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
|
//------------------------------------------------------------------------------
// GB_subassign_15_template: C(I,J)<!M> += scalar ; using S
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2025, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
// Method 15: C(I,J)<!M> += scalar ; using S
// M: present
// Mask_comp: true
// Mask_struct: true or false
// C_replace: false
// accum: present
// A: scalar
// S: constructed
// C: not bitmap, but can be full since no zombies are inserted in that case
// M: not bitmap
{
//--------------------------------------------------------------------------
// get inputs
//--------------------------------------------------------------------------
GB_EMPTY_TASKLIST ;
GB_GET_C ; // C must not be bitmap
GB_GET_MASK ;
GB_GET_MASK_HYPER_HASH ;
GB_GET_S ;
GB_GET_ACCUM_SCALAR ;
//--------------------------------------------------------------------------
// Method 15: C(I,J)<!M> += scalar ; using S
//--------------------------------------------------------------------------
// Time: Close to optimal; must visit all IxJ, so Omega(|I|*|J|) is
// required. The sparsity of !M cannot be exploited.
// Methods 13, 15, 17, and 19 are very similar.
//--------------------------------------------------------------------------
// Parallel: all IxJ (Methods 01, 03, 13, 15, 17, 19)
//--------------------------------------------------------------------------
GB_SUBASSIGN_IXJ_SLICE ;
//--------------------------------------------------------------------------
// phase 1: create zombies, update entries, and count pending tuples
//--------------------------------------------------------------------------
#pragma omp parallel for num_threads(nthreads) schedule(dynamic,1) \
reduction(+:nzombies)
for (taskid = 0 ; taskid < ntasks ; taskid++)
{
//----------------------------------------------------------------------
// get the task descriptor
//----------------------------------------------------------------------
GB_GET_IXJ_TASK_DESCRIPTOR_PHASE1 (iA_start, iA_end) ;
//----------------------------------------------------------------------
// compute all vectors in this task
//----------------------------------------------------------------------
for (int64_t j = kfirst ; j <= klast ; j++)
{
//------------------------------------------------------------------
// get jC, the corresponding vector of C
//------------------------------------------------------------------
int64_t jC = GB_IJLIST (J, j, GB_J_KIND, Jcolon) ;
//------------------------------------------------------------------
// get S(iA_start:end,j) and M(iA_start:end,j)
//------------------------------------------------------------------
GB_LOOKUP_VECTOR_S_FOR_IXJ (j, pS, pS_end, iA_start) ;
GB_LOOKUP_VECTOR_M_FOR_IXJ (j, pM, pM_end, iA_start) ;
//------------------------------------------------------------------
// C(I(iA_start,iA_end-1),jC)<!M> += scalar
//------------------------------------------------------------------
for (int64_t iA = iA_start ; iA < iA_end ; iA++)
{
//--------------------------------------------------------------
// Get the indices at the top of each list.
