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
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
|
//------------------------------------------------------------------------------
// GB_AxB_saxpy3_template: C=A*B, C<M>=A*B, or C<!M>=A*B via saxpy3 method
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2022, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
// GB_AxB_saxpy3_template.c computes C=A*B for any semiring and matrix types,
// where C is sparse or hypersparse.
#include "GB_unused.h"
//------------------------------------------------------------------------------
// template code for C=A*B via the saxpy3 method
//------------------------------------------------------------------------------
{
#ifdef GB_TIMING
double ttt = omp_get_wtime ( ) ;
#endif
//--------------------------------------------------------------------------
// get the chunk size
//--------------------------------------------------------------------------
GB_GET_NTHREADS_MAX (nthreads_max, chunk, Context) ;
//--------------------------------------------------------------------------
// get M, A, B, and C
//--------------------------------------------------------------------------
int64_t *restrict Cp = C->p ;
// const int64_t *restrict Ch = C->h ;
const int64_t cvlen = C->vlen ;
const int64_t cnvec = C->nvec ;
const int64_t *restrict Bp = B->p ;
const int64_t *restrict Bh = B->h ;
const int8_t *restrict Bb = B->b ;
const int64_t *restrict Bi = B->i ;
const bool B_iso = B->iso ;
const int64_t bvlen = B->vlen ;
const bool B_is_sparse = GB_IS_SPARSE (B) ;
const bool B_is_hyper = GB_IS_HYPERSPARSE (B) ;
const bool B_is_bitmap = GB_IS_BITMAP (B) ;
const bool B_is_sparse_or_hyper = B_is_sparse || B_is_hyper ;
const int64_t *restrict Ap = A->p ;
const int64_t *restrict Ah = A->h ;
const int8_t *restrict Ab = A->b ;
const int64_t *restrict Ai = A->i ;
const int64_t anvec = A->nvec ;
const int64_t avlen = A->vlen ;
const bool A_is_sparse = GB_IS_SPARSE (A) ;
const bool A_is_hyper = GB_IS_HYPERSPARSE (A) ;
const bool A_is_bitmap = GB_IS_BITMAP (A) ;
const bool A_iso = A->iso ;
const bool A_jumbled = A->jumbled ;
const bool A_ok_for_binary_search =
((A_is_sparse || A_is_hyper) && !A_jumbled) ;
const int64_t *restrict A_Yp = NULL ;
const int64_t *restrict A_Yi = NULL ;
const int64_t *restrict A_Yx = NULL ;
int64_t A_hash_bits = 0 ;
if (A_is_hyper)
{
ASSERT_MATRIX_OK (A->Y, "A->Y hyper_hash", GB0) ;
A_Yp = A->Y->p ;
A_Yi = A->Y->i ;
A_Yx = A->Y->x ;
A_hash_bits = A->Y->vdim - 1 ;
}
#if ( !GB_NO_MASK )
const int64_t *restrict Mp = M->p ;
const int64_t *restrict Mh = M->h ;
const int8_t *restrict Mb = M->b ;
const int64_t *restrict Mi = M->i ;
const GB_void *restrict Mx = (GB_void *) (Mask_struct ? NULL : (M->x)) ;
const bool M_is_hyper = GB_IS_HYPERSPARSE (M) ;
const bool M_is_bitmap = GB_IS_BITMAP (M) ;
const bool M_jumbled = GB_JUMBLED (M) ;
size_t msize = M->type->size ;
int64_t mnvec = M->nvec ;
int64_t mvlen = M->vlen ;
// get the M hyper_hash
const int64_t *restrict M_Yp = NULL ;
const int64_t *restrict M_Yi = NULL ;
const int64_t *restrict M_Yx = NULL ;
int64_t M_hash_bits = 0 ;
{
if (M_is_hyper)
{
// mask is present, and hypersparse
M_Yp = M->Y->p ;
M_Yi = M->Y->i ;
M_Yx = M->Y->x ;
M_hash_bits = M->Y->vdim - 1 ;
}
}
#endif
#if !GB_A_IS_PATTERN
const GB_ATYPE *restrict Ax = (GB_ATYPE *) A->x ;
#endif
#if !GB_B_IS_PATTERN
const GB_BTYPE *restrict Bx = (GB_BTYPE *) B->x ;
#endif
//==========================================================================
// phase2: numeric work for fine tasks
//==========================================================================
// Coarse tasks: nothing to do in phase2.
