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
|
//------------------------------------------------------------------------------
// GB_AxB_saxpy_generic: compute C=A*B, C<M>=A*B, or C<!M>=A*B in parallel
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2025, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
// GB_AxB_saxpy_generic computes C=A*B, C<M>=A*B, or C<!M>=A*B in parallel,
// with arbitrary types and operators, via memcpy and function pointers. C can
// be hyper, sparse, or bitmap (never full). For all cases, the four matrices
// C, M (if present), A, and B have the same format (by-row or by-column), or
// they represent implicitly transposed matrices with the same effect. This
// method does not handle the dot-product methods, which compute C=A'*B if A
// and B are held by column, or equivalently A*B' if both are held by row.
// This method uses GB_AxB_saxpy3_generic_* and GB_AxB_saxbit_generic_*
// to implement two meta-methods, each of which can contain further specialized
// methods (such as the fine/ coarse x Gustavson/Hash, mask/no-mask methods in
// saxpy3):
// saxpy3: general purpose method, where C is sparse or hypersparse,
// via GB_AxB_saxpy3_template.c. SaxpyTasks holds the (fine/coarse x
// Gustavson/Hash) tasks constructed by GB_AxB_saxpy3_slice*.
// saxbit: general purpose method, where C is bitmap, via
// GB_AxB_saxbit_template.c.
// C is not iso, and it is never full.
//------------------------------------------------------------------------------
#include "mxm/GB_mxm.h"
#include "binaryop/GB_binop.h"
#include "mxm/GB_AxB_saxpy_generic.h"
GrB_Info GB_AxB_saxpy_generic
(
GrB_Matrix C, // any sparsity
const GrB_Matrix M,
bool Mask_comp,
const bool Mask_struct,
const bool M_in_place, // ignored if C is bitmap
const GrB_Matrix A,
bool A_is_pattern,
const GrB_Matrix B,
bool B_is_pattern,
const GrB_Semiring semiring, // semiring that defines C=A*B
const bool flipxy, // if true, do z=fmult(b,a) vs fmult(a,b)
const int saxpy_method, // saxpy3 or bitmap method
const int ntasks,
const int nthreads,
// for saxpy3 only:
GB_saxpy3task_struct *restrict SaxpyTasks, // NULL if C is bitmap
const int nfine,
const int do_sort, // if true, sort in saxpy3
GB_Werk Werk,
// for saxbit only:
const int nfine_tasks_per_vector,
const bool use_coarse_tasks,
const bool use_atomics,
const int64_t *restrict M_ek_slicing,
const int M_nthreads,
const int M_ntasks,
const int64_t *restrict A_slice,
const int64_t *restrict H_slice,
GB_void *restrict Wcx,
int8_t *restrict Wf
)
{
//--------------------------------------------------------------------------
// get operators, functions, workspace, contents of A, B, and C
//--------------------------------------------------------------------------
GrB_Info info = GrB_NO_VALUE ;
GrB_BinaryOp mult = semiring->multiply ;
GB_Opcode opcode = mult->opcode ;
//--------------------------------------------------------------------------
// C = A*B via saxpy3 or bitmap method, function pointers, and typecasting
//--------------------------------------------------------------------------
if (opcode == GB_FIRST_binop_code)
{
//----------------------------------------------------------------------
// generic semirings with FIRST multiply operators
//----------------------------------------------------------------------
GB_BURBLE_MATRIX (C, "(generic first C=A*B) ") ;
// t = A(i,k)
// mult->binop_function is not used and can be NULL. This is required
// for GB_reduce_to_vector for user-defined types.
ASSERT (!flipxy) ;
ASSERT (B_is_pattern) ;
if (saxpy_method == GB_SAXPY_METHOD_3)
{
// C is sparse or hypersparse
info = GB_AxB_saxpy3_generic_first (C, M, Mask_comp, Mask_struct,
M_in_place, A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, SaxpyTasks, nfine, do_sort, Werk) ;
}
else
{
// C is bitmap
info = GB_AxB_saxbit_generic_first (C, M, Mask_comp, Mask_struct,
M_in_place, A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, nfine_tasks_per_vector, use_coarse_tasks,
use_atomics, M_ek_slicing, M_nthreads, M_ntasks,
A_slice, H_slice, Wcx, Wf) ;
}
}
else if (opcode == GB_SECOND_binop_code)
{
//----------------------------------------------------------------------
// generic semirings with SECOND multiply operators
//----------------------------------------------------------------------
GB_BURBLE_MATRIX (C, "(generic second C=A*B) ") ;
// t = B(i,k)
// mult->binop_function is not used and can be NULL. This is required
// for GB_reduce_to_vector for user-defined types.
