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
|
//------------------------------------------------------------------------------
// GB_unop_transpose: C=op(cast(A')), transpose, typecast, and apply op
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2022, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
{
// Ax unused for some uses of this template
#include "GB_unused.h"
//--------------------------------------------------------------------------
// get A and C
//--------------------------------------------------------------------------
#ifndef GB_ISO_TRANSPOSE
const GB_ATYPE *restrict Ax = (GB_ATYPE *) A->x ;
GB_CTYPE *restrict Cx = (GB_CTYPE *) C->x ;
#endif
//--------------------------------------------------------------------------
// C = op (cast (A'))
//--------------------------------------------------------------------------
if (Workspaces == NULL)
{
//----------------------------------------------------------------------
// A and C are both full or both bitmap
//----------------------------------------------------------------------
// A is avlen-by-avdim; C is avdim-by-avlen
int64_t avlen = A->vlen ;
int64_t avdim = A->vdim ;
int64_t anz = avlen * avdim ; // ignore integer overflow
const int8_t *restrict Ab = A->b ;
int8_t *restrict Cb = C->b ;
ASSERT ((Cb == NULL) == (Ab == NULL)) ;
// TODO: it would be faster to do this by tiles, not rows/columns, for
// large matrices, but in most of the cases in GraphBLAS, A and C will
// be tall-and-thin or short-and-fat.
if (Ab == NULL)
{
//------------------------------------------------------------------
// A and C are both full (no work if A and C are iso)
//------------------------------------------------------------------
#ifndef GB_ISO_TRANSPOSE
int tid ;
#pragma omp parallel for num_threads(nthreads) schedule(static)
for (tid = 0 ; tid < nthreads ; tid++)
{
int64_t pC_start, pC_end ;
GB_PARTITION (pC_start, pC_end, anz, tid, nthreads) ;
for (int64_t pC = pC_start ; pC < pC_end ; pC++)
{
// get i and j of the entry C(i,j)
// i = (pC % avdim) ;
// j = (pC / avdim) ;
// find the position of the entry A(j,i)
// pA = j + i * avlen
// Cx [pC] = op (Ax [pA])
GB_CAST_OP (pC, ((pC/avdim) + (pC%avdim) * avlen)) ;
}
}
#endif
}
else
{
//------------------------------------------------------------------
// A and C are both bitmap
//------------------------------------------------------------------
int tid ;
#pragma omp parallel for num_threads(nthreads) schedule(static)
for (tid = 0 ; tid < nthreads ; tid++)
{
int64_t pC_start, pC_end ;
GB_PARTITION (pC_start, pC_end, anz, tid, nthreads) ;
for (int64_t pC = pC_start ; pC < pC_end ; pC++)
{
// get i and j of the entry C(i,j)
// i = (pC % avdim) ;
// j = (pC / avdim) ;
// find the position of the entry A(j,i)
// pA = j + i * avlen
int64_t pA = ((pC / avdim) + (pC % avdim) * avlen) ;
int8_t cij_exists = Ab [pA] ;
Cb [pC] = cij_exists ;
#ifndef GB_ISO_TRANSPOSE
if (cij_exists)
{
// Cx [pC] = op (Ax [pA])
GB_CAST_OP (pC, pA) ;
}
#endif
}
}
}
}
else
{
//----------------------------------------------------------------------
// A is sparse or hypersparse; C is sparse
//----------------------------------------------------------------------
const int64_t *restrict Ap = A->p ;
const int64_t *restrict Ah = A->h ;
const int64_t *restrict Ai = A->i ;
const int64_t anvec = A->nvec ;
int64_t *restrict Ci = C->i ;
if (nthreads == 1)
{
//------------------------------------------------------------------
// sequential method
//------------------------------------------------------------------
int64_t *restrict workspace = Workspaces [0] ;
for (int64_t k = 0 ; k < anvec ; k++)
{
// iterate over the entries in A(:,j)
int64_t j = GBH (Ah, k) ;
int64_t pA_start = Ap [k] ;
int64_t pA_end = Ap [k+1] ;
for (int64_t pA = pA_start ; pA < pA_end ; pA++)
{
// C(j,i) = A(i,j)
int64_t i = Ai [pA] ;
int64_t pC = workspace [i]++ ;
Ci [pC] = j ;
#ifndef GB_ISO_TRANSPOSE
// Cx [pC] = op (Ax [pA])
GB_CAST_OP (pC, pA) ;
#endif
}
}
}
else if (nworkspaces == 1)
{
//------------------------------------------------------------------
// atomic method
//------------------------------------------------------------------
int64_t *restrict workspace = Workspaces [0] ;
int tid ;
#pragma omp parallel for num_threads(nthreads) schedule(static)
for (tid = 0 ; tid < nthreads ; tid++)
{
for (int64_t k = A_slice [tid] ; k < A_slice [tid+1] ; k++)
{
// iterate over the entries in A(:,j)
int64_t j = GBH (Ah, k) ;
int64_t pA_start = Ap [k] ;
int64_t pA_end = Ap [k+1] ;
for (int64_t pA = pA_start ; pA < pA_end ; pA++)
{
// C(j,i) = A(i,j)
int64_t i = Ai [pA] ;
// do this atomically: pC = workspace [i]++
int64_t pC ;
GB_ATOMIC_CAPTURE_INC64 (pC, workspace [i]) ;
Ci [pC] = j ;
#ifndef GB_ISO_TRANSPOSE
// Cx [pC] = op (Ax [pA])
GB_CAST_OP (pC, pA) ;
#endif
}
}
}
}
else
{
//------------------------------------------------------------------
// non-atomic method
//------------------------------------------------------------------
int tid ;
#pragma omp parallel for num_threads(nthreads) schedule(static)
for (tid = 0 ; tid < nthreads ; tid++)
{
int64_t *restrict workspace = Workspaces [tid] ;
for (int64_t k = A_slice [tid] ; k < A_slice [tid+1] ; k++)
{
// iterate over the entries in A(:,j)
int64_t j = GBH (Ah, k) ;
int64_t pA_start = Ap [k] ;
int64_t pA_end = Ap [k+1] ;
for (int64_t pA = pA_start ; pA < pA_end ; pA++)
{
// C(j,i) = A(i,j)
int64_t i = Ai [pA] ;
int64_t pC = workspace [i]++ ;
Ci [pC] = j ;
#ifndef GB_ISO_TRANSPOSE
// Cx [pC] = op (Ax [pA])
GB_CAST_OP (pC, pA) ;
#endif
}
}
}
}
}
}
#undef GB_ISO_TRANSPOSE
|
