1 `/* Implementation of the MATMUL intrinsic
2 Copyright 2002 Free Software Foundation, Inc.
3 Contributed by Paul Brook <paul@nowt.org>
5 This file is part of the GNU Fortran 95 runtime library (libgfor).
7 Libgfortran is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Lesser General Public
9 License as published by the Free Software Foundation; either
10 version 2.1 of the License, or (at your option) any later version.
12 Libgfortran is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU Lesser General Public License for more details.
17 You should have received a copy of the GNU Lesser General Public
18 License along with libgfor; see the file COPYING.LIB. If not,
19 write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
20 Boston, MA 02111-1307, USA. */
25 #include "libgfortran.h"'
27 define(rtype_kind, regexp(file, `_l\([0-9]+\)\.', `\1'))dnl
28 define(rtype_code,`l'rtype_kind)dnl
29 define(rtype,get_arraytype(l,rtype_kind))dnl
30 define(rtype_name, get_typename(l, rtype_kind))dnl
32 /* Dimensions: retarray(x,y) a(x, count) b(count,y).
33 Either a or b can be rank 1. In this case x or y is 1. */
35 `__matmul_'rtype_code (rtype * retarray, gfc_array_l4 * a, gfc_array_l4 * b)
56 assert (GFC_DESCRIPTOR_RANK (a) == 2
57 || GFC_DESCRIPTOR_RANK (b) == 2);
59 if (GFC_DESCRIPTOR_SIZE (a) != 4)
61 assert (GFC_DESCRIPTOR_SIZE (a) == 8);
62 abase = GFOR_POINTER_L8_TO_L4 (abase);
66 if (GFC_DESCRIPTOR_SIZE (b) != 4)
68 assert (GFC_DESCRIPTOR_SIZE (b) == 8);
69 bbase = GFOR_POINTER_L8_TO_L4 (bbase);
72 dest = retarray->data;
74 if (retarray->dim[0].stride == 0)
75 retarray->dim[0].stride = 1;
76 if (a->dim[0].stride == 0)
78 if (b->dim[0].stride == 0)
81 sinclude(`matmul_asm_'rtype_code`.m4')dnl
83 if (GFC_DESCRIPTOR_RANK (retarray) == 1)
85 rxstride = retarray->dim[0].stride;
90 rxstride = retarray->dim[0].stride;
91 rystride = retarray->dim[1].stride;
94 /* If we have rank 1 parameters, zero the absent stride, and set the size to
96 if (GFC_DESCRIPTOR_RANK (a) == 1)
98 astride = a->dim[0].stride;
99 count = a->dim[0].ubound + 1 - a->dim[0].lbound;
106 astride = a->dim[1].stride;
107 count = a->dim[1].ubound + 1 - a->dim[1].lbound;
108 xstride = a->dim[0].stride;
109 xcount = a->dim[0].ubound + 1 - a->dim[0].lbound;
111 if (GFC_DESCRIPTOR_RANK (b) == 1)
113 bstride = b->dim[0].stride;
114 assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
121 bstride = b->dim[0].stride;
122 assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
123 ystride = b->dim[1].stride;
124 ycount = b->dim[1].ubound + 1 - b->dim[1].lbound;
127 for (y = 0; y < ycount; y++)
129 for (x = 0; x < xcount; x++)
131 /* Do the summation for this element. For real and integer types
132 this is the same as DOT_PRODUCT. For complex types we use do
133 a*b, not conjg(a)*b. */
138 for (n = 0; n < count; n++)
152 abase -= xstride * xcount;
154 dest += rystride - (rxstride * xcount);