// Fine tasks: compute nnz (C(:,j)), and values in Hx via atomics.
int taskid ;
#pragma omp parallel for num_threads(nthreads) schedule(dynamic,1)
for (taskid = 0 ; taskid < nfine ; taskid++)
{
//----------------------------------------------------------------------
// get the task descriptor
//----------------------------------------------------------------------
int64_t kk = SaxpyTasks [taskid].vector ;
int team_size = SaxpyTasks [taskid].team_size ;
int64_t hash_size = SaxpyTasks [taskid].hsize ;
bool use_Gustavson = (hash_size == cvlen) ;
int64_t pB = SaxpyTasks [taskid].start ;
int64_t pB_end = SaxpyTasks [taskid].end + 1 ;
int64_t j = GBH (Bh, kk) ;
GB_GET_T_FOR_SECONDJ ;
#if !GB_IS_ANY_PAIR_SEMIRING
GB_CTYPE *restrict Hx = (GB_CTYPE *) SaxpyTasks [taskid].Hx ;
#endif
#if GB_IS_PLUS_FC32_MONOID
float *restrict Hx_real = (float *) Hx ;
float *restrict Hx_imag = Hx_real + 1 ;
#elif GB_IS_PLUS_FC64_MONOID
double *restrict Hx_real = (double *) Hx ;
double *restrict Hx_imag = Hx_real + 1 ;
#endif
if (use_Gustavson)
{
//------------------------------------------------------------------
// phase2: fine Gustavson task
//------------------------------------------------------------------
// Hf [i] == 0: unlocked, i has not been seen in C(:,j).
// Hx [i] is not initialized.
// M(i,j) is 0, or M is not present.
// if M: Hf [i] stays equal to 0 (or 3 if locked)
// if !M, or no M: C(i,j) is a new entry seen for 1st time
// Hf [i] == 1: unlocked, i has not been seen in C(:,j).
// Hx [i] is not initialized. M is present.
// M(i,j) is 1. (either M or !M case)
// if M: C(i,j) is a new entry seen for the first time.
// if !M: Hf [i] stays equal to 1 (or 3 if locked)
// Hf [i] == 2: unlocked, i has been seen in C(:,j).
// Hx [i] is initialized. This case is independent of M.
// Hf [i] == 3: locked. Hx [i] cannot be accessed.
int8_t *restrict
Hf = (int8_t *restrict) SaxpyTasks [taskid].Hf ;
#if ( GB_NO_MASK )
{
// phase2: fine Gustavson task, C(:,j)=A*B(:,j)
#include "GB_AxB_saxpy3_fineGus_phase2.c"
}
#elif ( !GB_MASK_COMP )
{
// phase2: fine Gustavson task, C(:,j)<M(:,j)>=A*B(:,j)
#include "GB_AxB_saxpy3_fineGus_M_phase2.c"
}
#else
{
// phase2: fine Gustavson task, C(:,j)<!M(:,j)>=A*B(:,j)
#include "GB_AxB_saxpy3_fineGus_notM_phase2.c"
}
#endif
}
else
{
//------------------------------------------------------------------
// phase2: fine hash task
//------------------------------------------------------------------
// Each hash entry Hf [hash] splits into two parts, (h,f). f
// is in the 2 least significant bits. h is 62 bits, and is
// the 1-based index i of the C(i,j) entry stored at that
// location in the hash table.
// If M is present (M or !M), and M(i,j)=1, then (i+1,1)
// has been inserted into the hash table, in phase0.
// Given Hf [hash] split into (h,f)
// h == 0, f == 0: unlocked and unoccupied.
// note that if f=0, h must be zero too.
// h == i+1, f == 1: unlocked, occupied by M(i,j)=1.
// C(i,j) has not been seen, or is ignored.
// Hx is not initialized. M is present.
// if !M: this entry will be ignored in C.
// h == i+1, f == 2: unlocked, occupied by C(i,j).
