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
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
|
/*
-- MAGMA (version 2.9.0) --
Univ. of Tennessee, Knoxville
Univ. of California, Berkeley
Univ. of Colorado, Denver
@date January 2025
@generated from sparse/control/magma_zparilut_kernels.cpp, normal z -> c, Wed Jan 22 14:42:33 2025
@author Hartwig Anzt
*/
#include "magmasparse_internal.h"
#ifdef _OPENMP
#include <omp.h>
#endif
#define SWAP(a, b) { val_swap = a; a = b; b = val_swap; }
/***************************************************************************//**
Purpose
-------
This function does an ParILUT sweep. The difference to the ParILU sweep is
that the nonzero pattern of A and the incomplete factors L and U can be
different.
The pattern determing which elements are iterated are hence the pattern
of L and U, not A.
This is the CPU version of the asynchronous ParILUT sweep.
Arguments
---------
@param[in]
A magma_c_matrix*
System matrix. The format is sorted CSR.
@param[in,out]
L magma_c_matrix*
Current approximation for the lower triangular factor
The format is MAGMA_CSRCOO. This is sorted CSR plus the
rowindexes being stored.
@param[in,out]
U magma_c_matrix*
Current approximation for the lower triangular factor
The format is MAGMA_CSRCOO. This is sorted CSR plus the
rowindexes being stored.
@param[in]
queue magma_queue_t
Queue to execute in.
@ingroup magmasparse_caux
*******************************************************************************/
extern "C" magma_int_t
magma_cparilut_sweep(
magma_c_matrix *A,
magma_c_matrix *L,
magma_c_matrix *U,
magma_queue_t queue)
{
magma_int_t info = 0;
#pragma omp parallel for
for (magma_int_t e=0; e<L->nnz; e++) {
magma_int_t i,j,icol,jcol,jold;
magma_index_t row = L->rowidx[ e ];
magma_index_t col = L->col[ e ];
// as we look at the lower triangular,
// col<row, i.e. disregard last element in row
if(col < row) {
magmaFloatComplex A_e = MAGMA_C_ZERO;
// check whether A contains element in this location
for (i = A->row[row]; i<A->row[row+1]; i++) {
if(A->col[i] == col) {
A_e = A->val[i];
break;
}
}
//now do the actual iteration
i = L->row[ row ];
j = U->row[ col ];
magma_int_t endi = L->row[ row+1 ];
magma_int_t endj = U->row[ col+1 ];
magmaFloatComplex sum = MAGMA_C_ZERO;
magmaFloatComplex lsum = MAGMA_C_ZERO;
do{
lsum = MAGMA_C_ZERO;
jold = j;
icol = L->col[i];
jcol = U->col[j];
if(icol == jcol) {
lsum = L->val[i] * U->val[j];
sum = sum + lsum;
i++;
j++;
}
else if(icol<jcol) {
i++;
}
else {
j++;
}
}while(i<endi && j<endj);
sum = sum - lsum;
// write back to location e
L->val[ e ] = (A_e - sum)/ U->val[jold];
} else if(row == col) { // end check whether part of L
L->val[ e ] = MAGMA_C_ONE; // lower triangular has diagonal equal 1
}
}// end omp parallel section
#pragma omp parallel for
for (magma_int_t e=0; e<U->nnz; e++) {
{
magma_int_t i,j,icol,jcol;
magma_index_t row = U->col[ e ];
magma_index_t col = U->rowidx[ e ];
magmaFloatComplex A_e = MAGMA_C_ZERO;
// check whether A contains element in this location
for (i = A->row[row]; i<A->row[row+1]; i++) {
if(A->col[i] == col) {
A_e = A->val[i];
break;
}
}
//now do the actual iteration
i = L->row[ row ];
j = U->row[ col ];
magma_int_t endi = L->row[ row+1 ];
magma_int_t endj = U->row[ col+1 ];
magmaFloatComplex sum = MAGMA_C_ZERO;
magmaFloatComplex lsum = MAGMA_C_ZERO;
do{
lsum = MAGMA_C_ZERO;
icol = L->col[i];
jcol = U->col[j];
if(icol == jcol) {
lsum = L->val[i] * U->val[j];
sum = sum + lsum;
i++;
j++;
}
else if(icol<jcol) {
i++;
}
else {
j++;
}
}while(i<endi && j<endj);
sum = sum - lsum;
// write back to location e
U->val[ e ] = (A_e - sum);
}
}// end omp parallel section
return info;
}
/***************************************************************************//**
Purpose
-------
This function does an ParILUT sweep. The difference to the ParILU sweep is
that the nonzero pattern of A and the incomplete factors L and U can be
different.
The pattern determing which elements are iterated are hence the pattern
of L and U, not A.
This is the CPU version of the synchronous ParILUT sweep.
Arguments
---------
@param[in]
A magma_c_matrix*
System matrix. The format is sorted CSR.
@param[in,out]
L magma_c_matrix*
Current approximation for the lower triangular factor
The format is MAGMA_CSRCOO. This is sorted CSR plus the
rowindexes being stored.
@param[in,out]
U magma_c_matrix*
Current approximation for the lower triangular factor
The format is MAGMA_CSRCOO. This is sorted CSR plus the
rowindexes being stored.
@param[in]
queue magma_queue_t
Queue to execute in.