//--------------------------------------------------------------
int64_t iS = (pS < pS_end) ? GBi_S (Si, pS, Svlen) : INT64_MAX ;
int64_t iM = (pM < pM_end) ? GBi_M (Mi, pM, Mvlen) : INT64_MAX ;
//--------------------------------------------------------------
// find the smallest index of [iS iA iM] (always iA)
//--------------------------------------------------------------
int64_t i = iA ;
//--------------------------------------------------------------
// get M(i,j)
//--------------------------------------------------------------
bool mij ;
if (i == iM)
{
// mij = (bool) M [pM]
mij = GBb_M (Mb, pM) && GB_MCAST (Mx, pM, msize) ;
pM++ ; // go to the next entry in M(:,j)
}
else
{
// mij not present, implicitly false
ASSERT (i < iM) ;
mij = false ;
}
// complement the mask entry mij since Mask_comp is true
mij = !mij ;
//--------------------------------------------------------------
// accumulate the entry
//--------------------------------------------------------------
if (i == iS)
{
ASSERT (i == iA) ;
{
// both S (i,j) and A (i,j) present
if (mij)
{
// ----[C A 1] or [X A 1]---------------------------
// [C A 1]: action: ( =C+A ): apply accum
// [X A 1]: action: ( undelete ): zombie lives
GB_C_S_LOOKUP ;
GB_withaccum_C_A_1_scalar ;
}
pS++ ; // go to the next entry in S(:,j)
}
}
else
{
ASSERT (i == iA) ;
{
// S (i,j) is not present, A (i,j) is present
if (mij)
{
// ----[. A 1]--------------------------------------
// [. A 1]: action: ( insert )
task_pending++ ;
}
}
}
}
}
GB_PHASE1_TASK_WRAPUP ;
}
//--------------------------------------------------------------------------
// phase 2: insert pending tuples
//--------------------------------------------------------------------------
GB_PENDING_CUMSUM ;
#pragma omp parallel for num_threads(nthreads) schedule(dynamic,1) \
reduction(&&:pending_sorted)
for (taskid = 0 ; taskid < ntasks ; taskid++)
{
//----------------------------------------------------------------------
// get the task descriptor
//----------------------------------------------------------------------
GB_GET_IXJ_TASK_DESCRIPTOR_PHASE2 (iA_start, iA_end) ;
//----------------------------------------------------------------------
// compute all vectors in this task
//----------------------------------------------------------------------
for (int64_t j = kfirst ; j <= klast ; j++)
{
//------------------------------------------------------------------
// get jC, the corresponding vector of C
//------------------------------------------------------------------
int64_t jC = GB_IJLIST (J, j, GB_J_KIND, Jcolon) ;
//------------------------------------------------------------------
// get S(iA_start:end,j) and M(iA_start:end,j)
//------------------------------------------------------------------
GB_LOOKUP_VECTOR_S_FOR_IXJ (j, pS, pS_end, iA_start) ;
GB_LOOKUP_VECTOR_M_FOR_IXJ (j, pM, pM_end, iA_start) ;
//------------------------------------------------------------------
// C(I(iA_start,iA_end-1),jC)<!M> += scalar
//------------------------------------------------------------------
for (int64_t iA = iA_start ; iA < iA_end ; iA++)
{
//--------------------------------------------------------------
// Get the indices at the top of each list.
//--------------------------------------------------------------
int64_t iS = (pS < pS_end) ? GBi_S (Si, pS, Svlen) : INT64_MAX ;
int64_t iM = (pM < pM_end) ? GBi_M (Mi, pM, Mvlen) : INT64_MAX ;
//--------------------------------------------------------------
// find the smallest index of [iS iA iM] (always iA)
//--------------------------------------------------------------
int64_t i = iA ;
//--------------------------------------------------------------
// get M(i,j)
//--------------------------------------------------------------
bool mij ;
if (i == iM)
{
// mij = (bool) M [pM]
mij = GBb_M (Mb, pM) && GB_MCAST (Mx, pM, msize) ;
pM++ ; // go to the next entry in M(:,j)
}
else
{
// mij not present, implicitly false
ASSERT (i < iM) ;
mij = false ;
}
// complement the mask entry mij since Mask_comp is true
mij = !mij ;
//--------------------------------------------------------------
// accumulate the entry
//--------------------------------------------------------------
if (i == iS)
{
ASSERT (i == iA) ;
{
pS++ ; // go to the next entry in S(:,j)
}
}
else
{
ASSERT (i == iA) ;
{
// S (i,j) is not present, A (i,j) is present
if (mij)
{
// ----[. A 1]--------------------------------------
// [. A 1]: action: ( insert )
int64_t iC = GB_IJLIST (I, iA, GB_I_KIND, Icolon) ;
GB_PENDING_INSERT_scalar ;
}
}
}
}
}
GB_PHASE2_TASK_WRAPUP ;
}
//--------------------------------------------------------------------------
// finalize the matrix and return result
//--------------------------------------------------------------------------
GB_SUBASSIGN_WRAPUP ;
}
|
