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
|
// Copyright (c) 2017-2023, University of Tennessee. All rights reserved.
// SPDX-License-Identifier: BSD-3-Clause
// This program is free software: you can redistribute it and/or modify it under
// the terms of the BSD 3-Clause license. See the accompanying LICENSE file.
#include "test.hh"
#include "lapack.hh"
#include "lapack/flops.hh"
#include "print_matrix.hh"
#include "error.hh"
#include "lapacke_wrappers.hh"
#include <vector>
// -----------------------------------------------------------------------------
template< typename scalar_t >
void test_ungrq_work( Params& params, bool run )
{
using real_t = blas::real_type< scalar_t >;
// get & mark input values
int64_t m = params.dim.m();
int64_t n = params.dim.n();
int64_t k = params.dim.k();
int64_t align = params.align();
params.matrix.mark();
real_t eps = std::numeric_limits< real_t >::epsilon();
real_t tol = params.tol() * eps;
// mark non-standard output values
params.ortho();
params.gflops();
params.ref_time();
params.ref_gflops();
params.msg();
if (! run)
return;
// skip invalid sizes
if (! (n >= m && m >= k)) {
params.msg() = "skipping: requires n >= m and m >= k";
return;
}
// ---------- setup
// For error check, R needs to be k-by-m;
// for ortho check, R needs to be m-by-m to store Q*Q^H.
// zrqt02.f has R to be m-by-n, which is bigger than needed.
int64_t lda = roundup( blas::max( 1, m ), align );
int64_t ldr = lda;
size_t size_A = (size_t) lda * n;
size_t size_tau = (size_t) (k);
size_t size_R = (size_t) lda * m;
std::vector< scalar_t > A_tst( size_A );
std::vector< scalar_t > A_ref( size_A );
std::vector< scalar_t > tau( size_tau );
std::vector< scalar_t > R( size_R );
lapack::generate_matrix( params.matrix, m, n, &A_tst[0], lda );
int64_t idist = 1;
int64_t iseed[4] = { 0, 1, 2, 3 };
lapack::larnv( idist, iseed, tau.size(), &tau[0] );
A_ref = A_tst;
// ---------- factor matrix
// A is m-by-n, but factor bottom k-by-n portion.
// (Compare with QR, which would factor the left m-by-k portion.)
int64_t info_rqf = lapack::gerqf( k, n, &A_tst[(m-k)], lda, &tau[0] );
if (info_rqf != 0) {
fprintf( stderr, "lapack::gerqf returned error %lld\n", llong( info_rqf ) );
}
// Copy R(m-k+1:m, 1:n) from factored A for check.
lapack::laset( lapack::MatrixType::General, k, m, 0.0, 0.0,
&R[(m-k)], ldr );
lapack::lacpy( lapack::MatrixType::Upper, k, k,
&A_tst[(m-k) + (n-k)*lda], lda,
&R[(m-k) + (m-k)*ldr], ldr );
// ---------- run test
testsweeper::flush_cache( params.cache() );
double time = testsweeper::get_wtime();
int64_t info_tst = lapack::ungrq( m, n, k, &A_tst[0], lda, &tau[0] );
time = testsweeper::get_wtime() - time;
if (info_tst != 0) {
fprintf( stderr, "lapack::ungrq returned error %lld\n", llong( info_tst ) );
}
params.time() = time;
double gflop = lapack::Gflop< scalar_t >::ungrq( m, n, k );
params.gflops() = gflop / time;
if (params.check() == 'y') {
// ---------- check error
// comparing to ref. solution doesn't work
// Following lapack/TESTING/LIN/zrqt02.f
// Note: n >= m; m >= k; lda >= m
// Compute R(m-k+1:m, 1:n) - A(m-k+1:m, 1:n) * Q(n-m+1:n, 1:n)^H
blas::gemm( blas::Layout::ColMajor,
blas::Op::NoTrans, blas::Op::ConjTrans, k, m, n,
-1.0, &A_ref[(m-k)], lda,
&A_tst[0], lda,
1.0, &R[(m-k)], ldr );
// Compute norm( R - Q^H*A ) / ( M * norm(A) * EPS ) .
real_t Anorm = lapack::lange( lapack::Norm::One, k, n, &A_ref[(m-k)], lda );
real_t error = lapack::lange( lapack::Norm::One, k, m, &R[(m-k)], ldr );
if (Anorm > 0)
error /= n * Anorm;
// Compute I - Q*Q^H
// Note Q has orthonormal rows (instead of cols).
lapack::laset( lapack::MatrixType::General, m, m, 0.0, 1.0, &R[0], ldr );
blas::herk( blas::Layout::ColMajor,
blas::Uplo::Upper, blas::Op::NoTrans, m, n,
-1.0, &A_tst[0], lda,
1.0, &R[0], ldr );
// Compute norm( I - Q*Q^H ) / ( N * EPS )
real_t ortho = lapack::lansy( lapack::Norm::One, lapack::Uplo::Upper, m, &R[0], ldr );
ortho /= n;
params.error() = error;
params.ortho() = ortho;
params.okay() = (error < tol) && (ortho < tol);
}
if (params.ref() == 'y') {
// ---------- run reference
testsweeper::flush_cache( params.cache() );
time = testsweeper::get_wtime();
int64_t info_ref = LAPACKE_ungrq( m, n, k, &A_ref[0], lda, &tau[0] );
time = testsweeper::get_wtime() - time;
if (info_ref != 0) {
fprintf( stderr, "LAPACKE_ungrq returned error %lld\n", llong( info_ref ) );
}
params.ref_time() = time;
params.ref_gflops() = gflop / time;
}
}
// -----------------------------------------------------------------------------
void test_ungrq( Params& params, bool run )
{
switch (params.datatype()) {
case testsweeper::DataType::Single:
test_ungrq_work< float >( params, run );
break;
case testsweeper::DataType::Double:
test_ungrq_work< double >( params, run );
break;
case testsweeper::DataType::SingleComplex:
test_ungrq_work< std::complex<float> >( params, run );
break;
case testsweeper::DataType::DoubleComplex:
test_ungrq_work< std::complex<double> >( params, run );
break;
default:
throw std::runtime_error( "unknown datatype" );
break;
}
}
|