// Hx is initialized. M is no longer
// relevant.
// h == (anything), f == 3: locked.
int64_t *restrict Hf = (int64_t *restrict) SaxpyTasks [taskid].Hf ;
int64_t hash_bits = (hash_size-1) ;
#if ( GB_NO_MASK )
{
//--------------------------------------------------------------
// phase2: fine hash task, C(:,j)=A*B(:,j)
//--------------------------------------------------------------
// no mask present, or mask ignored
#undef GB_CHECK_MASK_ij
#include "GB_AxB_saxpy3_fineHash_phase2.c"
}
#elif ( !GB_MASK_COMP )
{
//--------------------------------------------------------------
// phase2: fine hash task, C(:,j)<M(:,j)>=A*B(:,j)
//--------------------------------------------------------------
GB_GET_M_j ; // get M(:,j)
if (M_in_place)
{
// M is bitmap/as-if-full, thus not scattered into Hf
if (M_is_bitmap && Mask_struct)
{
// M is bitmap and structural
const int8_t *restrict Mjb = Mb + pM_start ;
#undef GB_CHECK_MASK_ij
#define GB_CHECK_MASK_ij \
if (!Mjb [i]) continue ;
#include "GB_AxB_saxpy3_fineHash_phase2.c"
}
else
{
// M is bitmap/dense
#undef GB_CHECK_MASK_ij
#define GB_CHECK_MASK_ij \
const int64_t pM = pM_start + i ; \
GB_GET_M_ij (pM) ; \
if (!mij) continue ;
#include "GB_AxB_saxpy3_fineHash_phase2.c"
}
}
else
{
// M(:,j) is sparse and scattered into Hf
#include "GB_AxB_saxpy3_fineHash_M_phase2.c"
}
}
#else
{
//--------------------------------------------------------------
// phase2: fine hash task, C(:,j)<!M(:,j)>=A*B(:,j)
//--------------------------------------------------------------
GB_GET_M_j ; // get M(:,j)
if (M_in_place)
{
// M is bitmap/as-if-full, thus not scattered into Hf
if (M_is_bitmap && Mask_struct)
{
// M is bitmap and structural
const int8_t *restrict Mjb = Mb + pM_start ;
#undef GB_CHECK_MASK_ij
#define GB_CHECK_MASK_ij \
if (Mjb [i]) continue ;
#include "GB_AxB_saxpy3_fineHash_phase2.c"
}
else
{
// M is bitmap/dense
#undef GB_CHECK_MASK_ij
#define GB_CHECK_MASK_ij \
const int64_t pM = pM_start + i ; \
GB_GET_M_ij (pM) ; \
if (mij) continue ;
#include "GB_AxB_saxpy3_fineHash_phase2.c"
}
}
else
{
// M(:,j) is sparse/hyper and scattered into Hf
#include "GB_AxB_saxpy3_fineHash_notM_phase2.c"
}
}
#endif
}
}
#ifdef GB_TIMING
ttt = omp_get_wtime ( ) - ttt ;
GB_Global_timing_add (9, ttt) ;
ttt = omp_get_wtime ( ) ;
#endif
//==========================================================================
// phase3/phase4: count nnz(C(:,j)) for fine tasks, cumsum of Cp
//==========================================================================
GB_AxB_saxpy3_cumsum (C, SaxpyTasks, nfine, chunk, nthreads, Context) ;
#ifdef GB_TIMING
ttt = omp_get_wtime ( ) - ttt ;
GB_Global_timing_add (10, ttt) ;
ttt = omp_get_wtime ( ) ;
#endif
//==========================================================================
// phase5: numeric phase for coarse tasks, gather for fine tasks
