1 `/* Implementation of the MATMUL intrinsic
2 Copyright 2002, 2005 Free Software Foundation, Inc.
3 Contributed by Paul Brook <paul@nowt.org>
5 This file is part of the GNU Fortran 95 runtime library (libgfortran).
7 Libgfortran is free software; you can redistribute it and/or
8 modify it under the terms of the GNU General Public
9 License as published by the Free Software Foundation; either
10 version 2 of the License, or (at your option) any later version.
12 In addition to the permissions in the GNU General Public License, the
13 Free Software Foundation gives you unlimited permission to link the
14 compiled version of this file into combinations with other programs,
15 and to distribute those combinations without any restriction coming
16 from the use of this file. (The General Public License restrictions
17 do apply in other respects; for example, they cover modification of
18 the file, and distribution when not linked into a combine
21 Libgfortran is distributed in the hope that it will be useful,
22 but WITHOUT ANY WARRANTY; without even the implied warranty of
23 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
24 GNU General Public License for more details.
26 You should have received a copy of the GNU General Public
27 License along with libgfortran; see the file COPYING. If not,
28 write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
29 Boston, MA 02110-1301, USA. */
34 #include "libgfortran.h"'
37 /* Dimensions: retarray(x,y) a(x, count) b(count,y).
38 Either a or b can be rank 1. In this case x or y is 1. */
40 extern void matmul_`'rtype_code (rtype *, gfc_array_l4 *, gfc_array_l4 *);
41 export_proto(matmul_`'rtype_code);
44 matmul_`'rtype_code (rtype * retarray, gfc_array_l4 * a, gfc_array_l4 * b)
65 assert (GFC_DESCRIPTOR_RANK (a) == 2
66 || GFC_DESCRIPTOR_RANK (b) == 2);
68 if (retarray->data == NULL)
70 if (GFC_DESCRIPTOR_RANK (a) == 1)
72 retarray->dim[0].lbound = 0;
73 retarray->dim[0].ubound = b->dim[1].ubound - b->dim[1].lbound;
74 retarray->dim[0].stride = 1;
76 else if (GFC_DESCRIPTOR_RANK (b) == 1)
78 retarray->dim[0].lbound = 0;
79 retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
80 retarray->dim[0].stride = 1;
84 retarray->dim[0].lbound = 0;
85 retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
86 retarray->dim[0].stride = 1;
88 retarray->dim[1].lbound = 0;
89 retarray->dim[1].ubound = b->dim[1].ubound - b->dim[1].lbound;
90 retarray->dim[1].stride = retarray->dim[0].ubound+1;
94 = internal_malloc_size (sizeof (rtype_name) * size0 ((array_t *) retarray));
99 if (GFC_DESCRIPTOR_SIZE (a) != 4)
101 assert (GFC_DESCRIPTOR_SIZE (a) == 8);
102 abase = GFOR_POINTER_L8_TO_L4 (abase);
105 if (GFC_DESCRIPTOR_SIZE (b) != 4)
107 assert (GFC_DESCRIPTOR_SIZE (b) == 8);
108 bbase = GFOR_POINTER_L8_TO_L4 (bbase);
110 dest = retarray->data;
112 if (retarray->dim[0].stride == 0)
113 retarray->dim[0].stride = 1;
114 if (a->dim[0].stride == 0)
115 a->dim[0].stride = 1;
116 if (b->dim[0].stride == 0)
117 b->dim[0].stride = 1;
119 sinclude(`matmul_asm_'rtype_code`.m4')dnl
121 if (GFC_DESCRIPTOR_RANK (retarray) == 1)
123 rxstride = retarray->dim[0].stride;
128 rxstride = retarray->dim[0].stride;
129 rystride = retarray->dim[1].stride;
132 /* If we have rank 1 parameters, zero the absent stride, and set the size to
134 if (GFC_DESCRIPTOR_RANK (a) == 1)
136 astride = a->dim[0].stride;
137 count = a->dim[0].ubound + 1 - a->dim[0].lbound;
144 astride = a->dim[1].stride;
145 count = a->dim[1].ubound + 1 - a->dim[1].lbound;
146 xstride = a->dim[0].stride;
147 xcount = a->dim[0].ubound + 1 - a->dim[0].lbound;
149 if (GFC_DESCRIPTOR_RANK (b) == 1)
151 bstride = b->dim[0].stride;
152 assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
159 bstride = b->dim[0].stride;
160 assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
161 ystride = b->dim[1].stride;
162 ycount = b->dim[1].ubound + 1 - b->dim[1].lbound;
165 for (y = 0; y < ycount; y++)
167 for (x = 0; x < xcount; x++)
169 /* Do the summation for this element. For real and integer types
170 this is the same as DOT_PRODUCT. For complex types we use do
171 a*b, not conjg(a)*b. */
176 for (n = 0; n < count; n++)
190 abase -= xstride * xcount;
192 dest += rystride - (rxstride * xcount);