIdx.x * blockDim.x;
int j = threadIdx.y + blockIdx.y * blockDim.y;
if (i >= sizex || j >= sizey) return;
int ncells = sizex*sizey;
for (int mat = 0; mat < Nmats; mat++) {
double nm = n[mat];
if (Vf[ncells*mat + i+sizex*j] > 0.0) {
p[ncells*mat + i+sizex*j] = (nm * rho[ncells*mat + i+sizex*j] * t[ncells*mat + i+sizex*j]) / Vf[ncells*mat + i+sizex*j];
}
else {
p[ncells*mat + i+sizex*j] = 0.0;
}
}
}
__global__ void cc_loop3(const double * __restrict rho, double *__restrict rho_mat_ave, const double * __restrict Vf,
const double * __restrict x, const double * __restrict y, int sizex, int sizey, int Nmats) {
int i = threadIdx.x + blockIdx.x * blockDim.x;
int j = threadIdx.y + blockIdx.y * blockDim.y;
if (i >= sizex-1 || j >= sizey-1 || i < 1 || j < 1) return;
// o: outer
double xo = x[i+sizex*j];
double yo = y[i+sizex*j];
// There are at most 9 neighbours in 2D case.
double dsqr[9];
for (int nj = -1; nj <= 1; nj++) {
for (int ni = -1; ni <= 1; ni++) {
dsqr[(nj+1)*3 + (ni+1)] = 0.0;
// i: inner
double xi = x[(i+ni)+sizex*(j+nj)];
double yi = y[(i+ni)+sizex*(j+nj)];
dsqr[(nj+1)*3 + (ni+1)] += (xo - xi) * (xo - xi);
dsqr[(nj+1)*3 + (ni+1)] += (yo - yi) * (yo - yi);
}
}
for (int mat = 0; mat < Nmats; mat++) {
if (Vf[(i+sizex*j)*Nmats+mat] > 0.0) {
double rho_sum = 0.0;
int Nn = 0;
for (int nj = -1; nj <= 1; nj++) {
for (int ni = -1; ni <= 1; ni++) {
if (Vf[((i+ni)+sizex*(j+nj))*Nmats+mat] > 0.0) {
rho_sum += rho[((i+ni)+sizex*(j+nj))*Nmats+mat] / dsqr[(nj+1)*3 + (ni+1)];
Nn += 1;
}
}
}
rho_mat_ave[(i+sizex*j)*Nmats+mat] = rho_sum / Nn;
}
else {
rho_mat_ave[(i+sizex*j)*Nmats+mat] = 0.0;
}
}
}
__global__ void mc_loop3(const double * __restrict rho, double * __restrict rho_mat_ave, const double * __restrict Vf,
const double * __restrict x, const double * __restrict y, int sizex, int sizey, int Nmats) {
int i = threadIdx.x + blockIdx.x * blockDim.x;
int j = threadIdx.y + blockIdx.y * blockDim.y;
if (i >= sizex-1 || j >= sizey-1 || i < 1 || j < 1) return;
int ncells = sizex*sizey;
for (int mat = 0; mat < Nmats; mat++) {
if (Vf[ncells*mat + i+sizex*j] > 0.0) {
// o: outer
double xo = x[i+sizex*j];
double yo = y[i+sizex*j];
double rho_sum = 0.0;
int Nn = 0;
for (int nj = -1; nj <= 1; nj++) {
for (int ni = -1; ni <= 1; ni++) {
if (Vf[ncells*mat + (i+ni)+sizex*(j+nj)] > 0.0) {
double dsqr = 0.0;
// i: inner
double xi = x[(i+ni)+sizex*(j+nj)];
double yi = y[(i+ni)+sizex*(j+nj)];
dsqr += (xo - xi) * (xo - xi);
dsqr += (yo - yi) * (yo - yi);
rho_sum += rho[ncells*mat + i+sizex*j] / dsqr;
Nn += 1;
}
}
}
rho_mat_ave[ncells*mat + i+sizex*j] = rho_sum / Nn;
}
else {
rho_mat_ave[ncells*mat + i+sizex*j] = 0.0;
}
}
}
void full_matrix_cell_centric(full_data cc)
{
int sizex = cc.sizex;
int sizey = cc.sizey;
int Nmats = cc.Nmats;
double *d_rho = (double *)cp_to_device((char*)cc.rho, sizex*sizey*Nmats*sizeof(double));
double *d_rho_mat_ave = (double *)cp_to_device((char*)cc.rho_mat_ave, sizex*sizey*Nmats*sizeof(double));
double *d_p = (double *)cp_to_device((char*)cc.p, sizex*sizey*Nmats*sizeof(double));
