1
/* ---------------------------------------------------------------------
3
* -- Automatically Tuned Linear Algebra Software (ATLAS)
4
* (C) Copyright 2000 All Rights Reserved
6
* -- ATLAS routine -- Version 3.2 -- December 25, 2000
8
* Author : Antoine P. Petitet
9
* Originally developed at the University of Tennessee,
10
* Innovative Computing Laboratory, Knoxville TN, 37996-1301, USA.
12
* ---------------------------------------------------------------------
14
* -- Copyright notice and Licensing terms:
16
* Redistribution and use in source and binary forms, with or without
17
* modification, are permitted provided that the following conditions
20
* 1. Redistributions of source code must retain the above copyright
21
* notice, this list of conditions and the following disclaimer.
22
* 2. Redistributions in binary form must reproduce the above copyright
23
* notice, this list of conditions, and the following disclaimer in
24
* the documentation and/or other materials provided with the distri-
26
* 3. The name of the University, the ATLAS group, or the names of its
27
* contributors may not be used to endorse or promote products deri-
28
* ved from this software without specific written permission.
32
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
33
* ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
34
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
35
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY
36
* OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPE-
37
* CIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
38
* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
39
* OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEO-
40
* RY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (IN-
41
* CLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
42
* THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
44
* ---------------------------------------------------------------------
49
#include "atlas_misc.h"
52
#include "atlas_level1.h"
53
#include "atlas_kernel2.h"
54
#include "atlas_reflvl2.h"
55
#include "atlas_lvl2.h"
57
void Mjoin( PATL, hbmvU )
72
* Mjoin( PATL, hbmvU ) performs the matrix-vector operation
74
* y := alpha * A * x + beta * y,
76
* where alpha and beta are scalars, x and y are n-element vectors and A
77
* is an n by n Hermitian band matrix, with k super-diagonals.
79
* This is a recursive version of the algorithm. For a more detailed
80
* description of the arguments of this function, see the reference im-
81
* plementation in the ATLAS/src/blas/reference directory.
83
* ---------------------------------------------------------------------
86
* .. Local Variables ..
88
void (*gbmvT)( const int, const int, const int,
89
const int, const SCALAR, const TYPE *, const int,
90
const TYPE *, const int, const SCALAR, TYPE *,
92
void (*gbmvN)( const int, const int, const int,
93
const int, const SCALAR, const TYPE *, const int,
94
const TYPE *, const int, const SCALAR, TYPE *,
101
const TYPE * beta0, one[2] = { ATL_rone, ATL_rzero };
102
const int lda2 = (LDA << 1);
104
int ia, ian, jb, kl, ku, ma, mb, na, nb;
106
* .. Executable Statements ..
109
ATL_GetPartSBMV( A, LDA, &mb, &nb );
113
if( SCALAR_IS_ZERO( beta0 ) )
115
gbmvN = Mjoin( PATL, gbmvN_a1_x1_b0_y1 );
116
gbmvT = Mjoin( PATL, gbmvC_a1_x1_b0_y1 );
118
else if( SCALAR_IS_ONE ( beta0 ) )
120
gbmvN = Mjoin( PATL, gbmvN_a1_x1_b1_y1 );
121
gbmvT = Mjoin( PATL, gbmvC_a1_x1_b1_y1 );
125
gbmvN = Mjoin( PATL, gbmvN_a1_x1_bX_y1 );
126
gbmvT = Mjoin( PATL, gbmvC_a1_x1_bX_y1 );
129
for( ian = ( ( N - 1 ) / nb ) * nb; ian >= 0; ian -= nb )
131
jb = N - ian; jb = Mmin( jb, nb );
135
ia = ian - K; ia = Mmax( ia, 0 ); ma = ian - ia; na = Mmin( jb, K );
136
kl = ma - 1; kl = Mmax( kl, 0 ); ku = K - 1 - kl; ku = Mmax( ku, 0 );
137
gbmvT( na, ma, kl, ku, one, A+ian*lda2, LDA, X+(ia SHIFT), 1, beta0,
139
gbmvN( ma, na, kl, ku, one, A+ian*lda2, LDA, X+(ian SHIFT), 1, beta0,
141
if( !( SCALAR_IS_ONE( beta0 ) ) )
144
Mjoin( PATL, scal )( jb-na, beta0, Y+((ian+na) SHIFT), 1 );
145
if( ia > 0 ) Mjoin( PATL, scal )( ia, beta0, Y, 1 );
149
Mjoin( PATL, refhbmvU )( jb, K, one, A+ian*lda2, LDA, X+(ian SHIFT), 1,
150
beta0, Y+(ian SHIFT), 1 );
151
gbmvN = Mjoin( PATL, gbmvN_a1_x1_b1_y1 );
152
gbmvT = Mjoin( PATL, gbmvC_a1_x1_b1_y1 );
156
* End of Mjoin( PATL, hbmvU )