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
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
|
/*
-- MAGMA (version 2.9.0) --
Univ. of Tennessee, Knoxville
Univ. of California, Berkeley
Univ. of Colorado, Denver
@date January 2025
@author Hartwig Anzt
@precisions normal z -> s d c
*/
#include "magmasparse_internal.h"
#include "magma_trisolve.h"
#define PRECISION_z
// This file is deprecated and will be removed in future.
// The ParILU/ParIC functionality is provided by
// src/zparilu_gpu.cpp and src/zparic_gpu.cpp
//
/**
Purpose
-------
Prepares the ILU preconditioner via the iterative ILU iteration.
Arguments
---------
@param[in]
A magma_z_matrix
input matrix A
@param[in]
b magma_z_matrix
input RHS b
@param[in,out]
precond magma_z_preconditioner*
preconditioner parameters
@param[in]
queue magma_queue_t
Queue to execute in.
@ingroup magmasparse_zgepr
********************************************************************/
extern "C"
magma_int_t
magma_zparilusetup(
magma_z_matrix A,
magma_z_matrix b,
magma_z_preconditioner *precond,
magma_queue_t queue )
{
magma_int_t info = 0;
magma_z_matrix hAh={Magma_CSR}, hA={Magma_CSR}, hL={Magma_CSR}, hU={Magma_CSR},
hAcopy={Magma_CSR}, hAL={Magma_CSR}, hAU={Magma_CSR}, hAUt={Magma_CSR},
hUT={Magma_CSR}, hAtmp={Magma_CSR}, hACSRCOO={Magma_CSR}, dAinitguess={Magma_CSR},
dL={Magma_CSR}, dU={Magma_CSR};
// copy original matrix as CSRCOO to device
CHECK( magma_zmtransfer(A, &hAh, A.memory_location, Magma_CPU, queue ));
CHECK( magma_zmconvert( hAh, &hA, hAh.storage_type, Magma_CSR , queue ));
magma_zmfree(&hAh, queue );
CHECK( magma_zmtransfer( hA, &hAcopy, Magma_CPU, Magma_CPU , queue ));
// in case using fill-in
CHECK( magma_zsymbilu( &hAcopy, precond->levels, &hAL, &hAUt, queue ));
// add a unit diagonal to L for the algorithm
CHECK( magma_zmLdiagadd( &hAL , queue ));
// transpose U for the algorithm
CHECK( magma_z_cucsrtranspose( hAUt, &hAU , queue ));
magma_zmfree( &hAUt , queue );
// ---------------- initial guess ------------------- //
CHECK( magma_zmconvert( hAcopy, &hACSRCOO, Magma_CSR, Magma_CSRCOO , queue ));
CHECK( magma_zmtransfer( hACSRCOO, &dAinitguess, Magma_CPU, Magma_DEV , queue ));
magma_zmfree(&hACSRCOO, queue );
magma_zmfree(&hAcopy, queue );
// transfer the factor L and U
CHECK( magma_zmtransfer( hAL, &dL, Magma_CPU, Magma_DEV , queue ));
CHECK( magma_zmtransfer( hAU, &dU, Magma_CPU, Magma_DEV , queue ));
magma_zmfree(&hAL, queue );
magma_zmfree(&hAU, queue );
for(int i=0; i<precond->sweeps; i++){
CHECK( magma_zparilu_csr( dAinitguess, dL, dU , queue ));
}
CHECK( magma_zmtransfer( dL, &hL, Magma_DEV, Magma_CPU , queue ));
CHECK( magma_zmtransfer( dU, &hU, Magma_DEV, Magma_CPU , queue ));
CHECK( magma_z_cucsrtranspose( hU, &hUT , queue ));
magma_zmfree(&dL, queue );
magma_zmfree(&dU, queue );
magma_zmfree(&hU, queue );
CHECK( magma_zmlumerge( hL, hUT, &hAtmp, queue ));
