forked from FFTW/fftw3
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrdft-problem.c
148 lines (132 loc) · 4.32 KB
/
rdft-problem.c
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
/*
* Copyright (c) 2003, 2007 Matteo Frigo
* Copyright (c) 2003, 2007 Massachusetts Institute of Technology
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
*/
#include "mpi-rdft.h"
static void destroy(problem *ego_)
{
problem_mpi_rdft *ego = (problem_mpi_rdft *) ego_;
XM(dtensor_destroy)(ego->sz);
MPI_Comm_free(&ego->comm);
#if !defined(STRUCT_HACK_C99) && !defined(STRUCT_HACK_KR)
X(ifree0)(ego->kind);
#endif
X(ifree)(ego_);
}
static void hash(const problem *p_, md5 *m)
{
const problem_mpi_rdft *p = (const problem_mpi_rdft *) p_;
int i;
X(md5puts)(m, "mpi-dft");
X(md5int)(m, p->I == p->O);
X(md5int)(m, X(alignment_of)(p->I));
X(md5int)(m, X(alignment_of)(p->O));
XM(dtensor_md5)(m, p->sz);
X(md5INT)(m, p->vn);
for (i = 0; i < p->sz->rnk; ++i)
X(md5int)(m, p->kind[i]);
X(md5int)(m, p->flags);
MPI_Comm_size(p->comm, &i); X(md5int)(m, i);
}
static void print(const problem *ego_, printer *p)
{
const problem_mpi_rdft *ego = (const problem_mpi_rdft *) ego_;
int i;
p->print(p, "(mpi-rdft %d %d %d ",
ego->I == ego->O,
X(alignment_of)(ego->I),
X(alignment_of)(ego->O));
XM(dtensor_print)(ego->sz, p);
for (i = 0; i < ego->sz->rnk; ++i)
p->print(p, " %d", (int)ego->kind[i]);
p->print(p, " %D %d", ego->vn, ego->flags);
MPI_Comm_size(ego->comm, &i); p->print(p, " %d)", i);
}
static void zero(const problem *ego_)
{
const problem_mpi_rdft *ego = (const problem_mpi_rdft *) ego_;
R *I = ego->I;
INT i, N;
int my_pe;
MPI_Comm_rank(ego->comm, &my_pe);
N = ego->vn * XM(total_block)(ego->sz, IB, my_pe);
for (i = 0; i < N; ++i) I[i] = K(0.0);
}
static const problem_adt padt =
{
PROBLEM_MPI_RDFT,
hash,
zero,
print,
destroy
};
problem *XM(mkproblem_rdft)(const dtensor *sz, INT vn,
R *I, R *O,
MPI_Comm comm,
const rdft_kind *kind, unsigned flags)
{
problem_mpi_rdft *ego;
int i, rnk = sz->rnk;
int n_pes;
A(XM(dtensor_validp)(sz) && FINITE_RNK(sz->rnk));
MPI_Comm_size(comm, &n_pes);
A(n_pes >= XM(num_blocks_total)(sz, IB)
&& n_pes >= XM(num_blocks_total)(sz, OB));
A(vn >= 0);
#if defined(STRUCT_HACK_KR)
ego = (problem_mpi_rdft *) X(mkproblem)(sizeof(problem_mpi_rdft)
+ sizeof(rdft_kind)
* (rnk > 0 ? rnk - 1 : 0), &padt);
#elif defined(STRUCT_HACK_C99)
ego = (problem_mpi_rdft *) X(mkproblem)(sizeof(problem_mpi_rdft)
+ sizeof(rdft_kind) * rnk, &padt);
#else
ego = (problem_mpi_rdft *) X(mkproblem)(sizeof(problem_mpi_rdft), &padt);
ego->kind = (rdft_kind *) MALLOC(sizeof(rdft_kind) * rnk, PROBLEMS);
#endif
/* enforce pointer equality if untainted pointers are equal */
if (UNTAINT(I) == UNTAINT(O))
I = O = JOIN_TAINT(I, O);
ego->sz = XM(dtensor_canonical)(sz, 0);
ego->vn = vn;
ego->I = I;
ego->O = O;
for (i = 0; i< ego->sz->rnk; ++i)
ego->kind[i] = kind[i];
/* canonicalize: replace TRANSPOSED_IN with TRANSPOSED_OUT by
swapping the first two dimensions (for rnk > 1) */
if ((flags & TRANSPOSED_IN) && ego->sz->rnk > 1) {
ddim dim0 = ego->sz->dims[0];
ego->sz->dims[0] = ego->sz->dims[1];
ego->sz->dims[1] = dim0;
flags &= ~TRANSPOSED_IN;
flags ^= TRANSPOSED_OUT;
}
ego->flags = flags;
MPI_Comm_dup(comm, &ego->comm);
return &(ego->super);
}
problem *XM(mkproblem_rdft_d)(dtensor *sz, INT vn,
R *I, R *O,
MPI_Comm comm,
const rdft_kind *kind, unsigned flags)
{
problem *p = XM(mkproblem_rdft)(sz, vn, I, O, comm, kind, flags);
XM(dtensor_destroy)(sz);
return p;
}