double *d_t = (double *)cp_to_device((char*)cc.t, sizex*sizey*Nmats*sizeof(double));
double *d_Vf = (double *)cp_to_device((char*)cc.Vf, sizex*sizey*Nmats*sizeof(double));
double *d_V = (double *)cp_to_device((char*)cc.V, sizex*sizey*sizeof(double));
double *d_x = (double *)cp_to_device((char*)cc.x, sizex*sizey*sizeof(double));
double *d_y = (double *)cp_to_device((char*)cc.y, sizex*sizey*sizeof(double));
double *d_n = (double *)cp_to_device((char*)cc.n, Nmats*sizeof(double));
double *d_rho_ave = (double *)cp_to_device((char*)cc.rho_ave, sizex*sizey*sizeof(double));
int thx = 32;
int thy = 4;
dim3 threads(thx,thy,1);
dim3 blocks((sizex-1)/thx+1, (sizey-1)/thy+1, 1);
// Cell-centric algorithms
// Computational loop 1 - average density in cell
double t1 = omp_get_wtime();
cc_loop1<<<blocks, threads>>>(d_rho, d_Vf, d_V, d_rho_ave, sizex, sizey, Nmats);
cudaDeviceSynchronize();
printf("Full matrix, cell centric, alg 1: %g sec\n", omp_get_wtime()-t1);
// Computational loop 2 - Pressure for each cell and each material
t1 = omp_get_wtime();
cc_loop2<<<blocks, threads>>>(d_rho, d_Vf, d_t, d_n, d_p, sizex, sizey, Nmats);
cudaDeviceSynchronize();
printf("Full matrix, cell centric, alg 2: %g sec\n", omp_get_wtime()-t1);
// Computational loop 3 - Average density of each material over neighborhood of each cell
t1 = omp_get_wtime();
cc_loop3<<<blocks, threads>>>(d_rho, d_rho_mat_ave, d_Vf, d_x, d_y, sizex, sizey, Nmats);
cudaDeviceSynchronize();
printf("Full matrix, cell centric, alg 3: %g sec\n", omp_get_wtime()-t1);
cp_to_host((char*)cc.rho, (char*)d_rho, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)cc.rho_mat_ave, (char*)d_rho_mat_ave, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)cc.p, (char*)d_p, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)cc.t, (char*)d_t, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)cc.Vf, (char*)d_Vf, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)cc.V, (char*)d_V, sizex*sizey*sizeof(double));
cp_to_host((char*)cc.x, (char*)d_x, sizex*sizey*sizeof(double));
cp_to_host((char*)cc.y, (char*)d_y, sizex*sizey*sizeof(double));
cp_to_host((char*)cc.n, (char*)d_n, Nmats*sizeof(double));
cp_to_host((char*)cc.rho_ave, (char*)d_rho_ave, sizex*sizey*sizeof(double));
}
void full_matrix_material_centric(full_data cc, full_data mc)
{
int sizex = cc.sizex;
int sizey = cc.sizey;
int Nmats = cc.Nmats;
double *d_rho = (double *)cp_to_device((char*)mc.rho, sizex*sizey*Nmats*sizeof(double));
double *d_p = (double *)cp_to_device((char*)mc.p, sizex*sizey*Nmats*sizeof(double));
double *d_t = (double *)cp_to_device((char*)mc.t, sizex*sizey*Nmats*sizeof(double));
double *d_Vf = (double *)cp_to_device((char*)mc.Vf, sizex*sizey*Nmats*sizeof(double));
double *d_V = (double *)cp_to_device((char*)mc.V, sizex*sizey*sizeof(double));
double *d_x = (double *)cp_to_device((char*)mc.x, sizex*sizey*sizeof(double));
double *d_y = (double *)cp_to_device((char*)mc.y, sizex*sizey*sizeof(double));
double *d_n = (double *)cp_to_device((char*)mc.n, Nmats*sizeof(double));
