/src/fftw3/dft/dftw-directsq.c
Line | Count | Source (jump to first uncovered line) |
1 | | /* |
2 | | * Copyright (c) 2003, 2007-14 Matteo Frigo |
3 | | * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology |
4 | | * |
5 | | * This program is free software; you can redistribute it and/or modify |
6 | | * it under the terms of the GNU General Public License as published by |
7 | | * the Free Software Foundation; either version 2 of the License, or |
8 | | * (at your option) any later version. |
9 | | * |
10 | | * This program is distributed in the hope that it will be useful, |
11 | | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
12 | | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
13 | | * GNU General Public License for more details. |
14 | | * |
15 | | * You should have received a copy of the GNU General Public License |
16 | | * along with this program; if not, write to the Free Software |
17 | | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
18 | | * |
19 | | */ |
20 | | |
21 | | |
22 | | #include "dft/ct.h" |
23 | | |
24 | | typedef struct { |
25 | | ct_solver super; |
26 | | const ct_desc *desc; |
27 | | kdftwsq k; |
28 | | } S; |
29 | | |
30 | | typedef struct { |
31 | | plan_dftw super; |
32 | | kdftwsq k; |
33 | | INT r; |
34 | | stride rs, vs; |
35 | | INT m, ms, v, mb, me; |
36 | | twid *td; |
37 | | const S *slv; |
38 | | } P; |
39 | | |
40 | | |
41 | | static void apply(const plan *ego_, R *rio, R *iio) |
42 | 14 | { |
43 | 14 | const P *ego = (const P *) ego_; |
44 | 14 | INT mb = ego->mb, ms = ego->ms; |
45 | 14 | ego->k(rio + mb*ms, iio + mb*ms, ego->td->W, ego->rs, ego->vs, |
46 | 14 | mb, ego->me, ms); |
47 | 14 | } |
48 | | |
49 | | static void awake(plan *ego_, enum wakefulness wakefulness) |
50 | 8 | { |
51 | 8 | P *ego = (P *) ego_; |
52 | | |
53 | 8 | X(twiddle_awake)(wakefulness, &ego->td, ego->slv->desc->tw, |
54 | 8 | ego->r * ego->m, ego->r, ego->m); |
55 | 8 | } |
56 | | |
57 | | static void destroy(plan *ego_) |
58 | 22 | { |
59 | 22 | P *ego = (P *) ego_; |
60 | 22 | X(stride_destroy)(ego->rs); |
61 | 22 | X(stride_destroy)(ego->vs); |
62 | 22 | } |
63 | | |
64 | | static void print(const plan *ego_, printer *p) |
65 | 0 | { |
66 | 0 | const P *ego = (const P *) ego_; |
67 | 0 | const S *slv = ego->slv; |
68 | 0 | const ct_desc *e = slv->desc; |
69 | |
|
70 | 0 | p->print(p, "(dftw-directsq-%D/%D%v \"%s\")", |
71 | 0 | ego->r, X(twiddle_length)(ego->r, e->tw), ego->v, e->nam); |
72 | 0 | } |
73 | | |
74 | | static int applicable(const S *ego, |
75 | | INT r, INT irs, INT ors, |
76 | | INT m, INT ms, |
77 | | INT v, INT ivs, INT ovs, |
78 | | INT mb, INT me, |
79 | | R *rio, R *iio, |
80 | | const planner *plnr) |
81 | 22 | { |
82 | 22 | const ct_desc *e = ego->desc; |
83 | 22 | UNUSED(v); |
84 | | |
85 | 22 | return ( |
86 | 22 | 1 |
87 | 22 | && r == e->radix |
88 | | |
89 | | /* transpose r, v */ |
90 | 22 | && r == v |
91 | 22 | && irs == ovs |
92 | 22 | && ivs == ors |
93 | | |
94 | | /* check for alignment/vector length restrictions */ |
95 | 22 | && e->genus->okp(e, rio, iio, irs, ivs, m, mb, me, ms, plnr) |
96 | | |
97 | 22 | ); |
98 | 22 | } |
99 | | |
100 | | static plan *mkcldw(const ct_solver *ego_, |
101 | | INT r, INT irs, INT ors, |
102 | | INT m, INT ms, |
103 | | INT v, INT ivs, INT ovs, |
104 | | INT mstart, INT mcount, |
105 | | R *rio, R *iio, |
106 | | planner *plnr) |
107 | 22 | { |
108 | 22 | const S *ego = (const S *) ego_; |
109 | 22 | P *pln; |
110 | 22 | const ct_desc *e = ego->desc; |
111 | | |
112 | 22 | static const plan_adt padt = { |
113 | 22 | 0, awake, print, destroy |
114 | 22 | }; |
115 | | |
116 | 22 | A(mstart >= 0 && mstart + mcount <= m); |
117 | 22 | if (!applicable(ego, |
118 | 22 | r, irs, ors, m, ms, v, ivs, ovs, mstart, mstart + mcount, |
119 | 22 | rio, iio, plnr)) |
120 | 0 | return (plan *)0; |
121 | | |
122 | 22 | pln = MKPLAN_DFTW(P, &padt, apply); |
123 | | |
124 | 22 | pln->k = ego->k; |
125 | 22 | pln->rs = X(mkstride)(r, irs); |
126 | 22 | pln->vs = X(mkstride)(v, ivs); |
127 | 22 | pln->td = 0; |
128 | 22 | pln->r = r; |
129 | 22 | pln->m = m; |
130 | 22 | pln->ms = ms; |
131 | 22 | pln->v = v; |
132 | 22 | pln->mb = mstart; |
133 | 22 | pln->me = mstart + mcount; |
134 | 22 | pln->slv = ego; |
135 | | |
136 | 22 | X(ops_zero)(&pln->super.super.ops); |
137 | 22 | X(ops_madd2)(mcount/e->genus->vl, &e->ops, &pln->super.super.ops); |
138 | | |
139 | 22 | return &(pln->super.super); |
140 | 22 | } |
141 | | |
142 | | static void regone(planner *plnr, kdftwsq codelet, |
143 | | const ct_desc *desc, int dec) |
144 | 6 | { |
145 | 6 | S *slv = (S *)X(mksolver_ct)(sizeof(S), desc->radix, dec, mkcldw, 0); |
146 | 6 | slv->k = codelet; |
147 | 6 | slv->desc = desc; |
148 | 6 | REGISTER_SOLVER(plnr, &(slv->super.super)); |
149 | 6 | if (X(mksolver_ct_hook)) { |
150 | 0 | slv = (S *)X(mksolver_ct_hook)(sizeof(S), desc->radix, dec, |
151 | 0 | mkcldw, 0); |
152 | 0 | slv->k = codelet; |
153 | 0 | slv->desc = desc; |
154 | 0 | REGISTER_SOLVER(plnr, &(slv->super.super)); |
155 | 0 | } |
156 | 6 | } |
157 | | |
158 | | void X(regsolver_ct_directwsq)(planner *plnr, kdftwsq codelet, |
159 | | const ct_desc *desc, int dec) |
160 | 6 | { |
161 | 6 | regone(plnr, codelet, desc, dec+TRANSPOSE); |
162 | 6 | } |