1 `/* Implementation of the MATMUL intrinsic
2 Copyright 2002, 2005, 2006 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 `#if defined (HAVE_'rtype_name`)'
39 /* Dimensions: retarray(x,y) a(x, count) b(count,y).
40 Either a or b can be rank 1. In this case x or y is 1. */
42 extern void matmul_`'rtype_code (rtype * const restrict,
43 gfc_array_l4 * const restrict, gfc_array_l4 * const restrict);
44 export_proto(matmul_`'rtype_code);
47 matmul_`'rtype_code (rtype * const restrict retarray,
48 gfc_array_l4 * const restrict a, gfc_array_l4 * const restrict b)
50 const GFC_INTEGER_4 * restrict abase;
51 const GFC_INTEGER_4 * restrict bbase;
52 rtype_name * restrict dest;
62 const GFC_INTEGER_4 * restrict pa;
63 const GFC_INTEGER_4 * restrict pb;
69 assert (GFC_DESCRIPTOR_RANK (a) == 2
70 || GFC_DESCRIPTOR_RANK (b) == 2);
72 if (retarray->data == NULL)
74 if (GFC_DESCRIPTOR_RANK (a) == 1)
76 retarray->dim[0].lbound = 0;
77 retarray->dim[0].ubound = b->dim[1].ubound - b->dim[1].lbound;
78 retarray->dim[0].stride = 1;
80 else if (GFC_DESCRIPTOR_RANK (b) == 1)
82 retarray->dim[0].lbound = 0;
83 retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
84 retarray->dim[0].stride = 1;
88 retarray->dim[0].lbound = 0;
89 retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
90 retarray->dim[0].stride = 1;
92 retarray->dim[1].lbound = 0;
93 retarray->dim[1].ubound = b->dim[1].ubound - b->dim[1].lbound;
94 retarray->dim[1].stride = retarray->dim[0].ubound+1;
98 = internal_malloc_size (sizeof (rtype_name) * size0 ((array_t *) retarray));
103 if (GFC_DESCRIPTOR_SIZE (a) != 4)
105 assert (GFC_DESCRIPTOR_SIZE (a) == 8);
106 abase = GFOR_POINTER_L8_TO_L4 (abase);
109 if (GFC_DESCRIPTOR_SIZE (b) != 4)
111 assert (GFC_DESCRIPTOR_SIZE (b) == 8);
112 bbase = GFOR_POINTER_L8_TO_L4 (bbase);
114 dest = retarray->data;
116 sinclude(`matmul_asm_'rtype_code`.m4')dnl
118 if (GFC_DESCRIPTOR_RANK (retarray) == 1)
120 rxstride = retarray->dim[0].stride;
125 rxstride = retarray->dim[0].stride;
126 rystride = retarray->dim[1].stride;
129 /* If we have rank 1 parameters, zero the absent stride, and set the size to
131 if (GFC_DESCRIPTOR_RANK (a) == 1)
133 astride = a->dim[0].stride;
134 count = a->dim[0].ubound + 1 - a->dim[0].lbound;
141 astride = a->dim[1].stride;
142 count = a->dim[1].ubound + 1 - a->dim[1].lbound;
143 xstride = a->dim[0].stride;
144 xcount = a->dim[0].ubound + 1 - a->dim[0].lbound;
146 if (GFC_DESCRIPTOR_RANK (b) == 1)
148 bstride = b->dim[0].stride;
149 assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
156 bstride = b->dim[0].stride;
157 assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
158 ystride = b->dim[1].stride;
159 ycount = b->dim[1].ubound + 1 - b->dim[1].lbound;
162 for (y = 0; y < ycount; y++)
164 for (x = 0; x < xcount; x++)
166 /* Do the summation for this element. For real and integer types
167 this is the same as DOT_PRODUCT. For complex types we use do
168 a*b, not conjg(a)*b. */
173 for (n = 0; n < count; n++)
187 abase -= xstride * xcount;
189 dest += rystride - (rxstride * xcount);