double *d_rho_ave = (double *)cp_to_device((char*)mc.rho_ave, sizex*sizey*sizeof(double));
double *d_rho_mat_ave = (double *)cp_to_device((char*)mc.rho_mat_ave, sizex*sizey*Nmats*sizeof(double));
int thx = 32;
int thy = 4;
dim3 threads(thx,thy,1);
dim3 blocks((sizex-1)/thx+1, (sizey-1)/thy+1, 1);
// Material-centric algorithms
// Computational loop 1 - average density in cell
double t1 = omp_get_wtime();
mc_loop1<<<blocks, threads>>>(d_rho, d_Vf, d_V, d_rho_ave, sizex, sizey, Nmats);
cudaDeviceSynchronize();
printf("Full matrix, material centric, alg 1: %g sec\n", omp_get_wtime()-t1);
// Computational loop 2 - Pressure for each cell and each material
t1 = omp_get_wtime();
mc_loop2<<<blocks, threads>>>(d_rho, d_Vf, d_t, d_n, d_p, sizex, sizey, Nmats);
cudaDeviceSynchronize();
printf("Full matrix, material centric, alg 2: %g sec\n", omp_get_wtime()-t1);
// Computational loop 3 - Average density of each material over neighborhood of each cell
t1 = omp_get_wtime();
mc_loop3<<<blocks, threads>>>(d_rho, d_rho_mat_ave, d_Vf, d_x, d_y, sizex, sizey, Nmats);
cudaDeviceSynchronize();
printf("Full matrix, material centric, alg 2: %g sec\n", omp_get_wtime()-t1);
cp_to_host((char*)mc.rho, (char*)d_rho, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)mc.p, (char*)d_p, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)mc.t, (char*)d_t, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)mc.Vf, (char*)d_Vf, sizex*sizey*Nmats*sizeof(double));
cp_to_host((char*)mc.V, (char*)d_V, sizex*sizey*sizeof(double));
cp_to_host((char*)mc.x, (char*)d_x, sizex*sizey*sizeof(double));
cp_to_host((char*)mc.y, (char*)d_y, sizex*sizey*sizeof(double));
cp_to_host((char*)mc.n, (char*)d_n, Nmats*sizeof(double));
cp_to_host((char*)mc.rho_ave, (char*)d_rho_ave, sizex*sizey*sizeof(double));
cp_to_host((char*)mc.rho_mat_ave, (char*)d_rho_mat_ave, sizex*sizey*Nmats*sizeof(double));
}
bool full_matrix_check_results(full_data cc, full_data mc)
{
int sizex = cc.sizex;
int sizey = cc.sizey;
int Nmats = cc.Nmats;
int ncells = sizex * sizey;
printf("Checking results of full matrix representation... ");
for (int j = 0; j < sizey; j++) {
for (int i = 0; i < sizex; i++) {
if (fabs(cc.rho_ave[i+sizex*j] - mc.rho_ave[i+sizex*j]) > 0.0001) {
printf("1. cell-centric and material-centric values are not equal! (%f, %f, %d, %d)\n",
cc.rho_ave[i+sizex*j], mc.rho_ave[i+sizex*j], i, j);
return false;
}
for (int mat = 0; mat < Nmats; mat++) {
if (fabs(cc.p[(i+sizex*j)*Nmats+mat] - mc.p[ncells*mat + i+sizex*j]) > 0.0001) {
printf("2. cell-centric and material-centric values are not equal! (%f, %f, %d, %d, %d)\n",
cc.p[(i+sizex*j)*Nmats+mat], mc.p[ncells*mat + i+sizex*j], i, j, mat);
return false;
}
if (fabs(cc.rho_mat_ave[(i+sizex*j)*Nmats+mat] - mc.rho_mat_ave[ncells*mat + i+sizex*j]) > 0.0001) {
printf("3. cell-centric and material-centric values are not equal! (%f, %f, %d, %d, %d)\n",
cc.rho_mat_ave[(i+sizex*j)*Nmats+mat], mc.rho_mat_ave[ncells*mat + i+sizex*j], i, j, mat);
return false;
}
}
}
}
printf("All tests passed!\n");
return true;
}