root / src / pgesv / HPL_perm.c @ 8
Historique | Voir | Annoter | Télécharger (5,84 ko)
| 1 | 1 | equemene | /*
|
|---|---|---|---|
| 2 | 1 | equemene | * -- High Performance Computing Linpack Benchmark (HPL)
|
| 3 | 1 | equemene | * HPL - 2.0 - September 10, 2008
|
| 4 | 1 | equemene | * Antoine P. Petitet
|
| 5 | 1 | equemene | * University of Tennessee, Knoxville
|
| 6 | 1 | equemene | * Innovative Computing Laboratory
|
| 7 | 1 | equemene | * (C) Copyright 2000-2008 All Rights Reserved
|
| 8 | 1 | equemene | *
|
| 9 | 1 | equemene | * -- Copyright notice and Licensing terms:
|
| 10 | 1 | equemene | *
|
| 11 | 1 | equemene | * Redistribution and use in source and binary forms, with or without
|
| 12 | 1 | equemene | * modification, are permitted provided that the following conditions
|
| 13 | 1 | equemene | * are met:
|
| 14 | 1 | equemene | *
|
| 15 | 1 | equemene | * 1. Redistributions of source code must retain the above copyright
|
| 16 | 1 | equemene | * notice, this list of conditions and the following disclaimer.
|
| 17 | 1 | equemene | *
|
| 18 | 1 | equemene | * 2. Redistributions in binary form must reproduce the above copyright
|
| 19 | 1 | equemene | * notice, this list of conditions, and the following disclaimer in the
|
| 20 | 1 | equemene | * documentation and/or other materials provided with the distribution.
|
| 21 | 1 | equemene | *
|
| 22 | 1 | equemene | * 3. All advertising materials mentioning features or use of this
|
| 23 | 1 | equemene | * software must display the following acknowledgement:
|
| 24 | 1 | equemene | * This product includes software developed at the University of
|
| 25 | 1 | equemene | * Tennessee, Knoxville, Innovative Computing Laboratory.
|
| 26 | 1 | equemene | *
|
| 27 | 1 | equemene | * 4. The name of the University, the name of the Laboratory, or the
|
| 28 | 1 | equemene | * names of its contributors may not be used to endorse or promote
|
| 29 | 1 | equemene | * products derived from this software without specific written
|
| 30 | 1 | equemene | * permission.
|
| 31 | 1 | equemene | *
|
| 32 | 1 | equemene | * -- Disclaimer:
|
| 33 | 1 | equemene | *
|
| 34 | 1 | equemene | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
| 35 | 1 | equemene | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
| 36 | 1 | equemene | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
| 37 | 1 | equemene | * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY
|
| 38 | 1 | equemene | * OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
|
| 39 | 1 | equemene | * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
|
| 40 | 1 | equemene | * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
|
| 41 | 1 | equemene | * DATA OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
|
| 42 | 1 | equemene | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
|
| 43 | 1 | equemene | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
|
| 44 | 1 | equemene | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
| 45 | 1 | equemene | * ---------------------------------------------------------------------
|
| 46 | 1 | equemene | */
|
| 47 | 1 | equemene | /*
|
| 48 | 1 | equemene | * Include files
|
| 49 | 1 | equemene | */
|
| 50 | 1 | equemene | #include "hpl.h" |
| 51 | 1 | equemene | |
| 52 | 1 | equemene | #ifdef STDC_HEADERS
|
| 53 | 1 | equemene | void HPL_perm
|
| 54 | 1 | equemene | ( |
| 55 | 1 | equemene | const int N, |
| 56 | 1 | equemene | int * LINDXA,
|
| 57 | 1 | equemene | int * LINDXAU,
|
| 58 | 1 | equemene | int * IWORK
|
| 59 | 1 | equemene | ) |
| 60 | 1 | equemene | #else
|
| 61 | 1 | equemene | void HPL_perm
|
| 62 | 1 | equemene | ( N, LINDXA, LINDXAU, IWORK ) |
| 63 | 1 | equemene | const int N; |
| 64 | 1 | equemene | int * LINDXA;
|
| 65 | 1 | equemene | int * LINDXAU;
|
| 66 | 1 | equemene | int * IWORK;
|
| 67 | 1 | equemene | #endif
|
| 68 | 1 | equemene | {
|
| 69 | 1 | equemene | /*
|
| 70 | 1 | equemene | * Purpose
|
| 71 | 1 | equemene | * =======
|
| 72 | 1 | equemene | *
|
| 73 | 1 | equemene | * HPL_perm combines two index arrays and generate the corresponding
|
| 74 | 1 | equemene | * permutation. First, this function computes the inverse of LINDXA, and
|
| 75 | 1 | equemene | * then combine it with LINDXAU. Second, in order to be able to perform
|
| 76 | 1 | equemene | * the permutation in place, LINDXAU is overwritten by the sequence of
|
| 77 | 1 | equemene | * permutation producing the same result. What we ultimately want to
|
| 78 | 1 | equemene | * achieve is: U[LINDXAU[i]] := U[LINDXA[i]] for i in [0..N). After the
|
| 79 | 1 | equemene | * call to this function, this in place permutation can be performed by
|
| 80 | 1 | equemene | * for i in [0..N) swap U[i] with U[LINDXAU[i]].
