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
|
//==============================================================================
// sfmult_xA: Y = X*A, unblocked
//==============================================================================
// SFMULT, Copyright (c) 2009, Timothy A Davis. All Rights Reserved.
// SPDX-License-Identifier: BSD-3-clause
// This kernel is unique; it operates on all of X and Y, not just a few rows or
// columns. It is used only when A is very sparse and X is large since it can
// be very slow otherwise.
#include "sfmult.h"
void sfmult_xA // y = (A'*x')' = x*A, x is k-by-m, and y is k-by-n
(
// --- outputs, not initialized on input
double *Yx, // k-by-n
double *Yz, // k-by-n if Y is complex (TO DO)
// --- inputs, not modified
const Int *Ap, // size n+1 column pointers
const Int *Ai, // size nz = Ap[n] row indices
const double *Ax, // size nz values
const double *Az, // size nz imaginary values if A is complex (TO DO)
Int m, // A is m-by-n
Int n,
const double *Xx, // k-by-m
const double *Xz, // k-by-m if X complex (TO DO)
int ac, // true: use conj(A), otherwise use A (TO DO)
int xc, // true: use conj(X), otherwise use X (TO DO)
int yc // true: compute conj(Y), otherwise compute Y (TO DO)
, Int k
)
{
double a ;
const double *xx, *xz ;
Int p, pend, j, i, k1 ;
p = 0 ;
for (j = 0 ; j < n ; j++)
{
pend = Ap [j+1] ;
for (k1 = 0 ; k1 < k ; k1++)
{
Yx [k1] = 0 ;
}
for ( ; p < pend ; p++)
{
i = Ai [p] ;
a = Ax [p] ;
xx = Xx + i*k ;
xz = Xz + i*k ;
k1 = k % 4 ;
switch (k1)
{
case 3: Yx [2] += a * xx [2] ;
case 2: Yx [1] += a * xx [1] ;
case 1: Yx [0] += a * xx [0] ;
case 0: ;
}
for ( ; k1 < k ; k1 += 4)
{
Yx [k1 ] += a * xx [k1 ] ;
Yx [k1+1] += a * xx [k1+1] ;
Yx [k1+2] += a * xx [k1+2] ;
Yx [k1+3] += a * xx [k1+3] ;
}
}
Yx += k ;
Yz += k ;
}
}
|
