|
Chris@82
|
1 /*
|
|
Chris@82
|
2 * Copyright (c) 2003, 2007-14 Matteo Frigo
|
|
Chris@82
|
3 * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology
|
|
Chris@82
|
4 *
|
|
Chris@82
|
5 * This program is free software; you can redistribute it and/or modify
|
|
Chris@82
|
6 * it under the terms of the GNU General Public License as published by
|
|
Chris@82
|
7 * the Free Software Foundation; either version 2 of the License, or
|
|
Chris@82
|
8 * (at your option) any later version.
|
|
Chris@82
|
9 *
|
|
Chris@82
|
10 * This program is distributed in the hope that it will be useful,
|
|
Chris@82
|
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
Chris@82
|
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
Chris@82
|
13 * GNU General Public License for more details.
|
|
Chris@82
|
14 *
|
|
Chris@82
|
15 * You should have received a copy of the GNU General Public License
|
|
Chris@82
|
16 * along with this program; if not, write to the Free Software
|
|
Chris@82
|
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
Chris@82
|
18 *
|
|
Chris@82
|
19 */
|
|
Chris@82
|
20
|
|
Chris@82
|
21 #include "ifftw-mpi.h"
|
|
Chris@82
|
22
|
|
Chris@82
|
23 /* common functions for rearrangements of the data for the *-rank1-bigvec
|
|
Chris@82
|
24 solvers */
|
|
Chris@82
|
25
|
|
Chris@82
|
26 static int div_mult(INT b, INT a) {
|
|
Chris@82
|
27 return (a > b && a % b == 0);
|
|
Chris@82
|
28 }
|
|
Chris@82
|
29 static int div_mult2(INT b, INT a, INT n) {
|
|
Chris@82
|
30 return (div_mult(b, a) && div_mult(n, b));
|
|
Chris@82
|
31 }
|
|
Chris@82
|
32
|
|
Chris@82
|
33 int XM(rearrange_applicable)(rearrangement rearrange,
|
|
Chris@82
|
34 ddim dim0, INT vn, int n_pes)
|
|
Chris@82
|
35 {
|
|
Chris@82
|
36 /* note: it is important that cases other than CONTIG be
|
|
Chris@82
|
37 applicable only when the resulting transpose dimension
|
|
Chris@82
|
38 is divisible by n_pes; otherwise, the allocation size
|
|
Chris@82
|
39 returned by the API will be incorrect */
|
|
Chris@82
|
40 return ((rearrange != DISCONTIG || div_mult(n_pes, vn))
|
|
Chris@82
|
41 && (rearrange != SQUARE_BEFORE
|
|
Chris@82
|
42 || div_mult2(dim0.b[IB], vn, n_pes))
|
|
Chris@82
|
43 && (rearrange != SQUARE_AFTER
|
|
Chris@82
|
44 || (dim0.b[IB] != dim0.b[OB]
|
|
Chris@82
|
45 && div_mult2(dim0.b[OB], vn, n_pes)))
|
|
Chris@82
|
46 && (rearrange != SQUARE_MIDDLE
|
|
Chris@82
|
47 || div_mult(dim0.n * n_pes, vn)));
|
|
Chris@82
|
48 }
|
|
Chris@82
|
49
|
|
Chris@82
|
50 INT XM(rearrange_ny)(rearrangement rearrange, ddim dim0, INT vn, int n_pes)
|
|
Chris@82
|
51 {
|
|
Chris@82
|
52 switch (rearrange) {
|
|
Chris@82
|
53 case CONTIG:
|
|
Chris@82
|
54 return vn;
|
|
Chris@82
|
55 case DISCONTIG:
|
|
Chris@82
|
56 return n_pes;
|
|
Chris@82
|
57 case SQUARE_BEFORE:
|
|
Chris@82
|
58 return dim0.b[IB];
|
|
Chris@82
|
59 case SQUARE_AFTER:
|
|
Chris@82
|
60 return dim0.b[OB];
|
|
Chris@82
|
61 case SQUARE_MIDDLE:
|
|
Chris@82
|
62 return dim0.n * n_pes;
|
|
Chris@82
|
63 }
|
|
Chris@82
|
64 return 0;
|
|
Chris@82
|
65 }
|
