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
|
//------------------------------------------------------------------------------
// GB_AxB_meta_adotb_control: determine method for computing C=A'*B
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2022, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
#include "GB_mxm.h"
void GB_AxB_meta_adotb_control
(
// output:
int *axb_method,
// input:
const GrB_Matrix C_in,
const GrB_Matrix M,
bool Mask_comp,
const GrB_Matrix A,
const GrB_Matrix B,
const GrB_BinaryOp accum,
const GrB_Semiring semiring, // semiring that defines C=A*B
bool flipxy,
bool can_do_in_place,
bool allow_scale,
bool B_is_diagonal,
GrB_Desc_Value AxB_method,
GB_Context Context
)
{
// C=A'*B is being computed: use the dot product without computing A'
// or use the saxpy (Gustavson) method
// use saxpy by default, unless selecting other methods below
(*axb_method) = GB_USE_SAXPY ;
// If the mask is present, only entries for which M(i,j)=1 are
// computed, which makes this method very efficient when the mask is
// very sparse (triangle counting, for example). Each entry C(i,j) for
// which M(i,j)=1 is computed via a dot product, C(i,j) =
// A(:,i)'*B(:,j). If the mask is not present, the dot-product method
// is very slow in general, and thus the saxpy method is usually used
// instead.
if (allow_scale && M == NULL
&& !GB_IS_BITMAP (A) // TODO: A'*D colscale with A bitmap
&& B_is_diagonal)
{
// C = A'*D, col scale
(*axb_method) = GB_USE_COLSCALE ;
}
else if (allow_scale && M == NULL
&& !GB_IS_BITMAP (B) // TODO: D*B rowscale with B bitmap
&& GB_is_diagonal (A, Context))
{
// C = D*B, row scale
(*axb_method) = GB_USE_ROWSCALE ;
}
else if (AxB_method == GxB_DEFAULT)
{
// auto selection for A'*B
bool C_out_iso = false ; // ignored unless C can be done in-place
if (can_do_in_place && C_in != NULL)
{
// check if C will be iso on output (for dot4 control only).
// Ignored if dot4 C_in is not present or C cannot be
// computed in-place.
C_out_iso = GB_iso_AxB (NULL, A, B, A->vlen, semiring, flipxy,
false) ;
}
if (GB_AxB_dot4_control (C_out_iso, can_do_in_place ? C_in : NULL,
M, Mask_comp, accum, semiring))
{
// C+=A'*B can be done with dot4
(*axb_method) = GB_USE_DOT ;
}
else if (GB_AxB_dot3_control (M, Mask_comp))
{
// C<M>=A'*B uses the masked dot product method (dot3)
(*axb_method) = GB_USE_DOT ;
}
else if (GB_AxB_dot2_control (A, B, Context))
{
// C=A'*B or C<!M>=A'B* can efficiently use the dot2 method
(*axb_method) = GB_USE_DOT ;
}
}
else if (AxB_method == GxB_AxB_DOT)
{
// user selection for A'*B
(*axb_method) = GB_USE_DOT ;
}
}
|
