File: test_pprfs.cc

package info (click to toggle)
lapackpp 2024.10.26-1
  • links: PTS, VCS
  • area: main
  • in suites: forky, sid, trixie
  • size: 6,500 kB
  • sloc: cpp: 80,181; ansic: 27,660; python: 4,838; xml: 182; perl: 99; makefile: 53; sh: 23
file content (154 lines) | stat: -rw-r--r-- 5,113 bytes parent folder | download
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
// 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_pprfs_work( Params& params, bool run )
{
    using real_t = blas::real_type< scalar_t >;

    // get & mark input values
    lapack::Uplo uplo = params.uplo();
    int64_t n = params.dim.n();
    int64_t nrhs = params.nrhs();
    int64_t align = params.align();

    real_t eps = std::numeric_limits< real_t >::epsilon();

    // mark non-standard output values
    params.ref_time();
    //params.ref_gflops();
    //params.gflops();

    if (! run)
        return;

    // ---------- setup
    int64_t ldb = roundup( blas::max( 1, n ), align );
    int64_t ldx = roundup( blas::max( 1, n ), align );
    size_t size_AP = (size_t) (n*(n+1)/2);
    size_t size_AFP = (size_t) (n*(n+1)/2);
    size_t size_B = (size_t) ldb * nrhs;
    size_t size_X = (size_t) ldx * nrhs;
    size_t size_ferr = (size_t) (nrhs);
    size_t size_berr = (size_t) (nrhs);

    std::vector< scalar_t > AP( size_AP );
    std::vector< scalar_t > AFP( size_AFP );
    std::vector< scalar_t > B( size_B );
    std::vector< scalar_t > X_tst( size_X );
    std::vector< scalar_t > X_ref( size_X );
    std::vector< real_t > ferr_tst( size_ferr );
    std::vector< real_t > ferr_ref( size_ferr );
    std::vector< real_t > berr_tst( size_berr );
    std::vector< real_t > berr_ref( size_berr );

    int64_t idist = 1;
    int64_t iseed[4] = { 0, 1, 2, 3 };
    lapack::larnv( idist, iseed, AP.size(), &AP[0] );
    lapack::larnv( idist, iseed, AFP.size(), &AFP[0] );
    lapack::larnv( idist, iseed, B.size(), &B[0] );
    lapack::larnv( idist, iseed, X_tst.size(), &X_tst[0] );
    X_ref = X_tst;

    // diagonally dominant -> positive definite
    if (uplo == lapack::Uplo::Upper) {
        for (int64_t i = 0; i < n; ++i) {
            AP[ i + 0.5*(i+1)*i ] += n;
        }
    }
    else { // lower
        for (int64_t i = 0; i < n; ++i) {
            AP[ i + n*i - 0.5*i*(i+1) ] += n;
        }
    }

    AFP = AP;

    // factor AFP
    int64_t info = lapack::pptrf( uplo, n, &AFP[0] );
    if (info != 0) {
        fprintf( stderr, "lapack::pptrf returned error %lld\n", llong( info ) );
    }

    // compute solution
    info = lapack::pptrs( uplo, n, nrhs, &AFP[0], &B[0], ldb );
    if (info != 0) {
        fprintf( stderr, "lapack::pptrs returned error %lld\n", llong( info ) );
    }

    // ---------- run test
    testsweeper::flush_cache( params.cache() );
    double time = testsweeper::get_wtime();
    int64_t info_tst = lapack::pprfs( uplo, n, nrhs, &AP[0], &AFP[0], &B[0], ldb, &X_tst[0], ldx, &ferr_tst[0], &berr_tst[0] );
    time = testsweeper::get_wtime() - time;
    if (info_tst != 0) {
        fprintf( stderr, "lapack::pprfs returned error %lld\n", llong( info_tst ) );
    }

    params.time() = time;
    //double gflop = lapack::Gflop< scalar_t >::pprfs( n, nrhs );
    //params.gflops() = gflop / time;

    if (params.ref() == 'y' || params.check() == 'y') {
        // ---------- run reference
        testsweeper::flush_cache( params.cache() );
        time = testsweeper::get_wtime();
        int64_t info_ref = LAPACKE_pprfs( to_char( uplo ), n, nrhs, &AP[0], &AFP[0], &B[0], ldb, &X_ref[0], ldx, &ferr_ref[0], &berr_ref[0] );
        time = testsweeper::get_wtime() - time;
        if (info_ref != 0) {
            fprintf( stderr, "LAPACKE_pprfs returned error %lld\n", llong( info_ref ) );
        }

        params.ref_time() = time;
        //params.ref_gflops() = gflop / time;

        // ---------- check error compared to reference
        real_t error = 0;
        if (info_tst != info_ref) {
            error = 1;
        }
        error += abs_error( X_tst, X_ref );
        error += abs_error( ferr_tst, ferr_ref );
        error += abs_error( berr_tst, berr_ref );
        params.error() = error;
        params.okay() = (error < 3*eps);
    }
}

// -----------------------------------------------------------------------------
void test_pprfs( Params& params, bool run )
{
    switch (params.datatype()) {
        case testsweeper::DataType::Single:
            test_pprfs_work< float >( params, run );
            break;

        case testsweeper::DataType::Double:
            test_pprfs_work< double >( params, run );
            break;

        case testsweeper::DataType::SingleComplex:
            test_pprfs_work< std::complex<float> >( params, run );
            break;

        case testsweeper::DataType::DoubleComplex:
            test_pprfs_work< std::complex<double> >( params, run );
            break;

        default:
            throw std::runtime_error( "unknown datatype" );
            break;
    }
}