|
| 81 | 1 | equemene | *
|
| 82 | 1 | equemene | * Arguments
|
| 83 | 1 | equemene | * =========
|
| 84 | 1 | equemene | *
|
| 85 | 1 | equemene | * N (global input) const int
|
| 86 | 1 | equemene | * On entry, N specifies the length of the arrays LINDXA and
|
| 87 | 1 | equemene | * LINDXAU. N should be at least zero.
|
| 88 | 1 | equemene | *
|
| 89 | 1 | equemene | * LINDXA (global input/output) int *
|
| 90 | 1 | equemene | * On entry, LINDXA is an array of dimension N containing the
|
| 91 | 1 | equemene | * source indexes. On exit, LINDXA contains the combined index
|
| 92 | 1 | equemene | * array.
|
| 93 | 1 | equemene | *
|
| 94 | 1 | equemene | * LINDXAU (global input/output) int *
|
| 95 | 1 | equemene | * On entry, LINDXAU is an array of dimension N containing the
|
| 96 | 1 | equemene | * target indexes. On exit, LINDXAU contains the sequence of
|
| 97 | 1 | equemene | * permutation, that should be applied in increasing order to
|
| 98 | 1 | equemene | * permute the underlying array U in place.
|
| 99 | 1 | equemene | *
|
| 100 | 1 | equemene | * IWORK (workspace) int *
|
| 101 | 1 | equemene | * On entry, IWORK is a workarray of dimension N.
|
| 102 | 1 | equemene | *
|
| 103 | 1 | equemene | * ---------------------------------------------------------------------
|
| 104 | 1 | equemene | */
|
| 105 | 1 | equemene | /*
|
| 106 | 1 | equemene | * .. Local Variables ..
|
| 107 | 1 | equemene | */
|
| 108 | 1 | equemene | int i, j, k, fndd;
|
| 109 | 1 | equemene | /* ..
|
| 110 | 1 | equemene | * .. Executable Statements ..
|
| 111 | 1 | equemene | */
|
| 112 | 1 | equemene | /*
|
| 113 | 1 | equemene | * Inverse LINDXA - combine LINDXA and LINDXAU - Initialize IWORK
|
| 114 | 1 | equemene | */
|
| 115 | 1 | equemene | for( i = 0; i < N; i++ ) { IWORK[LINDXA[i]] = i; } |
| 116 | 1 | equemene | for( i = 0; i < N; i++ ) { LINDXA[i] = LINDXAU[IWORK[i]]; IWORK[i] = i; } |
| 117 | 1 | equemene | |
| 118 | 1 | equemene | for( i = 0; i < N; i++ ) |
| 119 | 1 | equemene | {
|
| 120 | 1 | equemene | /* search LINDXA such that LINDXA[j] == i */
|
| 121 | 1 | equemene | j = 0; do { fndd = ( LINDXA[j] == i ); j++; } while( !fndd ); j--; |
| 122 | 1 | equemene | /* search IWORK such that IWORK[k] == j */
|
| 123 | 1 | equemene | k = 0; do { fndd = ( IWORK[k] == j ); k++; } while( !fndd ); k--; |
| 124 | 1 | equemene | /* swap IWORK[i] and IWORK[k]; LINDXAU[i] = k */
|
| 125 | 1 | equemene | j = IWORK[i]; IWORK[i] = IWORK[k]; IWORK[k] = j; |
| 126 | 1 | equemene | LINDXAU[i] = k; |
| 127 | 1 | equemene | } |
| 128 | 1 | equemene | /*
|
| 129 | 1 | equemene | * End of HPL_perm
|
| 130 | 1 | equemene | */
|
| 131 | 1 | equemene | } |