Mercurial > hg > sv-dependency-builds

annotate src/fftw-3.3.5/dft/dftw-directsq.c @ 42:2cd0e3b3e1fd

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Current fftw source
author Chris Cannam
date Tue, 18 Oct 2016 13:40:26 +0100
parents
children
rev   line source
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
Chris@42 22 #include "ct.h"
Chris@42 23
Chris@42 24 typedef struct {
Chris@42 25 ct_solver super;
Chris@42 26 const ct_desc *desc;
Chris@42 27 kdftwsq k;
Chris@42 28 } S;
Chris@42 29
Chris@42 30 typedef struct {
Chris@42 31 plan_dftw super;
Chris@42 32 kdftwsq k;
Chris@42 33 INT r;
Chris@42 34 stride rs, vs;
Chris@42 35 INT m, ms, v, mb, me;
Chris@42 36 twid *td;
Chris@42 37 const S *slv;
Chris@42 38 } P;
Chris@42 39
Chris@42 40
Chris@42 41 static void apply(const plan *ego_, R *rio, R *iio)
Chris@42 42 {
Chris@42 43 const P *ego = (const P *) ego_;
Chris@42 44 INT mb = ego->mb, ms = ego->ms;
Chris@42 45 ego->k(rio + mb*ms, iio + mb*ms, ego->td->W, ego->rs, ego->vs,
Chris@42 46 mb, ego->me, ms);
Chris@42 47 }
Chris@42 48
Chris@42 49 static void awake(plan *ego_, enum wakefulness wakefulness)
Chris@42 50 {
Chris@42 51 P *ego = (P *) ego_;
Chris@42 52
Chris@42 53 X(twiddle_awake)(wakefulness, &ego->td, ego->slv->desc->tw,
Chris@42 54 ego->r * ego->m, ego->r, ego->m);
Chris@42 55 }
Chris@42 56
Chris@42 57 static void destroy(plan *ego_)
Chris@42 58 {
Chris@42 59 P *ego = (P *) ego_;
Chris@42 60 X(stride_destroy)(ego->rs);
Chris@42 61 X(stride_destroy)(ego->vs);
Chris@42 62 }
Chris@42 63
Chris@42 64 static void print(const plan *ego_, printer *p)
Chris@42 65 {
Chris@42 66 const P *ego = (const P *) ego_;
Chris@42 67 const S *slv = ego->slv;
Chris@42 68 const ct_desc *e = slv->desc;
Chris@42 69
Chris@42 70 p->print(p, "(dftw-directsq-%D/%D%v \"%s\")",
Chris@42 71 ego->r, X(twiddle_length)(ego->r, e->tw), ego->v, e->nam);
Chris@42 72 }
Chris@42 73
Chris@42 74 static int applicable(const S *ego,
Chris@42 75 INT r, INT irs, INT ors,
Chris@42 76 INT m, INT ms,
Chris@42 77 INT v, INT ivs, INT ovs,
Chris@42 78 INT mb, INT me,
Chris@42 79 R *rio, R *iio,
Chris@42 80 const planner *plnr)
Chris@42 81 {
Chris@42 82 const ct_desc *e = ego->desc;
Chris@42 83 UNUSED(v);
Chris@42 84
Chris@42 85 return (
Chris@42 86 1
Chris@42 87 && r == e->radix
Chris@42 88
Chris@42 89 /* transpose r, v */
Chris@42 90 && r == v
Chris@42 91 && irs == ovs
Chris@42 92 && ivs == ors
Chris@42 93
Chris@42 94 /* check for alignment/vector length restrictions */
Chris@42 95 && e->genus->okp(e, rio, iio, irs, ivs, m, mb, me, ms, plnr)
Chris@42 96
Chris@42 97 );
Chris@42 98 }
Chris@42 99
Chris@42 100 static plan *mkcldw(const ct_solver *ego_,
Chris@42 101 INT r, INT irs, INT ors,
Chris@42 102 INT m, INT ms,
Chris@42 103 INT v, INT ivs, INT ovs,
Chris@42 104 INT mstart, INT mcount,
Chris@42 105 R *rio, R *iio,
Chris@42 106 planner *plnr)
Chris@42 107 {
Chris@42 108 const S *ego = (const S *) ego_;
Chris@42 109 P *pln;
Chris@42 110 const ct_desc *e = ego->desc;
Chris@42 111
Chris@42 112 static const plan_adt padt = {
Chris@42 113 0, awake, print, destroy
Chris@42 114 };
Chris@42 115
Chris@42 116 A(mstart >= 0 && mstart + mcount <= m);
Chris@42 117 if (!applicable(ego,
Chris@42 118 r, irs, ors, m, ms, v, ivs, ovs, mstart, mstart + mcount,
Chris@42 119 rio, iio, plnr))
Chris@42 120 return (plan *)0;
Chris@42 121
Chris@42 122 pln = MKPLAN_DFTW(P, &padt, apply);
Chris@42 123
Chris@42 124 pln->k = ego->k;
Chris@42 125 pln->rs = X(mkstride)(r, irs);
Chris@42 126 pln->vs = X(mkstride)(v, ivs);
Chris@42 127 pln->td = 0;
Chris@42 128 pln->r = r;
Chris@42 129 pln->m = m;
Chris@42 130 pln->ms = ms;
Chris@42 131 pln->v = v;
Chris@42 132 pln->mb = mstart;
Chris@42 133 pln->me = mstart + mcount;
Chris@42 134 pln->slv = ego;
Chris@42 135
Chris@42 136 X(ops_zero)(&pln->super.super.ops);
Chris@42 137 X(ops_madd2)(mcount/e->genus->vl, &e->ops, &pln->super.super.ops);
Chris@42 138
Chris@42 139 return &(pln->super.super);
Chris@42 140 }
Chris@42 141
Chris@42 142 static void regone(planner *plnr, kdftwsq codelet,
Chris@42 143 const ct_desc *desc, int dec)
Chris@42 144 {
Chris@42 145 S *slv = (S *)X(mksolver_ct)(sizeof(S), desc->radix, dec, mkcldw, 0);
Chris@42 146 slv->k = codelet;
Chris@42 147 slv->desc = desc;
Chris@42 148 REGISTER_SOLVER(plnr, &(slv->super.super));
Chris@42 149 if (X(mksolver_ct_hook)) {
Chris@42 150 slv = (S *)X(mksolver_ct_hook)(sizeof(S), desc->radix, dec,
Chris@42 151 mkcldw, 0);
Chris@42 152 slv->k = codelet;
Chris@42 153 slv->desc = desc;
Chris@42 154 REGISTER_SOLVER(plnr, &(slv->super.super));
Chris@42 155 }
Chris@42 156 }
Chris@42 157
Chris@42 158 void X(regsolver_ct_directwsq)(planner *plnr, kdftwsq codelet,
Chris@42 159 const ct_desc *desc, int dec)
Chris@42 160 {
Chris@42 161 regone(plnr, codelet, desc, dec+TRANSPOSE);
Chris@42 162 }