diff options
Diffstat (limited to 'src/fftw3/dft/rank0.c')
-rw-r--r-- | src/fftw3/dft/rank0.c | 235 |
1 files changed, 235 insertions, 0 deletions
diff --git a/src/fftw3/dft/rank0.c b/src/fftw3/dft/rank0.c new file mode 100644 index 0000000..bef719f --- /dev/null +++ b/src/fftw3/dft/rank0.c @@ -0,0 +1,235 @@ +/* + * Copyright (c) 2003 Matteo Frigo + * Copyright (c) 2003 Massachusetts Institute of Technology + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + */ + +/* $Id: rank0.c,v 1.1 2008/10/17 06:11:08 scuri Exp $ */ + +/* plans for rank-0 DFTs (copy operations) */ + +#include "dft.h" + +#ifdef HAVE_STRING_H +#include <string.h> /* for memcpy() */ +#endif + + +typedef struct { + dftapply apply; + int (*applicable)(const problem_dft *p); + const char *nam; +} rnk0adt; + +typedef struct { + solver super; + const rnk0adt *adt; +} S; + +typedef struct { + plan_dft super; + int vl; + int ivs, ovs; + const S *slv; +} P; + +/* generic applicability function */ +static int applicable(const solver *ego_, const problem *p_) +{ + if (DFTP(p_)) { + const S *ego = (const S *) ego_; + const problem_dft *p = (const problem_dft *) p_; + return (1 + && p->ri != p->ro + && p->sz->rnk == 0 + && ego->adt->applicable(p) + ); + } + return 0; +} + +/*-----------------------------------------------------------------------*/ +/* rank-0 dft, vl == 1: just a copy */ +static void apply_1(const plan *ego_, R *ri, R *ii, R *ro, R *io) +{ + UNUSED(ego_); + *ro = *ri; + *io = *ii; +} + +static int applicable_1(const problem_dft *p) +{ + return (p->vecsz->rnk == 0); +} + +static const rnk0adt adt_cpy1 = +{ + apply_1, applicable_1, "dft-rank0-cpy1" +}; + +/*-----------------------------------------------------------------------*/ +/* rank-0 dft, vl > 1: just a copy loop (unroll 4) */ +static void apply_vec(const plan *ego_, R *ri, R *ii, R *ro, R *io) +{ + const P *ego = (const P *) ego_; + int i, vl = ego->vl; + int ivs = ego->ivs, ovs = ego->ovs; + int imi = ii - ri, imo = io - ro; + for (i = vl; i > 0; --i) { + R r0 = ri[0], i0 = ri[imi]; + ro[0] = r0; ro[imo] = i0; + ri += ivs; ro += ovs; + } +} + +static int applicable_vec(const problem_dft *p) +{ + return (p->vecsz->rnk == 1 && p->ro != p->ri); +} + +static const rnk0adt adt_vec = +{ + apply_vec, applicable_vec, "dft-rank0-vec" +}; + +/*-----------------------------------------------------------------------*/ +/* rank-0 dft, vl > 1, [io]vs == 1, using memcpy */ +static void apply_io1(const plan *ego_, R *ri, R *ii, R *ro, R *io) +{ + const P *ego = (const P *) ego_; + int vl = ego->vl; + memcpy(ro, ri, vl * sizeof(R)); + memcpy(io, ii, vl * sizeof(R)); +} + +static int applicable_io1(const problem_dft *p) +{ + return (1 + && applicable_vec(p) + && p->vecsz->dims[0].is == 1 + && p->vecsz->dims[0].os == 1 + ); +} + +static const rnk0adt adt_io1 = +{ + apply_io1, applicable_io1, "dft-rank0-io1-memcpy" +}; + +/*-----------------------------------------------------------------------*/ +/* rank-0 dft, vl > 1, [io]vs == 2 (interleaved) using memcpy */ +static void apply_io2r(const plan *ego_, R *ri, R *ii, R *ro, R *io) +{ + const P *ego = (const P *) ego_; + int vl = ego->vl; + UNUSED(ii); + UNUSED(io); /* i{in,out}put == r{in,out}put + 1 */ + memcpy(ro, ri, vl * sizeof(R) * 2); +} + +static int applicable_io2r(const problem_dft *p) +{ + return (1 + && applicable_vec(p) + && p->vecsz->dims[0].is == 2 + && p->vecsz->dims[0].os == 2 + && p->ii == p->ri + 1 && p->io == p->ro + 1 + ); +} + +static const rnk0adt adt_io2r = +{ + apply_io2r, applicable_io2r, "dft-rank0-io2r-memcpy" +}; + +static void apply_io2i(const plan *ego_, R *ri, R *ii, R *ro, R *io) +{ + const P *ego = (const P *) ego_; + int vl = ego->vl; + UNUSED(ri); + UNUSED(ro); /* r{in,out}put == i{in,out}put + 1 */ + memcpy(io, ii, vl * sizeof(R) * 2); +} + +static int applicable_io2i(const problem_dft *p) +{ + return (1 + && applicable_vec(p) + && p->vecsz->dims[0].is == 2 + && p->vecsz->dims[0].os == 2 + && p->ri == p->ii + 1 && p->ro == p->io + 1 + ); +} + +static const rnk0adt adt_io2i = +{ + apply_io2i, applicable_io2i, "dft-rank0-io2i-memcpy" +}; + +/*-----------------------------------------------------------------------*/ +/* generic stuff: */ + +static void print(const plan *ego_, printer *p) +{ + const P *ego = (const P *) ego_; + p->print(p, "(%s%v)", ego->slv->adt->nam, ego->vl); +} + +static plan *mkplan(const solver *ego_, const problem *p_, planner *plnr) +{ + const S *ego = (const S *) ego_; + const problem_dft *p; + P *pln; + + static const plan_adt padt = { + X(dft_solve), X(null_awake), print, X(plan_null_destroy) + }; + + UNUSED(plnr); + + if (!applicable(ego_, p_)) + return (plan *) 0; + + p = (const problem_dft *) p_; + pln = MKPLAN_DFT(P, &padt, ego->adt->apply); + + X(tensor_tornk1)(p->vecsz, &pln->vl, &pln->ivs, &pln->ovs); + pln->slv = ego; + + /* 2*vl loads, 2*vl stores */ + X(ops_other)(4 * pln->vl, &pln->super.super.ops); + return &(pln->super.super); +} + +static solver *mksolver(const rnk0adt *adt) +{ + static const solver_adt sadt = { mkplan }; + S *slv = MKSOLVER(S, &sadt); + slv->adt = adt; + return &(slv->super); +} + +void X(dft_rank0_register)(planner *p) +{ + unsigned i; + static const rnk0adt *const adts[] = { + &adt_cpy1, &adt_vec, &adt_io1, &adt_io2r, &adt_io2i + }; + + for (i = 0; i < sizeof(adts) / sizeof(adts[0]); ++i) + REGISTER_SOLVER(p, mksolver(adts[i])); +} |