//==========================================================================
// C is iso for the ANY_PAIR semiring, and non-iso otherwise
// allocate Ci and Cx
int64_t cnz = Cp [cnvec] ;
// set C->iso = GB_IS_ANY_PAIR_SEMIRING OK
GrB_Info info = GB_bix_alloc (C, cnz, GxB_SPARSE, false, true,
GB_IS_ANY_PAIR_SEMIRING, Context) ;
if (info != GrB_SUCCESS)
{
// out of memory
return (GrB_OUT_OF_MEMORY) ;
}
C->nvals = cnz ;
int64_t *restrict Ci = C->i ;
#if ( !GB_IS_ANY_PAIR_SEMIRING )
GB_CTYPE *restrict Cx = (GB_CTYPE *) C->x ;
#endif
ASSERT (C->i_size == GB_Global_memtable_size (C->i)) ;
#ifdef GB_TIMING
ttt = omp_get_wtime ( ) - ttt ;
GB_Global_timing_add (11, ttt) ;
ttt = omp_get_wtime ( ) ;
#endif
bool C_jumbled = false ;
#pragma omp parallel for num_threads(nthreads) schedule(dynamic,1) \
reduction(||:C_jumbled)
for (taskid = 0 ; taskid < ntasks ; taskid++)
{
//----------------------------------------------------------------------
// get the task descriptor
//----------------------------------------------------------------------
#if !GB_IS_ANY_PAIR_SEMIRING
GB_CTYPE *restrict Hx = (GB_CTYPE *) SaxpyTasks [taskid].Hx ;
#endif
int64_t hash_size = SaxpyTasks [taskid].hsize ;
bool use_Gustavson = (hash_size == cvlen) ;
bool task_C_jumbled = false ;
if (taskid < nfine)
{
//------------------------------------------------------------------
// fine task: gather pattern and values
//------------------------------------------------------------------
int64_t kk = SaxpyTasks [taskid].vector ;
int team_size = SaxpyTasks [taskid].team_size ;
int leader = SaxpyTasks [taskid].leader ;
int my_teamid = taskid - leader ;
int64_t pC = Cp [kk] ;
if (use_Gustavson)
{
//--------------------------------------------------------------
// phase5: fine Gustavson task, C=A*B, C<M>=A*B, or C<!M>=A*B
//--------------------------------------------------------------
// Hf [i] == 2 if C(i,j) is an entry in C(:,j)
int8_t *restrict
Hf = (int8_t *restrict) SaxpyTasks [taskid].Hf ;
int64_t cjnz = Cp [kk+1] - pC ;
int64_t istart, iend ;
GB_PARTITION (istart, iend, cvlen, my_teamid, team_size) ;
if (cjnz == cvlen)
{
// C(:,j) is dense
for (int64_t i = istart ; i < iend ; i++)
{
Ci [pC + i] = i ;
}
// copy Hx [istart:iend-1] into Cx [pC+istart:pC+iend-1]
GB_CIJ_MEMCPY (pC + istart, istart, iend - istart) ;
}
else
{
// C(:,j) is sparse
pC += SaxpyTasks [taskid].my_cjnz ;
for (int64_t i = istart ; i < iend ; i++)
{
if (Hf [i] == 2)
{
GB_CIJ_GATHER (pC, i) ; // Cx [pC] = Hx [i]
Ci [pC++] = i ;
}
}
}
}
else
{
//--------------------------------------------------------------
// phase5: fine hash task, C=A*B, C<M>=A*B, C<!M>=A*B
//--------------------------------------------------------------
// (Hf [hash] & 3) == 2 if C(i,j) is an entry in C(:,j),
// and the index i of the entry is (Hf [hash] >> 2) - 1.