magma_zmfree(&hL, queue );
magma_zmfree(&hUT, queue );
CHECK( magma_zmtransfer( hAtmp, &precond->M, Magma_CPU, Magma_DEV , queue ));
hAL.diagorder_type = Magma_UNITY;
CHECK( magma_zmconvert(hAtmp, &hAL, Magma_CSR, Magma_CSRL, queue ));
hAL.storage_type = Magma_CSR;
CHECK( magma_zmconvert(hAtmp, &hAU, Magma_CSR, Magma_CSRU, queue ));
hAU.storage_type = Magma_CSR;
magma_zmfree(&hAtmp, queue );
// CHECK( magma_zcsrsplit( 0, 256, hAL, &DL, &RL , queue ));
// CHECK( magma_zcsrsplit( 0, 256, hAU, &DU, &RU , queue ));
//
// CHECK( magma_zmtransfer( DL, &precond->LD, Magma_CPU, Magma_DEV , queue ));
// CHECK( magma_zmtransfer( DU, &precond->UD, Magma_CPU, Magma_DEV , queue ));
// for cusparse uncomment this
CHECK( magma_zmtransfer( hAL, &precond->L, Magma_CPU, Magma_DEV , queue ));
CHECK( magma_zmtransfer( hAU, &precond->U, Magma_CPU, Magma_DEV , queue ));
/*
//-- for ba-solve uncomment this
if( RL.nnz != 0 )
CHECK( magma_zmtransfer( RL, &precond->L, Magma_CPU, Magma_DEV , queue ));
else {
precond->L.nnz = 0;
precond->L.val = NULL;
precond->L.col = NULL;
precond->L.row = NULL;
precond->L.blockinfo = NULL;
}
if( RU.nnz != 0 )
CHECK( magma_zmtransfer( RU, &precond->U, Magma_CPU, Magma_DEV , queue ));
else {
precond->U.nnz = 0;
precond->L.val = NULL;
precond->L.col = NULL;
precond->L.row = NULL;
precond->L.blockinfo = NULL;
}
//-- for ba-solve uncomment this
*/
// extract the diagonal of L into precond->d
CHECK( magma_zjacobisetup_diagscal( precond->L, &precond->d, queue ));
CHECK( magma_zvinit( &precond->work1, Magma_DEV, hA.num_rows, 1, MAGMA_Z_ZERO, queue ));
// extract the diagonal of U into precond->d2
CHECK( magma_zjacobisetup_diagscal( precond->U, &precond->d2, queue ));
CHECK( magma_zvinit( &precond->work2, Magma_DEV, hA.num_rows, 1, MAGMA_Z_ZERO, queue ));
magma_zmfree(&hAL, queue );
magma_zmfree(&hAU, queue );
// magma_zmfree(&DL, queue );
// magma_zmfree(&RL, queue );
// magma_zmfree(&DU, queue );
// magma_zmfree(&RU, queue );
CHECK(magma_ztrisolve_analysis(precond->L, &precond->cuinfoL, false, false, false, queue));
CHECK(magma_ztrisolve_analysis(precond->U, &precond->cuinfoU, true, false, false, queue));
cleanup:
magma_zmfree( &hAh, queue );
magma_zmfree( &hA, queue );
magma_zmfree( &hL, queue );
magma_zmfree( &hU, queue );
magma_zmfree( &hAcopy, queue );
magma_zmfree( &hAL, queue );
magma_zmfree( &hAU, queue );
magma_zmfree( &hAUt, queue );
magma_zmfree( &hUT, queue );
magma_zmfree( &hAtmp, queue );
magma_zmfree( &hACSRCOO, queue );
magma_zmfree( &dAinitguess, queue );
magma_zmfree( &dL, queue );
magma_zmfree( &dU, queue );
// magma_zmfree( &DL, queue );
// magma_zmfree( &DU, queue );
// magma_zmfree( &RL, queue );
// magma_zmfree( &RU, queue );
return info;
}
/**
Purpose
-------
Updates an existing preconditioner via additional iterative ILU sweeps for
previous factorization initial guess (PFIG).
