|
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 /* tproblem.c: */
|
|
Chris@42
|
24 typedef struct {
|
|
Chris@42
|
25 problem super;
|
|
Chris@42
|
26 INT vn; /* vector length (vector stride 1) */
|
|
Chris@42
|
27 INT nx, ny; /* nx x ny transposed to ny x nx */
|
|
Chris@42
|
28 R *I, *O; /* contiguous real arrays (both same size!) */
|
|
Chris@42
|
29
|
|
Chris@42
|
30 unsigned flags; /* TRANSPOSED_IN: input is *locally* transposed
|
|
Chris@42
|
31 TRANSPOSED_OUT: output is *locally* transposed */
|
|
Chris@42
|
32
|
|
Chris@42
|
33 INT block, tblock; /* block size, slab decomposition;
|
|
Chris@42
|
34 tblock is for transposed blocks on output */
|
|
Chris@42
|
35
|
|
Chris@42
|
36 MPI_Comm comm;
|
|
Chris@42
|
37 } problem_mpi_transpose;
|
|
Chris@42
|
38
|
|
Chris@42
|
39 problem *XM(mkproblem_transpose)(INT nx, INT ny, INT vn,
|
|
Chris@42
|
40 R *I, R *O,
|
|
Chris@42
|
41 INT block, INT tblock,
|
|
Chris@42
|
42 MPI_Comm comm,
|
|
Chris@42
|
43 unsigned flags);
|
|
Chris@42
|
44
|
|
Chris@42
|
45 /* tsolve.c: */
|
|
Chris@42
|
46 void XM(transpose_solve)(const plan *ego_, const problem *p_);
|
|
Chris@42
|
47
|
|
Chris@42
|
48 /* plans have same operands as rdft plans, so just re-use */
|
|
Chris@42
|
49 typedef plan_rdft plan_mpi_transpose;
|
|
Chris@42
|
50 #define MKPLAN_MPI_TRANSPOSE(type, adt, apply) \
|
|
Chris@42
|
51 (type *)X(mkplan_rdft)(sizeof(type), adt, apply)
|
|
Chris@42
|
52
|
|
Chris@42
|
53 /* transpose-pairwise.c: */
|
|
Chris@42
|
54 int XM(mkplans_posttranspose)(const problem_mpi_transpose *p, planner *plnr,
|
|
Chris@42
|
55 R *I, R *O, int my_pe,
|
|
Chris@42
|
56 plan **cld2, plan **cld2rest, plan **cld3,
|
|
Chris@42
|
57 INT *rest_Ioff, INT *rest_Ooff);
|
|
Chris@42
|
58 /* various solvers */
|
|
Chris@42
|
59 void XM(transpose_pairwise_register)(planner *p);
|
|
Chris@42
|
60 void XM(transpose_alltoall_register)(planner *p);
|
|
Chris@42
|
61 void XM(transpose_recurse_register)(planner *p);
|
