Chris@10: /*
Chris@10:  * Copyright (c) 2003, 2007-11 Matteo Frigo
Chris@10:  * Copyright (c) 2003, 2007-11 Massachusetts Institute of Technology
Chris@10:  *
Chris@10:  * This program is free software; you can redistribute it and/or modify
Chris@10:  * it under the terms of the GNU General Public License as published by
Chris@10:  * the Free Software Foundation; either version 2 of the License, or
Chris@10:  * (at your option) any later version.
Chris@10:  *
Chris@10:  * This program is distributed in the hope that it will be useful,
Chris@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
Chris@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
Chris@10:  * GNU General Public License for more details.
Chris@10:  *
Chris@10:  * You should have received a copy of the GNU General Public License
Chris@10:  * along with this program; if not, write to the Free Software
Chris@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301  USA
Chris@10:  *
Chris@10:  */
Chris@10: 
Chris@10: 
Chris@10: #include "ct.h"
Chris@10: 
Chris@10: typedef struct {
Chris@10:      ct_solver super;
Chris@10:      const ct_desc *desc;
Chris@10:      kdftwsq k;
Chris@10: } S;
Chris@10: 
Chris@10: typedef struct {
Chris@10:      plan_dftw super;
Chris@10:      kdftwsq k;
Chris@10:      INT r;
Chris@10:      stride rs, vs;
Chris@10:      INT m, ms, v, mb, me;
Chris@10:      twid *td;
Chris@10:      const S *slv;
Chris@10: } P;
Chris@10: 
Chris@10: 
Chris@10: static void apply(const plan *ego_, R *rio, R *iio)
Chris@10: {
Chris@10:      const P *ego = (const P *) ego_;
Chris@10:      INT mb = ego->mb, ms = ego->ms;
Chris@10:      ego->k(rio + mb*ms, iio + mb*ms, ego->td->W, ego->rs, ego->vs,
Chris@10: 	    mb, ego->me, ms);
Chris@10: }
Chris@10: 
Chris@10: static void awake(plan *ego_, enum wakefulness wakefulness)
Chris@10: {
Chris@10:      P *ego = (P *) ego_;
Chris@10: 
Chris@10:      X(twiddle_awake)(wakefulness, &ego->td, ego->slv->desc->tw,
Chris@10: 		      ego->r * ego->m, ego->r, ego->m);
Chris@10: }
Chris@10: 
Chris@10: static void destroy(plan *ego_)
Chris@10: {
Chris@10:      P *ego = (P *) ego_;
Chris@10:      X(stride_destroy)(ego->rs);
Chris@10:      X(stride_destroy)(ego->vs);
Chris@10: }
Chris@10: 
Chris@10: static void print(const plan *ego_, printer *p)
Chris@10: {
Chris@10:      const P *ego = (const P *) ego_;
Chris@10:      const S *slv = ego->slv;
Chris@10:      const ct_desc *e = slv->desc;
Chris@10: 
Chris@10:      p->print(p, "(dftw-directsq-%D/%D%v \"%s\")",
Chris@10: 	      ego->r, X(twiddle_length)(ego->r, e->tw), ego->v, e->nam);
Chris@10: }
Chris@10: 
Chris@10: static int applicable(const S *ego,
Chris@10: 		      INT r, INT irs, INT ors,
Chris@10: 		      INT m, INT ms,
Chris@10: 		      INT v, INT ivs, INT ovs,
Chris@10: 		      INT mb, INT me,
Chris@10: 		      R *rio, R *iio,
Chris@10: 		      const planner *plnr)
Chris@10: {
Chris@10:      const ct_desc *e = ego->desc;
Chris@10:      UNUSED(v);
Chris@10: 
Chris@10:      return (
Chris@10: 	  1
Chris@10: 	  && r == e->radix
Chris@10: 
Chris@10: 	  /* transpose r, v */
Chris@10: 	  && r == v
Chris@10: 	  && irs == ovs
Chris@10: 	  && ivs == ors
Chris@10: 
Chris@10: 	  /* check for alignment/vector length restrictions */
Chris@10: 	  && e->genus->okp(e, rio, iio, irs, ivs, m, mb, me, ms, plnr)
Chris@10: 
Chris@10: 	  );
Chris@10: }
Chris@10: 
Chris@10: static plan *mkcldw(const ct_solver *ego_,
Chris@10: 		    INT r, INT irs, INT ors,
Chris@10: 		    INT m, INT ms,
Chris@10: 		    INT v, INT ivs, INT ovs,
Chris@10: 		    INT mstart, INT mcount,
Chris@10: 		    R *rio, R *iio,
Chris@10: 		    planner *plnr)
Chris@10: {
Chris@10:      const S *ego = (const S *) ego_;
Chris@10:      P *pln;
Chris@10:      const ct_desc *e = ego->desc;
Chris@10: 
Chris@10:      static const plan_adt padt = {
Chris@10: 	  0, awake, print, destroy
Chris@10:      };
Chris@10: 
Chris@10:      A(mstart >= 0 && mstart + mcount <= m);
Chris@10:      if (!applicable(ego,
Chris@10: 		     r, irs, ors, m, ms, v, ivs, ovs, mstart, mstart + mcount,
Chris@10: 		     rio, iio, plnr))
Chris@10:           return (plan *)0;
Chris@10: 
Chris@10:      pln = MKPLAN_DFTW(P, &padt, apply);
Chris@10: 
Chris@10:      pln->k = ego->k;
Chris@10:      pln->rs = X(mkstride)(r, irs);
Chris@10:      pln->vs = X(mkstride)(v, ivs);
Chris@10:      pln->td = 0;
Chris@10:      pln->r = r;
Chris@10:      pln->m = m;
Chris@10:      pln->ms = ms;
Chris@10:      pln->v = v;
Chris@10:      pln->mb = mstart;
Chris@10:      pln->me = mstart + mcount;
Chris@10:      pln->slv = ego;
Chris@10: 
Chris@10:      X(ops_zero)(&pln->super.super.ops);
Chris@10:      X(ops_madd2)(mcount/e->genus->vl, &e->ops, &pln->super.super.ops);
Chris@10: 
Chris@10:      return &(pln->super.super);
Chris@10: }
Chris@10: 
Chris@10: static void regone(planner *plnr, kdftwsq codelet,
Chris@10: 		   const ct_desc *desc, int dec)
Chris@10: {
Chris@10:      S *slv = (S *)X(mksolver_ct)(sizeof(S), desc->radix, dec, mkcldw, 0);
Chris@10:      slv->k = codelet;
Chris@10:      slv->desc = desc;
Chris@10:      REGISTER_SOLVER(plnr, &(slv->super.super));
Chris@10:      if (X(mksolver_ct_hook)) {
Chris@10: 	  slv = (S *)X(mksolver_ct_hook)(sizeof(S), desc->radix, dec,
Chris@10: 					 mkcldw, 0);
Chris@10: 	  slv->k = codelet;
Chris@10: 	  slv->desc = desc;
Chris@10: 	  REGISTER_SOLVER(plnr, &(slv->super.super));
Chris@10:      }
Chris@10: }
Chris@10: 
Chris@10: void X(regsolver_ct_directwsq)(planner *plnr, kdftwsq codelet,
Chris@10: 			       const ct_desc *desc, int dec)
Chris@10: {
Chris@10:      regone(plnr, codelet, desc, dec+TRANSPOSE);
Chris@10: }