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
|
//------------------------------------------------------------------------------
// GB_select_phase1: count entries in each vector for C=select(A,thunk)
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2022, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
//--------------------------------------------------------------------------
// get A and its slicing
//--------------------------------------------------------------------------
const int64_t *restrict kfirst_Aslice = A_ek_slicing ;
const int64_t *restrict klast_Aslice = A_ek_slicing + A_ntasks ;
const int64_t *restrict pstart_Aslice = A_ek_slicing + A_ntasks * 2 ;
const int64_t *restrict Ap = A->p ;
const int64_t *restrict Ah = A->h ;
const int64_t *restrict Ai = A->i ;
int64_t avlen = A->vlen ;
int64_t anvec = A->nvec ;
#if defined ( GB_ENTRY_SELECTOR )
//==========================================================================
// entry selector
//==========================================================================
ASSERT (GB_JUMBLED_OK (A)) ;
// The count of live entries kth vector A(:,k) is reduced to the kth scalar
// Cp(k). Each thread computes the reductions on roughly the same number
// of entries, which means that a vector A(:,k) may be reduced by more than
// one thread. The first vector A(:,kfirst) reduced by thread tid may be
// partial, where the prior thread tid-1 (and other prior threads) may also
// do some of the reductions for this same vector A(:,kfirst). The thread
// tid reduces all vectors A(:,k) for k in the range kfirst+1 to klast-1.
// The last vector A(:,klast) reduced by thread tid may also be partial.
// Thread tid+1, and following threads, may also do some of the reduces for
// A(:,klast).
//--------------------------------------------------------------------------
// get A
//--------------------------------------------------------------------------
const GB_ATYPE *restrict Ax = (GB_ATYPE *) A->x ;
size_t asize = A->type->size ;
int64_t avdim = A->vdim ;
ASSERT (GB_JUMBLED_OK (A)) ;
//--------------------------------------------------------------------------
// reduce each slice
//--------------------------------------------------------------------------
// each thread reduces its own part in parallel
int tid ;
#pragma omp parallel for num_threads(A_nthreads) schedule(dynamic,1)
for (tid = 0 ; tid < A_ntasks ; tid++)
{
// if kfirst > klast then thread tid does no work at all
int64_t kfirst = kfirst_Aslice [tid] ;
int64_t klast = klast_Aslice [tid] ;
Wfirst [tid] = 0 ;
Wlast [tid] = 0 ;
//----------------------------------------------------------------------
// reduce vectors kfirst to klast
//----------------------------------------------------------------------
for (int64_t k = kfirst ; k <= klast ; k++)
{
//------------------------------------------------------------------
// find the part of A(:,k) to be reduced by this thread
//------------------------------------------------------------------
int64_t j = GBH (Ah, k) ;
int64_t pA, pA_end ;
GB_get_pA (&pA, &pA_end, tid, k,
kfirst, klast, pstart_Aslice, Ap, avlen) ;
//------------------------------------------------------------------
// count entries in Ax [pA ... pA_end-1]
//------------------------------------------------------------------
int64_t cjnz = 0 ;
for ( ; pA < pA_end ; pA++)
{
ASSERT (Ai != NULL) ;
int64_t i = Ai [pA] ;
GB_TEST_VALUE_OF_ENTRY (keep, pA) ;
if (keep) cjnz++ ;
}
if (k == kfirst)
{
Wfirst [tid] = cjnz ;
}
else if (k == klast)
{
Wlast [tid] = cjnz ;
}
else
{
Cp [k] = cjnz ;
}
}
}
//--------------------------------------------------------------------------
// reduce the first and last vector of each slice using a single thread
//--------------------------------------------------------------------------
GB_ek_slice_merge1 (Cp, Wfirst, Wlast, A_ek_slicing, A_ntasks) ;
#else
//==========================================================================
// positional selector (tril, triu, diag, offdiag, resize, row*)
//==========================================================================
ASSERT (!GB_JUMBLED (A)) ;
//--------------------------------------------------------------------------
// tril, triu, diag, offdiag, resize: binary search in each vector
//--------------------------------------------------------------------------
int64_t k ;
#pragma omp parallel for num_threads(A_nthreads) schedule(guided)
for (k = 0 ; k < anvec ; k++)
{
//----------------------------------------------------------------------
// get A(:,k)
//----------------------------------------------------------------------
int64_t pA_start = GBP (Ap, k, avlen) ;
int64_t pA_end = GBP (Ap, k+1, avlen) ;
int64_t p = pA_start ;
int64_t cjnz = 0 ;
int64_t ajnz = pA_end - pA_start ;
bool found = false ;
if (ajnz > 0)
{
//------------------------------------------------------------------
// search for the entry A(i,k)
//------------------------------------------------------------------
int64_t ifirst = GBI (Ai, pA_start, avlen) ;
int64_t ilast = GBI (Ai, pA_end-1, avlen) ;
#if defined ( GB_ROWINDEX_SELECTOR )
int64_t i = -ithunk ;
#elif defined ( GB_ROWLE_SELECTOR ) || defined ( GB_ROWGT_SELECTOR )
int64_t i = ithunk ;
#else
// TRIL, TRIU, DIAG, OFFDIAG
int64_t j = GBH (Ah, k) ;
int64_t i = j-ithunk ;
#endif
if (i < ifirst)
{
// all entries in A(:,k) come after i
;
}
else if (i > ilast)
{
// all entries in A(:,k) come before i
p = pA_end ;
}
else if (ajnz == avlen)
{
// A(:,k) is dense
found = true ;
p += i ;
ASSERT (GBI (Ai, p, avlen) == i) ;
}
else
{
// binary search for A (i,k)
int64_t pright = pA_end - 1 ;
GB_SPLIT_BINARY_SEARCH (i, Ai, p, pright, found) ;
}
#if defined ( GB_TRIL_SELECTOR )
// keep p to pA_end-1
cjnz = pA_end - p ;
#elif defined ( GB_ROWGT_SELECTOR )
// if found, keep p+1 to pA_end-1
// else keep p to pA_end-1
if (found)
{
p++ ;
// now in both cases, keep p to pA_end-1
}
// keep p to pA_end-1
cjnz = pA_end - p ;
#elif defined ( GB_TRIU_SELECTOR ) \
|| defined ( GB_ROWLE_SELECTOR )
// if found, keep pA_start to p
// else keep pA_start to p-1
if (found)
{
p++ ;
// now in both cases, keep pA_start to p-1
}
// keep pA_start to p-1
cjnz = p - pA_start ;
#elif defined ( GB_DIAG_SELECTOR )
// if found, keep p
// else keep nothing
cjnz = found ;
if (!found) p = -1 ;
// if (cjnz >= 0) keep p, else keep nothing
#elif defined ( GB_OFFDIAG_SELECTOR ) || \
defined ( GB_ROWINDEX_SELECTOR )
// if found, keep pA_start to p-1 and p+1 to pA_end-1
// else keep pA_start to pA_end
cjnz = ajnz - found ;
if (!found)
{
p = pA_end ;
// now just keep pA_start to p-1; p+1 to pA_end is
// now empty
}
// in both cases, keep pA_start to p-1 and
// p+1 to pA_end-1. If the entry is not found, then
// p == pA_end, and all entries are kept.