int64_t *restrict
Hf = (int64_t *restrict) SaxpyTasks [taskid].Hf ;
int64_t mystart, myend ;
GB_PARTITION (mystart, myend, hash_size, my_teamid, team_size) ;
pC += SaxpyTasks [taskid].my_cjnz ;
for (int64_t hash = mystart ; hash < myend ; hash++)
{
int64_t hf = Hf [hash] ;
if ((hf & 3) == 2)
{
int64_t i = (hf >> 2) - 1 ; // found C(i,j) in hash
Ci [pC] = i ;
GB_CIJ_GATHER (pC, hash) ; // Cx [pC] = Hx [hash]
pC++ ;
}
}
task_C_jumbled = true ;
}
}
else
{
//------------------------------------------------------------------
// numeric coarse task: compute C(:,kfirst:klast)
//------------------------------------------------------------------
int64_t *restrict
Hf = (int64_t *restrict) SaxpyTasks [taskid].Hf ;
int64_t kfirst = SaxpyTasks [taskid].start ;
int64_t klast = SaxpyTasks [taskid].end ;
int64_t nk = klast - kfirst + 1 ;
int64_t mark = 2*nk + 1 ;
if (use_Gustavson)
{
//--------------------------------------------------------------
// phase5: coarse Gustavson task
//--------------------------------------------------------------
#if ( GB_NO_MASK )
{
// phase5: coarse Gustavson task, C=A*B
#include "GB_AxB_saxpy3_coarseGus_noM_phase5.c"
}
#elif ( !GB_MASK_COMP )
{
// phase5: coarse Gustavson task, C<M>=A*B
#include "GB_AxB_saxpy3_coarseGus_M_phase5.c"
}
#else
{
// phase5: coarse Gustavson task, C<!M>=A*B
#include "GB_AxB_saxpy3_coarseGus_notM_phase5.c"
}
#endif
}
else
{
//--------------------------------------------------------------
// phase5: coarse hash task
//--------------------------------------------------------------
int64_t *restrict Hi = SaxpyTasks [taskid].Hi ;
int64_t hash_bits = (hash_size-1) ;
#if ( GB_NO_MASK )
{
//----------------------------------------------------------
// phase5: coarse hash task, C=A*B
//----------------------------------------------------------
// no mask present, or mask ignored (see below)
#undef GB_CHECK_MASK_ij
#include "GB_AxB_saxpy3_coarseHash_phase5.c"
}
#elif ( !GB_MASK_COMP )
{
//----------------------------------------------------------
// phase5: coarse hash task, C<M>=A*B
//----------------------------------------------------------
if (M_in_place)
{
// M is bitmap/as-if-full, thus not scattered into Hf
if (M_is_bitmap && Mask_struct)
{
// M is bitmap and structural
#define GB_MASK_IS_BITMAP_AND_STRUCTURAL
#undef GB_CHECK_MASK_ij
#define GB_CHECK_MASK_ij \
if (!Mjb [i]) continue ;
#include "GB_AxB_saxpy3_coarseHash_phase5.c"
}
else
{
// M is bitmap/dense
#undef GB_CHECK_MASK_ij
#define GB_CHECK_MASK_ij \
const int64_t pM = pM_start + i ; \
GB_GET_M_ij (pM) ; \
if (!mij) continue ;
#include "GB_AxB_saxpy3_coarseHash_phase5.c"
}
}
else
{
// M is sparse and scattered into Hf
#include "GB_AxB_saxpy3_coarseHash_M_phase5.c"
}
}
#else
{
//----------------------------------------------------------
// phase5: coarse hash task, C<!M>=A*B
//----------------------------------------------------------
if (M_in_place)
{
// M is bitmap/as-if-full, thus not scattered into Hf
if (M_is_bitmap && Mask_struct)
{
// M is bitmap and structural
#define GB_MASK_IS_BITMAP_AND_STRUCTURAL
#undef GB_CHECK_MASK_ij
#define GB_CHECK_MASK_ij \
if (Mjb [i]) continue ;
#include "GB_AxB_saxpy3_coarseHash_phase5.c"
}
else
{
// M is bitmap/dense
#undef GB_CHECK_MASK_ij
#define GB_CHECK_MASK_ij \
const int64_t pM = pM_start + i ; \
GB_GET_M_ij (pM) ; \
if (mij) continue ;
#include "GB_AxB_saxpy3_coarseHash_phase5.c"
}
}
else
{
// M is sparse and scattered into Hf
#include "GB_AxB_saxpy3_coarseHash_notM_phase5.c"
}
}
#endif
}
}
C_jumbled = C_jumbled || task_C_jumbled ;
}
//--------------------------------------------------------------------------
// log the state of C->jumbled
//--------------------------------------------------------------------------
C->jumbled = C_jumbled ; // C is jumbled if any task left it jumbled
#ifdef GB_TIMING
ttt = omp_get_wtime ( ) - ttt ;
GB_Global_timing_add (12, ttt) ;
#endif
}
#undef GB_NO_MASK
#undef GB_MASK_COMP
|
