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
|
//------------------------------------------------------------------------------
// GB_dense_subassign_05d: C(:,:)<M> = scalar where C is as-if-full
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2022, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
// Method 05d: C(:,:)<M> = scalar ; no S, C is dense
// M: present
// Mask_comp: false
// C_replace: false
// accum: NULL
// A: scalar
// S: none
// C can have any sparsity structure, but it must be entirely dense with
// all entries present.
#include "GB_subassign_methods.h"
#include "GB_dense.h"
#include "GB_unused.h"
#ifndef GBCUDA_DEV
#include "GB_type__include.h"
#endif
#undef GB_FREE_WORKSPACE
#define GB_FREE_WORKSPACE \
{ \
GB_WERK_POP (M_ek_slicing, int64_t) ; \
}
#undef GB_FREE_ALL
#define GB_FREE_ALL GB_FREE_WORKSPACE
GrB_Info GB_dense_subassign_05d
(
GrB_Matrix C,
// input:
const GrB_Matrix M,
const bool Mask_struct,
const void *scalar,
const GrB_Type atype,
GB_Context Context
)
{
//--------------------------------------------------------------------------
// check inputs
//--------------------------------------------------------------------------
ASSERT (!GB_aliased (C, M)) ; // NO ALIAS of C==M
//--------------------------------------------------------------------------
// get inputs
//--------------------------------------------------------------------------
GrB_Info info ;
GB_WERK_DECLARE (M_ek_slicing, int64_t) ;
ASSERT_MATRIX_OK (C, "C for subassign method_05d", GB0) ;
ASSERT (!GB_ZOMBIES (C)) ;
ASSERT (!GB_JUMBLED (C)) ;
ASSERT (!GB_PENDING (C)) ;
ASSERT (GB_as_if_full (C)) ;
ASSERT_MATRIX_OK (M, "M for subassign method_05d", GB0) ;
ASSERT (!GB_ZOMBIES (M)) ;
ASSERT (GB_JUMBLED_OK (M)) ;
ASSERT (!GB_PENDING (M)) ;
GB_ENSURE_FULL (C) ; // convert C to full, if sparsity control allows it
if (C->iso)
{
// work has already been done by GB_assign_prep
return (GrB_SUCCESS) ;
}
const GB_Type_code ccode = C->type->code ;
const size_t csize = C->type->size ;
GB_GET_SCALAR ;
//--------------------------------------------------------------------------
// Method 05d: C(:,:)<M> = scalar ; no S; C is dense
//--------------------------------------------------------------------------
// Time: Optimal: the method must iterate over all entries in M,
// and the time is O(nnz(M)).
//--------------------------------------------------------------------------
// Parallel: slice M into equal-sized chunks
//--------------------------------------------------------------------------
GB_GET_NTHREADS_MAX (nthreads_max, chunk, Context) ;
//--------------------------------------------------------------------------
// slice the entries for each task
//--------------------------------------------------------------------------
int M_ntasks, M_nthreads ;
GB_SLICE_MATRIX (M, 8, chunk) ;
//--------------------------------------------------------------------------
// C<M> = x for built-in types
//--------------------------------------------------------------------------
bool done = false ;
#ifndef GBCUDA_DEV
//----------------------------------------------------------------------
// define the worker for the switch factory
//----------------------------------------------------------------------
#define GB_Cdense_05d(cname) GB (_Cdense_05d_ ## cname)
#define GB_WORKER(cname) \
{ \
info = GB_Cdense_05d(cname) (C, M, Mask_struct, cwork, \
M_ek_slicing, M_ntasks, M_nthreads) ; \
done = (info != GrB_NO_VALUE) ; \
} \
break ;
//----------------------------------------------------------------------
// launch the switch factory
//----------------------------------------------------------------------
// C<M> = x
switch (ccode)
{
case GB_BOOL_code : GB_WORKER (_bool )
case GB_INT8_code : GB_WORKER (_int8 )
case GB_INT16_code : GB_WORKER (_int16 )
case GB_INT32_code : GB_WORKER (_int32 )
case GB_INT64_code : GB_WORKER (_int64 )
case GB_UINT8_code : GB_WORKER (_uint8 )
case GB_UINT16_code : GB_WORKER (_uint16)
case GB_UINT32_code : GB_WORKER (_uint32)
case GB_UINT64_code : GB_WORKER (_uint64)
case GB_FP32_code : GB_WORKER (_fp32 )
case GB_FP64_code : GB_WORKER (_fp64 )
case GB_FC32_code : GB_WORKER (_fc32 )
case GB_FC64_code : GB_WORKER (_fc64 )
default: ;
}
#endif
//--------------------------------------------------------------------------
// C<M> = x for user-defined types
//--------------------------------------------------------------------------
if (!done)
{
//----------------------------------------------------------------------
// get operators, functions, workspace, contents of A and C
//----------------------------------------------------------------------
GB_BURBLE_MATRIX (M, "(generic C(:,:)<M>=x assign) ") ;
const size_t csize = C->type->size ;
// Cx [p] = scalar
#define GB_COPY_SCALAR_TO_C(p,x) \
memcpy (Cx + ((p)*csize), x, csize)
#define GB_CTYPE GB_void
// no vectorization
#define GB_PRAGMA_SIMD_VECTORIZE ;
#include "GB_dense_subassign_05d_template.c"
}
//--------------------------------------------------------------------------
// free workspace and return result
//--------------------------------------------------------------------------
GB_FREE_WORKSPACE ;
ASSERT_MATRIX_OK (C, "C output for subassign method_05d", GB0) ;
return (GrB_SUCCESS) ;
}
|