See Anzt et al., Parallel Computing, 2015.
Arguments
---------
@param[in]
A magma_z_matrix
input matrix A, current target system
@param[in]
precond magma_z_preconditioner*
preconditioner parameters
@param[in]
updates magma_int_t
number of updates
@param[in]
queue magma_queue_t
Queue to execute in.
@ingroup magmasparse_zhepr
********************************************************************/
extern "C"
magma_int_t
magma_zpariluupdate(
magma_z_matrix A,
magma_z_preconditioner *precond,
magma_int_t updates,
magma_queue_t queue )
{
magma_int_t info = 0;
magma_z_matrix hALt={Magma_CSR};
magma_z_matrix d_h={Magma_CSR};
magma_z_matrix hL={Magma_CSR}, hU={Magma_CSR},
hAcopy={Magma_CSR}, hAL={Magma_CSR}, hAU={Magma_CSR}, hAUt={Magma_CSR},
hUT={Magma_CSR}, hAtmp={Magma_CSR},
dL={Magma_CSR}, dU={Magma_CSR};
if ( updates > 0 ){
CHECK( magma_zmtransfer( precond->M, &hAcopy, Magma_DEV, Magma_CPU , queue ));
// in case using fill-in
CHECK( magma_zsymbilu( &hAcopy, precond->levels, &hAL, &hAUt, queue ));
// add a unit diagonal to L for the algorithm
CHECK( magma_zmLdiagadd( &hAL , queue ));
// transpose U for the algorithm
CHECK( magma_z_cucsrtranspose( hAUt, &hAU , queue ));
// transfer the factor L and U
CHECK( magma_zmtransfer( hAL, &dL, Magma_CPU, Magma_DEV , queue ));
CHECK( magma_zmtransfer( hAU, &dU, Magma_CPU, Magma_DEV , queue ));
magma_zmfree(&hAL, queue );
magma_zmfree(&hAU, queue );
magma_zmfree(&hAUt, queue );
magma_zmfree(&precond->M, queue );
magma_zmfree(&hAcopy, queue );
// copy original matrix as CSRCOO to device
for(int i=0; i<updates; i++){
CHECK( magma_zparilu_csr( A, dL, dU, queue ));
}
CHECK( magma_zmtransfer( dL, &hL, Magma_DEV, Magma_CPU , queue ));
CHECK( magma_zmtransfer( dU, &hU, Magma_DEV, Magma_CPU , queue ));
CHECK( magma_z_cucsrtranspose( hU, &hUT , queue ));
magma_zmfree(&dL, queue );
magma_zmfree(&dU, queue );
magma_zmfree(&hU, queue );
CHECK( magma_zmlumerge( hL, hUT, &hAtmp, queue ));
// for CUSPARSE
CHECK( magma_zmtransfer( hAtmp, &precond->M, Magma_CPU, Magma_DEV , queue ));
magma_zmfree(&hL, queue );
magma_zmfree(&hUT, queue );
hAL.diagorder_type = Magma_UNITY;
CHECK( magma_zmconvert(hAtmp, &hAL, Magma_CSR, Magma_CSRL, queue ));
hAL.storage_type = Magma_CSR;
CHECK( magma_zmconvert(hAtmp, &hAU, Magma_CSR, Magma_CSRU, queue ));
hAU.storage_type = Magma_CSR;
magma_zmfree(&hAtmp, queue );
CHECK( magma_zmtransfer( hAL, &precond->L, Magma_CPU, Magma_DEV , queue ));
CHECK( magma_zmtransfer( hAU, &precond->U, Magma_CPU, Magma_DEV , queue ));
magma_zmfree(&hAL, queue );
magma_zmfree(&hAU, queue );
magma_zmfree( &precond->d , queue );
magma_zmfree( &precond->d2 , queue );
CHECK( magma_zjacobisetup_diagscal( precond->L, &precond->d, queue ));
CHECK( magma_zjacobisetup_diagscal( precond->U, &precond->d2, queue ));
}
cleanup:
magma_zmfree(&d_h, queue );
magma_zmfree(&hALt, queue );
return info;
}
/**
Purpose
-------
Prepares the IC preconditioner via the iterative IC iteration.
