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., 59 Temple Place - Suite 330,
29 Boston, MA 02111-1307, USA. */
34 #include "libgfortran.h"
36 /* Dimensions: retarray(x,y) a(x, count) b(count,y).
37 Either a or b can be rank 1. In this case x or y is 1. */
39 extern void matmul_l8 (gfc_array_l8 *, gfc_array_l4 *, gfc_array_l4 *);
40 export_proto(matmul_l8);
43 matmul_l8 (gfc_array_l8 * retarray, gfc_array_l4 * a, gfc_array_l4 * b)
64 assert (GFC_DESCRIPTOR_RANK (a) == 2
65 || GFC_DESCRIPTOR_RANK (b) == 2);
67 if (retarray->data == NULL)
69 if (GFC_DESCRIPTOR_RANK (a) == 1)
71 retarray->dim[0].lbound = 0;
72 retarray->dim[0].ubound = b->dim[1].ubound - b->dim[1].lbound;
73 retarray->dim[0].stride = 1;
75 else if (GFC_DESCRIPTOR_RANK (b) == 1)
77 retarray->dim[0].lbound = 0;
78 retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
79 retarray->dim[0].stride = 1;
83 retarray->dim[0].lbound = 0;
84 retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
85 retarray->dim[0].stride = 1;
87 retarray->dim[1].lbound = 0;
88 retarray->dim[1].ubound = b->dim[1].ubound - b->dim[1].lbound;
89 retarray->dim[1].stride = retarray->dim[0].ubound+1;
93 = internal_malloc_size (sizeof (GFC_LOGICAL_8) * size0 ((array_t *) retarray));
98 if (GFC_DESCRIPTOR_SIZE (a) != 4)
100 assert (GFC_DESCRIPTOR_SIZE (a) == 8);
101 abase = GFOR_POINTER_L8_TO_L4 (abase);
104 if (GFC_DESCRIPTOR_SIZE (b) != 4)
106 assert (GFC_DESCRIPTOR_SIZE (b) == 8);
107 bbase = GFOR_POINTER_L8_TO_L4 (bbase);
109 dest = retarray->data;
111 if (retarray->dim[0].stride == 0)
112 retarray->dim[0].stride = 1;
113 if (a->dim[0].stride == 0)
114 a->dim[0].stride = 1;
115 if (b->dim[0].stride == 0)
116 b->dim[0].stride = 1;
119 if (GFC_DESCRIPTOR_RANK (retarray) == 1)
121 rxstride = retarray->dim[0].stride;
126 rxstride = retarray->dim[0].stride;
127 rystride = retarray->dim[1].stride;
130 /* If we have rank 1 parameters, zero the absent stride, and set the size to
132 if (GFC_DESCRIPTOR_RANK (a) == 1)
134 astride = a->dim[0].stride;
135 count = a->dim[0].ubound + 1 - a->dim[0].lbound;
142 astride = a->dim[1].stride;
143 count = a->dim[1].ubound + 1 - a->dim[1].lbound;
144 xstride = a->dim[0].stride;
145 xcount = a->dim[0].ubound + 1 - a->dim[0].lbound;
147 if (GFC_DESCRIPTOR_RANK (b) == 1)
149 bstride = b->dim[0].stride;
150 assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
157 bstride = b->dim[0].stride;
158 assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
159 ystride = b->dim[1].stride;
160 ycount = b->dim[1].ubound + 1 - b->dim[1].lbound;
163 for (y = 0; y < ycount; y++)
165 for (x = 0; x < xcount; x++)
167 /* Do the summation for this element. For real and integer types
168 this is the same as DOT_PRODUCT. For complex types we use do
169 a*b, not conjg(a)*b. */
174 for (n = 0; n < count; n++)
188 abase -= xstride * xcount;
190 dest += rystride - (rxstride * xcount);