@ingroup magmasparse_caux
*******************************************************************************/
extern "C" magma_int_t
magma_cparilut_sweep_sync(
magma_c_matrix *A,
magma_c_matrix *L,
magma_c_matrix *U,
magma_queue_t queue)
{
magma_int_t info = 0;
magmaFloatComplex *L_new_val = NULL, *U_new_val = NULL, *val_swap = NULL;
CHECK(magma_cmalloc_cpu(&L_new_val, L->nnz));
CHECK(magma_cmalloc_cpu(&U_new_val, U->nnz));
#pragma omp parallel for
for (magma_int_t e=0; e<U->nnz; e++) {
magma_int_t i,j,icol,jcol;
magma_index_t row = U->col[ e ];
magma_index_t col = U->rowidx[ e ];
{
magmaFloatComplex A_e = MAGMA_C_ZERO;
// check whether A contains element in this location
for (i = A->row[row]; i<A->row[row+1]; i++) {
if(A->col[i] == col) {
A_e = A->val[i];
break;
}
}
//now do the actual iteration
i = L->row[ row ];
j = U->row[ col ];
magma_int_t endi = L->row[ row+1 ];
magma_int_t endj = U->row[ col+1 ];
magmaFloatComplex sum = MAGMA_C_ZERO;
magmaFloatComplex lsum = MAGMA_C_ZERO;
do{
lsum = MAGMA_C_ZERO;
icol = L->col[i];
jcol = U->col[j];
if(icol == jcol) {
lsum = L->val[i] * U->val[j];
sum = sum + lsum;
i++;
j++;
}
else if(icol<jcol) {
i++;
}
else {
j++;
}
}while(i<endi && j<endj);
sum = sum - lsum;
// write back to location e
U_new_val[ e ] = (A_e - sum);
}
}// end omp parallel section
#pragma omp parallel for
for (magma_int_t e=0; e<L->nnz; e++) {
magma_int_t i,j,icol,jcol,jold;
magma_index_t row = L->rowidx[ e ];
magma_index_t col = L->col[ e ];
// as we look at the lower triangular,
// col<row, i.e. disregard last element in row
if(row == col) {
L_new_val[ e ] = MAGMA_C_ONE; // lower triangular has 1-diagonal
} else {
magmaFloatComplex A_e = MAGMA_C_ZERO;
// check whether A contains element in this location
for (i = A->row[row]; i<A->row[row+1]; i++) {
if(A->col[i] == col) {
A_e = A->val[i];
break;
}
}
//now do the actual iteration
i = L->row[ row ];
j = U->row[ col ];
magma_int_t endi = L->row[ row+1 ];
magma_int_t endj = U->row[ col+1 ];
magmaFloatComplex sum = MAGMA_C_ZERO;
magmaFloatComplex lsum = MAGMA_C_ZERO;
do{
lsum = MAGMA_C_ZERO;
jold = j;
icol = L->col[i];
jcol = U->col[j];
if(icol == jcol) {
lsum = L->val[i] * U_new_val[j];
sum = sum + lsum;
i++;
j++;
}
else if(icol<jcol) {
i++;
}
else {
j++;
}
}while(i<endi && j<endj);
sum = sum - lsum;
// write back to location e
L_new_val[ e ] = (A_e - sum)/ U->val[jold];
}
}// end omp parallel section
// swap old and new values
SWAP(L_new_val, L->val);
SWAP(U_new_val, U->val);
cleanup:
magma_free_cpu(L_new_val);
magma_free_cpu(U_new_val);
return info;
}
/***************************************************************************//**
Purpose
-------
This function computes the ILU residual in the locations included in the
sparsity pattern of R.
Arguments
---------
@param[in]
A magma_c_matrix
System matrix A.
@param[in]
L magma_c_matrix
Current approximation for the lower triangular factor.
The format is sorted CSR.
@param[in]
U magma_c_matrix
Current approximation for the upper triangular factor.
The format is sorted CSR.
@param[in,out]
R magma_c_matrix*
Sparsity pattern on which the ILU residual is computed.
R is in COO format. On output, R contains the ILU residual.
@param[in]
queue magma_queue_t
Queue to execute in.
@ingroup magmasparse_caux
*******************************************************************************/
extern "C" magma_int_t
magma_cparilut_residuals(
magma_c_matrix A,
magma_c_matrix L,
magma_c_matrix U,
magma_c_matrix *R,
magma_queue_t queue)
{
magma_int_t info = 0;
#pragma omp parallel for
for (magma_int_t e=0; e<R->nnz; e++) {
{
magma_int_t i,j,icol,jcol;
magma_index_t row = R->rowidx[ e ];
magma_index_t col = R->col[ e ];
magmaFloatComplex A_e = MAGMA_C_ZERO;
for (i = A.row[row]; i<A.row[row+1]; i++) {
if(A.col[i] == col) {
A_e = A.val[i];
i = A.row[row+1];
}
}
//now do the actual iteration
i = L.row[ row ];
j = U.row[ col ];
magma_int_t endi = L.row[ row+1 ];
magma_int_t endj = U.row[ col+1 ];
magmaFloatComplex sum = MAGMA_C_ZERO;
magmaFloatComplex lsum = MAGMA_C_ZERO;
do{
lsum = MAGMA_C_ZERO;
icol = L.col[i];
jcol = U.col[j];
if(icol == jcol) {
lsum = L.val[i] * U.val[j];
sum = sum + lsum;
i++;
}
else if(icol<jcol) {
i++;
}
else {
j++;
}
}while(i<endi && j<endj);
sum = sum - lsum;
// write back to location e
R->val[ e ] = (A_e - sum);
}
}// end omp parallel section
return info;
}
|