Arguments
---------
@param[in]
A magma_z_matrix
input matrix A
@param[in]
b magma_z_matrix
input RHS b
@param[in,out]
precond magma_z_preconditioner*
preconditioner parameters
@param[in]
queue magma_queue_t
Queue to execute in.
@ingroup magmasparse_zhepr
********************************************************************/
extern "C"
magma_int_t
magma_zparicsetup(
magma_z_matrix A,
magma_z_matrix b,
magma_z_preconditioner *precond,
magma_queue_t queue )
{
magma_int_t info = 0;
cusparseHandle_t cusparseHandle=NULL;
cusparseMatDescr_t descrL=NULL;
cusparseMatDescr_t descrU=NULL;
magma_z_matrix hAh={Magma_CSR}, hA={Magma_CSR}, hAtmp={Magma_CSR},
hAL={Magma_CSR}, hAUt={Magma_CSR}, hALt={Magma_CSR}, hM={Magma_CSR},
hACSRCOO={Magma_CSR}, dAinitguess={Magma_CSR}, dL={Magma_CSR};
magma_z_matrix d_h={Magma_CSR};
// copy original matrix as CSRCOO to device
CHECK( magma_zmtransfer(A, &hAh, A.memory_location, Magma_CPU, queue ));
CHECK( magma_zmconvert( hAh, &hA, hAh.storage_type, Magma_CSR , queue ));
magma_zmfree(&hAh, queue );
// in case using fill-in
CHECK( magma_zsymbilu( &hA, precond->levels, &hAL, &hAUt , queue ));
// need only lower triangular
magma_zmfree(&hAUt, queue );
magma_zmfree(&hAL, queue );
CHECK( magma_zmconvert( hA, &hAtmp, Magma_CSR, Magma_CSRL , queue ));
magma_zmfree(&hA, queue );
// ---------------- initial guess ------------------- //
CHECK( magma_zmconvert( hAtmp, &hACSRCOO, Magma_CSR, Magma_CSRCOO , queue ));
//int blocksize = 1;
//magma_zmreorder( hACSRCOO, n, blocksize, blocksize, blocksize, &hAinitguess , queue );
CHECK( magma_zmtransfer( hACSRCOO, &dAinitguess, Magma_CPU, Magma_DEV , queue ));
magma_zmfree(&hACSRCOO, queue );
CHECK( magma_zmtransfer( hAtmp, &dL, Magma_CPU, Magma_DEV , queue ));
magma_zmfree(&hAtmp, queue );
for(int i=0; i<precond->sweeps; i++){
CHECK( magma_zparic_csr( dAinitguess, dL , queue ));
}
CHECK( magma_zmtransfer( dL, &hAL, Magma_DEV, Magma_CPU , queue ));
magma_zmfree(&dL, queue );
magma_zmfree(&dAinitguess, queue );
// for CUSPARSE
CHECK( magma_zmtransfer( hAL, &precond->M, Magma_CPU, Magma_DEV , queue ));
// Jacobi setup
CHECK( magma_zjacobisetup_matrix( precond->M, &precond->L, &precond->d , queue ));
// for Jacobi, we also need U
CHECK( magma_z_cucsrtranspose( hAL, &hALt , queue ));
CHECK( magma_zjacobisetup_matrix( hALt, &hM, &d_h , queue ));
CHECK( magma_zmtransfer( hM, &precond->U, Magma_CPU, Magma_DEV , queue ));
magma_zmfree(&hM, queue );
magma_zmfree(&d_h, queue );
// copy the matrix to precond->L and (transposed) to precond->U
CHECK( magma_zmtransfer(precond->M, &(precond->L), Magma_DEV, Magma_DEV, queue ));
CHECK( magma_zmtranspose( precond->L, &(precond->U), queue ));
// extract the diagonal of L into precond->d
