static plan *mkcldw(const ct_solver *ego_,
INT r, INT irs, INT ors,
INT m, INT ms,
INT v, INT ivs, INT ovs,
INT mstart, INT mcount,
R *rio, R *iio,
planner *plnr)
{
const S *ego = (const S *) ego_;
P *pln;
const ct_desc *e = ego->desc;
INT extra_iter;
static const plan_adt padt = {
0, awake, print, destroy
};
A(mstart >= 0 && mstart + mcount <= m);
if (!applicable(ego,
r, irs, ors, m, ms, v, ivs, ovs, mstart, mstart + mcount,
rio, iio, plnr, &extra_iter))
return (plan *)0;
if (ego->bufferedp) {
pln = MKPLAN_DFTW(P, &padt, apply_buf);
} else {
pln = MKPLAN_DFTW(P, &padt, extra_iter ? apply_extra_iter : apply);
}
pln->k = ego->k;
pln->rs = X(mkstride)(r, irs);
pln->td = 0;
pln->r = r;
pln->m = m;
pln->ms = ms;
pln->v = v;
pln->vs = ivs;
pln->mb = mstart;
pln->me = mstart + mcount;
pln->slv = ego;
pln->brs = X(mkstride)(r, 2 * compute_batchsize(r));
pln->extra_iter = extra_iter;
X(ops_zero)(&pln->super.super.ops);
X(ops_madd2)(v * (mcount/e->genus->vl), &e->ops, &pln->super.super.ops);
if (ego->bufferedp) {
/* 8 load/stores * N * V */
pln->super.super.ops.other += 8 * r * mcount * v;
}
pln->super.super.could_prune_now_p =
(!ego->bufferedp && r >= 5 && r < 64 && m >= r);
return &(pln->super.super);
}
static plan *mkcldw(const ct_solver *ego_,
int dec, INT r, INT m, INT s, INT vl, INT vs,
INT mstart, INT mcount,
R *rio, R *iio,
planner *plnr)
{
const S *ego = (const S *)ego_;
P *pln;
plan *cld = 0;
static const plan_adt padt = {
0, awake, print, destroy
};
A(mstart >= 0 && mstart + mcount <= m);
if (!applicable(r, m, plnr))
return (plan *)0;
cld = X(mkplan_d)(plnr,
X(mkproblem_dft_d)(
X(mktensor_1d)(r, m * s, m * s),
X(mktensor_2d)(mcount, s, s, vl, vs, vs),
rio, iio, rio, iio)
);
if (!cld) goto nada;
pln = MKPLAN_DFTW(P, &padt, dec == DECDIT ? apply_dit : apply_dif);
pln->slv = ego;
pln->cld = cld;
pln->r = r;
pln->m = m;
pln->s = s;
pln->vl = vl;
pln->vs = vs;
pln->mstart = mstart;
pln->mcount = mcount;
pln->dec = dec;
pln->td = 0;
{
double n0 = (r - 1) * (mcount - 1) * vl;
pln->super.super.ops = cld->ops;
pln->super.super.ops.mul += 8 * n0;
pln->super.super.ops.add += 4 * n0;
pln->super.super.ops.other += 8 * n0;
}
return &(pln->super.super);
nada:
X(plan_destroy_internal)(cld);
return (plan *) 0;
}
static plan *mkcldw(const ct_solver *ego_,
INT r, INT irs, INT ors,
INT m, INT ms,
INT v, INT ivs, INT ovs,
INT mstart, INT mcount,
R *rio, R *iio,
planner *plnr)
{
const S *ego = (const S *) ego_;
P *pln;
const ct_desc *e = ego->desc;
static const plan_adt padt = {
0, awake, print, destroy
};
A(mstart >= 0 && mstart + mcount <= m);
if (!applicable(ego,
r, irs, ors, m, ms, v, ivs, ovs, mstart, mstart + mcount,
rio, iio, plnr))
return (plan *)0;
pln = MKPLAN_DFTW(P, &padt, apply);
pln->k = ego->k;
pln->rs = X(mkstride)(r, irs);
pln->vs = X(mkstride)(v, ivs);
pln->td = 0;
pln->r = r;
pln->m = m;
pln->ms = ms;
pln->v = v;
pln->mb = mstart;
pln->me = mstart + mcount;
pln->slv = ego;
X(ops_zero)(&pln->super.super.ops);
X(ops_madd2)(mcount/e->genus->vl, &e->ops, &pln->super.super.ops);
return &(pln->super.super);
}