#endif
}
//----------------------------------------------------------------------
// log the result for the kth vector
//----------------------------------------------------------------------
Zp [k] = p ;
Cp [k] = cjnz ;
}
//--------------------------------------------------------------------------
// compute Wfirst and Wlast for each task
//--------------------------------------------------------------------------
// Wfirst [0..A_ntasks-1] and Wlast [0..A_ntasks-1] are required for
// constructing C_start_slice [0..A_ntasks-1] in GB_selector.
for (int tid = 0 ; tid < A_ntasks ; tid++)
{
// if kfirst > klast then task tid does no work at all
int64_t kfirst = kfirst_Aslice [tid] ;
int64_t klast = klast_Aslice [tid] ;
Wfirst [tid] = 0 ;
Wlast [tid] = 0 ;
if (kfirst <= klast)
{
int64_t pA_start = pstart_Aslice [tid] ;
int64_t pA_end = GBP (Ap, kfirst+1, avlen) ;
pA_end = GB_IMIN (pA_end, pstart_Aslice [tid+1]) ;
if (pA_start < pA_end)
{
#if defined ( GB_TRIL_SELECTOR ) || \
defined ( GB_ROWGT_SELECTOR )
// keep Zp [kfirst] to pA_end-1
int64_t p = GB_IMAX (Zp [kfirst], pA_start) ;
Wfirst [tid] = GB_IMAX (0, pA_end - p) ;
#elif defined ( GB_TRIU_SELECTOR ) || \
defined ( GB_ROWLE_SELECTOR )
// keep pA_start to Zp [kfirst]-1
int64_t p = GB_IMIN (Zp [kfirst], pA_end) ;
Wfirst [tid] = GB_IMAX (0, p - pA_start) ;
#elif defined ( GB_DIAG_SELECTOR )
// task that owns the diagonal entry does this work
int64_t p = Zp [kfirst] ;
Wfirst [tid] = (pA_start <= p && p < pA_end) ? 1 : 0 ;
#elif defined ( GB_OFFDIAG_SELECTOR ) || \
defined ( GB_ROWINDEX_SELECTOR )
// keep pA_start to Zp [kfirst]-1
int64_t p = GB_IMIN (Zp [kfirst], pA_end) ;
Wfirst [tid] = GB_IMAX (0, p - pA_start) ;
// keep Zp [kfirst]+1 to pA_end-1
p = GB_IMAX (Zp [kfirst]+1, pA_start) ;
Wfirst [tid] += GB_IMAX (0, pA_end - p) ;
#endif
}
}
if (kfirst < klast)
{
int64_t pA_start = GBP (Ap, klast, avlen) ;
int64_t pA_end = pstart_Aslice [tid+1] ;
if (pA_start < pA_end)
{
#if defined ( GB_TRIL_SELECTOR ) || \
defined ( GB_ROWGT_SELECTOR )
// keep Zp [klast] to pA_end-1
int64_t p = GB_IMAX (Zp [klast], pA_start) ;
Wlast [tid] = GB_IMAX (0, pA_end - p) ;
#elif defined ( GB_TRIU_SELECTOR ) || \
defined ( GB_ROWLE_SELECTOR )
// keep pA_start to Zp [klast]-1
int64_t p = GB_IMIN (Zp [klast], pA_end) ;
Wlast [tid] = GB_IMAX (0, p - pA_start) ;
#elif defined ( GB_DIAG_SELECTOR )
// task that owns the diagonal entry does this work
int64_t p = Zp [klast] ;
Wlast [tid] = (pA_start <= p && p < pA_end) ? 1 : 0 ;
#elif defined ( GB_OFFDIAG_SELECTOR ) || \
defined ( GB_ROWINDEX_SELECTOR )
// keep pA_start to Zp [klast]-1
int64_t p = GB_IMIN (Zp [klast], pA_end) ;
Wlast [tid] = GB_IMAX (0, p - pA_start) ;
// keep Zp [klast]+1 to pA_end-1
p = GB_IMAX (Zp [klast]+1, pA_start) ;
Wlast [tid] += GB_IMAX (0, pA_end - p) ;
#endif
}
}
}
#endif
|