CHECK( magma_zjacobisetup_diagscal( precond->L, &precond->d, queue ));
CHECK( magma_zvinit( &precond->work1, Magma_DEV, hAL.num_rows, 1, MAGMA_Z_ZERO, queue ));
// extract the diagonal of U into precond->d2
CHECK( magma_zjacobisetup_diagscal( precond->U, &precond->d2, queue ));
CHECK( magma_zvinit( &precond->work2, Magma_DEV, hAL.num_rows, 1, MAGMA_Z_ZERO, queue ));
magma_zmfree(&hAL, queue );
magma_zmfree(&hALt, queue );
CHECK(magma_ztrisolve_analysis(precond->M, &precond->cuinfoL, false, false, false, queue));
CHECK(magma_ztrisolve_analysis(precond->M, &precond->cuinfoU, false, false, true, queue));
cleanup:
cusparseDestroy( cusparseHandle );
cusparseDestroyMatDescr( descrL );
cusparseDestroyMatDescr( descrU );
cusparseHandle=NULL;
descrL=NULL;
descrU=NULL;
magma_zmfree( &hAh, queue );
magma_zmfree( &hA, queue );
magma_zmfree( &hAtmp, queue );
magma_zmfree( &hAL, queue );
magma_zmfree( &hAUt, queue );
magma_zmfree( &hALt, queue );
magma_zmfree( &hM, queue );
magma_zmfree( &hACSRCOO, queue );
magma_zmfree( &dAinitguess, queue );
magma_zmfree( &dL, queue );
magma_zmfree( &d_h, queue );
return info;
}
/**
Purpose
-------
Updates an existing preconditioner via additional iterative IC sweeps for
previous factorization initial guess (PFIG).
See Anzt et al., Parallel Computing, 2015.
Arguments
---------
@param[in]
A magma_z_matrix
input matrix A, current target system
@param[in]
precond magma_z_preconditioner*
preconditioner parameters
@param[in]
updates magma_int_t
number of updates
@param[in]
queue magma_queue_t
Queue to execute in.
@ingroup magmasparse_zhepr
********************************************************************/
extern "C"
magma_int_t
magma_zparicupdate(
magma_z_matrix A,
magma_z_preconditioner *precond,
magma_int_t updates,
magma_queue_t queue )
{
magma_int_t info = 0;
magma_z_matrix hALt={Magma_CSR};
magma_z_matrix d_h={Magma_CSR};
if( updates > 0 ){
// copy original matrix as CSRCOO to device
for(int i=0; i<updates; i++){
CHECK( magma_zparic_csr( A, precond->M , queue ));
}
//magma_zmtransfer( precond->M, &hALt, Magma_DEV, Magma_CPU , queue );
magma_zmfree(&precond->L, queue );
magma_zmfree(&precond->U, queue );
magma_zmfree( &precond->d , queue );
magma_zmfree( &precond->d2 , queue );
// copy the matrix to precond->L and (transposed) to precond->U
CHECK( magma_zmtransfer(precond->M, &(precond->L), Magma_DEV, Magma_DEV, queue ));
CHECK( magma_zmtranspose( precond->L, &(precond->U), queue ));
CHECK( magma_zjacobisetup_diagscal( precond->L, &precond->d, queue ));
CHECK( magma_zjacobisetup_diagscal( precond->U, &precond->d2, queue ));
}
cleanup:
magma_zmfree(&d_h, queue );
magma_zmfree(&hALt, queue );
return info;
}
|
