|
Chris@42
|
1 /*
|
|
Chris@42
|
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
|
|
Chris@42
|
3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
|
|
Chris@42
|
4 *
|
|
Chris@42
|
5 * This program is free software; you can redistribute it and/or modify
|
|
Chris@42
|
6 * it under the terms of the GNU General Public License as published by
|
|
Chris@42
|
7 * the Free Software Foundation; either version 2 of the License, or
|
|
Chris@42
|
8 * (at your option) any later version.
|
|
Chris@42
|
9 *
|
|
Chris@42
|
10 * This program is distributed in the hope that it will be useful,
|
|
Chris@42
|
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
Chris@42
|
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
Chris@42
|
13 * GNU General Public License for more details.
|
|
Chris@42
|
14 *
|
|
Chris@42
|
15 * You should have received a copy of the GNU General Public License
|
|
Chris@42
|
16 * along with this program; if not, write to the Free Software
|
|
Chris@42
|
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
Chris@42
|
18 *
|
|
Chris@42
|
19 */
|
|
Chris@42
|
20
|
|
Chris@42
|
21 #include "ifftw-mpi.h"
|
|
Chris@42
|
22
|
|
Chris@42
|
23 /* r2c and c2r transforms. The sz dtensor, as usual, gives the size
|
|
Chris@42
|
24 of the "logical" complex array. For the last dimension N, however,
|
|
Chris@42
|
25 only N/2+1 complex numbers are stored for the complex data. Moreover,
|
|
Chris@42
|
26 for the real data, the last dimension is *always* padded to a size
|
|
Chris@42
|
27 2*(N/2+1). (Contrast this with the serial API, where there is only
|
|
Chris@42
|
28 padding for in-place plans.) */
|
|
Chris@42
|
29
|
|
Chris@42
|
30 /* problem.c: */
|
|
Chris@42
|
31 typedef struct {
|
|
Chris@42
|
32 problem super;
|
|
Chris@42
|
33 dtensor *sz;
|
|
Chris@42
|
34 INT vn; /* vector length (vector stride 1) */
|
|
Chris@42
|
35 R *I, *O; /* contiguous interleaved arrays */
|
|
Chris@42
|
36
|
|
Chris@42
|
37 rdft_kind kind; /* assert(kind < DHT) */
|
|
Chris@42
|
38 unsigned flags; /* TRANSPOSED_IN/OUT meaningful for rnk>1 only
|
|
Chris@42
|
39 SCRAMBLED_IN/OUT meaningful for 1d transforms only */
|
|
Chris@42
|
40
|
|
Chris@42
|
41 MPI_Comm comm;
|
|
Chris@42
|
42 } problem_mpi_rdft2;
|
|
Chris@42
|
43
|
|
Chris@42
|
44 problem *XM(mkproblem_rdft2)(const dtensor *sz, INT vn,
|
|
Chris@42
|
45 R *I, R *O, MPI_Comm comm,
|
|
Chris@42
|
46 rdft_kind kind, unsigned flags);
|
|
Chris@42
|
47 problem *XM(mkproblem_rdft2_d)(dtensor *sz, INT vn,
|
|
Chris@42
|
48 R *I, R *O, MPI_Comm comm,
|
|
Chris@42
|
49 rdft_kind kind, unsigned flags);
|
|
Chris@42
|
50
|
|
Chris@42
|
51 /* solve.c: */
|
|
Chris@42
|
52 void XM(rdft2_solve)(const plan *ego_, const problem *p_);
|
|
Chris@42
|
53
|
|
Chris@42
|
54 /* plans have same operands as rdft plans, so just re-use */
|
|
Chris@42
|
55 typedef plan_rdft plan_mpi_rdft2;
|
|
Chris@42
|
56 #define MKPLAN_MPI_RDFT2(type, adt, apply) \
|
|
Chris@42
|
57 (type *)X(mkplan_rdft)(sizeof(type), adt, apply)
|
|
Chris@42
|
58
|
|
Chris@42
|
59 int XM(rdft2_serial_applicable)(const problem_mpi_rdft2 *p);
|
|
Chris@42
|
60
|
|
Chris@42
|
61 /* various solvers */
|
|
Chris@42
|
62 void XM(rdft2_rank_geq2_register)(planner *p);
|
|
Chris@42
|
63 void XM(rdft2_rank_geq2_transposed_register)(planner *p);
|
|
Chris@42
|
64 void XM(rdft2_serial_register)(planner *p);
|
