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
|
//------------------------------------------------------------------------------
// GB_AxB_saxpy3_coarseGus_notM_phase5: C<!M>=A*B, coarse Gustavson, phase5
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2022, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
{
//--------------------------------------------------------------------------
// phase5: coarse Gustavson task, C<!M>=A*B
//--------------------------------------------------------------------------
// Since the mask is !M:
// Hf [i] < mark : M(i,j)=0, C(i,j) is not yet seen.
// Hf [i] == mark : M(i,j)=1, so C(i,j) is ignored.
// Hf [i] == mark+1 : M(i,j)=0, and C(i,j) has been seen.
for (int64_t kk = kfirst ; kk <= klast ; kk++)
{
int64_t pC = Cp [kk] ;
int64_t cjnz = Cp [kk+1] - pC ;
if (cjnz == 0) continue ; // nothing to do
GB_GET_B_j ; // get B(:,j)
#ifndef GB_GENERIC
if (cjnz == cvlen) // C(:,j) is dense
{
// This is not used for the generic saxpy3.
GB_COMPUTE_DENSE_C_j ; // C(:,j) = A*B(:,j)
continue ;
}
#endif
GB_GET_M_j ; // get M(:,j)
mark += 2 ;
int64_t mark1 = mark+1 ;
// scatter M(:,j) into the Gustavson workspace
GB_SCATTER_M_j (pM_start, pM_end, mark) ;
if (16 * cjnz > cvlen)
{
//------------------------------------------------------------------
// C(:,j) is not very sparse
//------------------------------------------------------------------
for ( ; pB < pB_end ; pB++) // scan B(:,j)
{
GB_GET_B_kj_INDEX ; // get k of B(k,j)
GB_GET_A_k ; // get A(:,k)
if (aknz == 0) continue ;
GB_GET_B_kj ; // bkj = B(k,j)
// scan A(:,k)
for (int64_t pA = pA_start ; pA < pA_end ; pA++)
{
GB_GET_A_ik_INDEX ; // get i of A(i,k)
int64_t hf = Hf [i] ;
if (hf < mark)
{
// C(i,j) = A(i,k) * B(k,j)
Hf [i] = mark1 ; // mark as seen
GB_MULT_A_ik_B_kj ; // t = A(i,k)*B(k,j)
GB_HX_WRITE (i, t) ; // Hx [i] = t
}
else if (hf == mark1)
{
// C(i,j) += A(i,k) * B(k,j)
GB_MULT_A_ik_B_kj ; // t =A(i,k)*B(k,j)
GB_HX_UPDATE (i, t) ; // Hx [i] += t
}
}
}
GB_GATHER_ALL_C_j(mark1) ; // gather into C(:,j)
}
else
{
//------------------------------------------------------------------
// C(:,j) is very sparse
//------------------------------------------------------------------
for ( ; pB < pB_end ; pB++) // scan B(:,j)
{
GB_GET_B_kj_INDEX ; // get k of B(k,j)
GB_GET_A_k ; // get A(:,k)
if (aknz == 0) continue ;
GB_GET_B_kj ; // bkj = B(k,j)
// scan A(:,k)
for (int64_t pA = pA_start ; pA < pA_end ; pA++)
{
GB_GET_A_ik_INDEX ; // get i of A(i,k)
int64_t hf = Hf [i] ;
if (hf < mark)
{
// C(i,j) = A(i,k) * B(k,j)
Hf [i] = mark1 ; // mark as seen
GB_MULT_A_ik_B_kj ; // t = A(i,k)*B(k,j)
GB_HX_WRITE (i, t) ; // Hx [i] = t
Ci [pC++] = i ; // create C(:,j) pattern
}
else if (hf == mark1)
{
// C(i,j) += A(i,k) * B(k,j)
GB_MULT_A_ik_B_kj ; // t =A(i,k)*B(k,j)
GB_HX_UPDATE (i, t) ; // Hx [i] += t
}
}
}
GB_SORT_AND_GATHER_C_j ; // gather into C(:,j)
}
}
}
|