ASSERT (!flipxy) ;
ASSERT (A_is_pattern) ;
if (saxpy_method == GB_SAXPY_METHOD_3)
{
// C is sparse or hypersparse
info = GB_AxB_saxpy3_generic_second (C, M, Mask_comp, Mask_struct,
M_in_place, A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, SaxpyTasks, nfine, do_sort, Werk) ;
}
else
{
// C is bitmap
info = GB_AxB_saxbit_generic_second (C, M, Mask_comp, Mask_struct,
M_in_place, A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, nfine_tasks_per_vector, use_coarse_tasks,
use_atomics, M_ek_slicing, M_nthreads, M_ntasks,
A_slice, H_slice, Wcx, Wf) ;
}
}
else if (mult->binop_function != NULL)
{
//----------------------------------------------------------------------
// generic semirings with standard multiply operators
//----------------------------------------------------------------------
// standard binary op
GB_BURBLE_MATRIX (C, "(generic C=A*B) ") ;
if (flipxy)
{
// t = B(k,j) * A(i,k)
if (saxpy_method == GB_SAXPY_METHOD_3)
{
// C is sparse or hypersparse, mult is flipped
info = GB_AxB_saxpy3_generic_flipped (C, M,
Mask_comp, Mask_struct, M_in_place,
A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, SaxpyTasks, nfine, do_sort, Werk) ;
}
else
{
// C is bitmap, mult is flipped
info = GB_AxB_saxbit_generic_flipped (C, M,
Mask_comp, Mask_struct, M_in_place,
A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, nfine_tasks_per_vector, use_coarse_tasks,
use_atomics, M_ek_slicing, M_nthreads, M_ntasks,
A_slice, H_slice, Wcx, Wf) ;
}
}
else
{
// t = A(i,k) * B(k,j)
if (saxpy_method == GB_SAXPY_METHOD_3)
{
// C is sparse or hypersparse, mult is unflipped
info = GB_AxB_saxpy3_generic_unflipped (C, M,
Mask_comp, Mask_struct, M_in_place,
A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, SaxpyTasks, nfine, do_sort, Werk) ;
}
else
{
// C is bitmap, mult is unflipped
info = GB_AxB_saxbit_generic_unflipped (C, M,
Mask_comp, Mask_struct, M_in_place,
A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, nfine_tasks_per_vector, use_coarse_tasks,
use_atomics, M_ek_slicing, M_nthreads, M_ntasks,
A_slice, H_slice, Wcx, Wf) ;
}
}
}
else
{
//----------------------------------------------------------------------
// generic semirings with index binary multiply operators
//----------------------------------------------------------------------
GB_BURBLE_MATRIX (C, "(generic index C=A*B) ") ;
ASSERT (mult->idxbinop_function != NULL) ;
if (flipxy)
{
// t = B(k,j) * A(i,k)
if (saxpy_method == GB_SAXPY_METHOD_3)
{
// C is sparse or hypersparse, mult is flipped
info = GB_AxB_saxpy3_generic_idx_flipped (C, M,
Mask_comp, Mask_struct, M_in_place,
A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, SaxpyTasks, nfine, do_sort, Werk) ;
}
else
{
// C is bitmap, mult is flipped
info = GB_AxB_saxbit_generic_idx_flipped (C, M,
Mask_comp, Mask_struct, M_in_place,
A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, nfine_tasks_per_vector, use_coarse_tasks,
use_atomics, M_ek_slicing, M_nthreads, M_ntasks,
A_slice, H_slice, Wcx, Wf) ;
}
}
else
{
// t = A(i,k) * B(k,j)
if (saxpy_method == GB_SAXPY_METHOD_3)
{
// C is sparse or hypersparse, mult is unflipped
info = GB_AxB_saxpy3_generic_idx_unflipped (C, M,
Mask_comp, Mask_struct, M_in_place,
A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, SaxpyTasks, nfine, do_sort, Werk) ;
}
else
{
// C is bitmap, mult is unflipped
info = GB_AxB_saxbit_generic_idx_unflipped (C, M,
Mask_comp, Mask_struct, M_in_place,
A, A_is_pattern, B, B_is_pattern, semiring,
ntasks, nthreads, nfine_tasks_per_vector, use_coarse_tasks,
use_atomics, M_ek_slicing, M_nthreads, M_ntasks,
A_slice, H_slice, Wcx, Wf) ;
}
}
}
return (info) ;
}
|
