Chris@82: /* Chris@82: * Copyright (c) 2003, 2007-14 Matteo Frigo Chris@82: * Copyright (c) 2003, 2007-14 Massachusetts Institute of Technology Chris@82: * Chris@82: * This program is free software; you can redistribute it and/or modify Chris@82: * it under the terms of the GNU General Public License as published by Chris@82: * the Free Software Foundation; either version 2 of the License, or Chris@82: * (at your option) any later version. Chris@82: * Chris@82: * This program is distributed in the hope that it will be useful, Chris@82: * but WITHOUT ANY WARRANTY; without even the implied warranty of Chris@82: * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the Chris@82: * GNU General Public License for more details. Chris@82: * Chris@82: * You should have received a copy of the GNU General Public License Chris@82: * along with this program; if not, write to the Free Software Chris@82: * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA Chris@82: * Chris@82: */ Chris@82: Chris@82: Chris@82: #include "kernel/ifftw.h" Chris@82: Chris@82: INT X(tensor_max_index)(const tensor *sz) Chris@82: { Chris@82: int i; Chris@82: INT ni = 0, no = 0; Chris@82: Chris@82: A(FINITE_RNK(sz->rnk)); Chris@82: for (i = 0; i < sz->rnk; ++i) { Chris@82: const iodim *p = sz->dims + i; Chris@82: ni += (p->n - 1) * X(iabs)(p->is); Chris@82: no += (p->n - 1) * X(iabs)(p->os); Chris@82: } Chris@82: return X(imax)(ni, no); Chris@82: } Chris@82: Chris@82: #define tensor_min_xstride(sz, xs) { \ Chris@82: A(FINITE_RNK(sz->rnk)); \ Chris@82: if (sz->rnk == 0) return 0; \ Chris@82: else { \ Chris@82: int i; \ Chris@82: INT s = X(iabs)(sz->dims[0].xs); \ Chris@82: for (i = 1; i < sz->rnk; ++i) \ Chris@82: s = X(imin)(s, X(iabs)(sz->dims[i].xs)); \ Chris@82: return s; \ Chris@82: } \ Chris@82: } Chris@82: Chris@82: INT X(tensor_min_istride)(const tensor *sz) tensor_min_xstride(sz, is) Chris@82: INT X(tensor_min_ostride)(const tensor *sz) tensor_min_xstride(sz, os) Chris@82: Chris@82: INT X(tensor_min_stride)(const tensor *sz) Chris@82: { Chris@82: return X(imin)(X(tensor_min_istride)(sz), X(tensor_min_ostride)(sz)); Chris@82: } Chris@82: Chris@82: int X(tensor_inplace_strides)(const tensor *sz) Chris@82: { Chris@82: int i; Chris@82: A(FINITE_RNK(sz->rnk)); Chris@82: for (i = 0; i < sz->rnk; ++i) { Chris@82: const iodim *p = sz->dims + i; Chris@82: if (p->is != p->os) Chris@82: return 0; Chris@82: } Chris@82: return 1; Chris@82: } Chris@82: Chris@82: int X(tensor_inplace_strides2)(const tensor *a, const tensor *b) Chris@82: { Chris@82: return X(tensor_inplace_strides(a)) && X(tensor_inplace_strides(b)); Chris@82: } Chris@82: Chris@82: /* return true (1) iff *any* strides of sz decrease when we Chris@82: tensor_inplace_copy(sz, k). */ Chris@82: static int tensor_strides_decrease(const tensor *sz, inplace_kind k) Chris@82: { Chris@82: if (FINITE_RNK(sz->rnk)) { Chris@82: int i; Chris@82: for (i = 0; i < sz->rnk; ++i) Chris@82: if ((sz->dims[i].os - sz->dims[i].is) Chris@82: * (k == INPLACE_OS ? (INT)1 : (INT)-1) < 0) Chris@82: return 1; Chris@82: } Chris@82: return 0; Chris@82: } Chris@82: Chris@82: /* Return true (1) iff *any* strides of sz decrease when we Chris@82: tensor_inplace_copy(k) *or* if *all* strides of sz are unchanged Chris@82: but *any* strides of vecsz decrease. This is used in indirect.c Chris@82: to determine whether to use INPLACE_IS or INPLACE_OS. Chris@82: Chris@82: Note: X(tensor_strides_decrease)(sz, vecsz, INPLACE_IS) Chris@82: || X(tensor_strides_decrease)(sz, vecsz, INPLACE_OS) Chris@82: || X(tensor_inplace_strides2)(p->sz, p->vecsz) Chris@82: must always be true. */ Chris@82: int X(tensor_strides_decrease)(const tensor *sz, const tensor *vecsz, Chris@82: inplace_kind k) Chris@82: { Chris@82: return(tensor_strides_decrease(sz, k) Chris@82: || (X(tensor_inplace_strides)(sz) Chris@82: && tensor_strides_decrease(vecsz, k))); Chris@82: }