]>
Commit | Line | Data |
---|---|---|
6de9cd9a | 1 | /* Implementation of the MATMUL intrinsic |
e3c063ce | 2 | Copyright (C) 2002-2013 Free Software Foundation, Inc. |
6de9cd9a DN |
3 | Contributed by Paul Brook <paul@nowt.org> |
4 | ||
21d1335b | 5 | This file is part of the GNU Fortran runtime library (libgfortran). |
6de9cd9a DN |
6 | |
7 | Libgfortran is free software; you can redistribute it and/or | |
57dea9f6 | 8 | modify it under the terms of the GNU General Public |
6de9cd9a | 9 | License as published by the Free Software Foundation; either |
748086b7 | 10 | version 3 of the License, or (at your option) any later version. |
6de9cd9a DN |
11 | |
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 | |
57dea9f6 | 15 | GNU General Public License for more details. |
6de9cd9a | 16 | |
748086b7 JJ |
17 | Under Section 7 of GPL version 3, you are granted additional |
18 | permissions described in the GCC Runtime Library Exception, version | |
19 | 3.1, as published by the Free Software Foundation. | |
20 | ||
21 | You should have received a copy of the GNU General Public License and | |
22 | a copy of the GCC Runtime Library Exception along with this program; | |
23 | see the files COPYING3 and COPYING.RUNTIME respectively. If not, see | |
24 | <http://www.gnu.org/licenses/>. */ | |
6de9cd9a | 25 | |
36ae8a61 | 26 | #include "libgfortran.h" |
6de9cd9a DN |
27 | #include <stdlib.h> |
28 | #include <assert.h> | |
36ae8a61 | 29 | |
6de9cd9a | 30 | |
644cb69f FXC |
31 | #if defined (HAVE_GFC_LOGICAL_4) |
32 | ||
6de9cd9a DN |
33 | /* Dimensions: retarray(x,y) a(x, count) b(count,y). |
34 | Either a or b can be rank 1. In this case x or y is 1. */ | |
7d7b8bfe | 35 | |
85206901 | 36 | extern void matmul_l4 (gfc_array_l4 * const restrict, |
28dc6b33 | 37 | gfc_array_l1 * const restrict, gfc_array_l1 * const restrict); |
7f68c75f | 38 | export_proto(matmul_l4); |
7d7b8bfe | 39 | |
6de9cd9a | 40 | void |
85206901 | 41 | matmul_l4 (gfc_array_l4 * const restrict retarray, |
28dc6b33 | 42 | gfc_array_l1 * const restrict a, gfc_array_l1 * const restrict b) |
6de9cd9a | 43 | { |
28dc6b33 TK |
44 | const GFC_LOGICAL_1 * restrict abase; |
45 | const GFC_LOGICAL_1 * restrict bbase; | |
85206901 | 46 | GFC_LOGICAL_4 * restrict dest; |
6de9cd9a DN |
47 | index_type rxstride; |
48 | index_type rystride; | |
49 | index_type xcount; | |
50 | index_type ycount; | |
51 | index_type xstride; | |
52 | index_type ystride; | |
53 | index_type x; | |
54 | index_type y; | |
28dc6b33 TK |
55 | int a_kind; |
56 | int b_kind; | |
6de9cd9a | 57 | |
28dc6b33 TK |
58 | const GFC_LOGICAL_1 * restrict pa; |
59 | const GFC_LOGICAL_1 * restrict pb; | |
6de9cd9a DN |
60 | index_type astride; |
61 | index_type bstride; | |
62 | index_type count; | |
63 | index_type n; | |
64 | ||
65 | assert (GFC_DESCRIPTOR_RANK (a) == 2 | |
66 | || GFC_DESCRIPTOR_RANK (b) == 2); | |
883c9d4d | 67 | |
21d1335b | 68 | if (retarray->base_addr == NULL) |
883c9d4d VL |
69 | { |
70 | if (GFC_DESCRIPTOR_RANK (a) == 1) | |
71 | { | |
dfb55fdc TK |
72 | GFC_DIMENSION_SET(retarray->dim[0], 0, |
73 | GFC_DESCRIPTOR_EXTENT(b,1) - 1, 1); | |
883c9d4d VL |
74 | } |
75 | else if (GFC_DESCRIPTOR_RANK (b) == 1) | |
76 | { | |
dfb55fdc TK |
77 | GFC_DIMENSION_SET(retarray->dim[0], 0, |
78 | GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); | |
883c9d4d VL |
79 | } |
80 | else | |
81 | { | |
dfb55fdc TK |
82 | GFC_DIMENSION_SET(retarray->dim[0], 0, |
83 | GFC_DESCRIPTOR_EXTENT(a,0) - 1, 1); | |
84 | ||
85 | GFC_DIMENSION_SET(retarray->dim[1], 0, | |
86 | GFC_DESCRIPTOR_EXTENT(b,1) - 1, | |
87 | GFC_DESCRIPTOR_EXTENT(retarray,0)); | |
883c9d4d | 88 | } |
efd4dc1a | 89 | |
21d1335b | 90 | retarray->base_addr |
1a0fd3d3 | 91 | = xmalloc (sizeof (GFC_LOGICAL_4) * size0 ((array_t *) retarray)); |
efd4dc1a | 92 | retarray->offset = 0; |
883c9d4d | 93 | } |
9731c4a3 | 94 | else if (unlikely (compile_options.bounds_check)) |
9ad13e91 TK |
95 | { |
96 | index_type ret_extent, arg_extent; | |
97 | ||
98 | if (GFC_DESCRIPTOR_RANK (a) == 1) | |
99 | { | |
dfb55fdc TK |
100 | arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); |
101 | ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); | |
9ad13e91 TK |
102 | if (arg_extent != ret_extent) |
103 | runtime_error ("Incorrect extent in return array in" | |
104 | " MATMUL intrinsic: is %ld, should be %ld", | |
105 | (long int) ret_extent, (long int) arg_extent); | |
106 | } | |
107 | else if (GFC_DESCRIPTOR_RANK (b) == 1) | |
108 | { | |
dfb55fdc TK |
109 | arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); |
110 | ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); | |
9ad13e91 TK |
111 | if (arg_extent != ret_extent) |
112 | runtime_error ("Incorrect extent in return array in" | |
113 | " MATMUL intrinsic: is %ld, should be %ld", | |
114 | (long int) ret_extent, (long int) arg_extent); | |
115 | } | |
116 | else | |
117 | { | |
dfb55fdc TK |
118 | arg_extent = GFC_DESCRIPTOR_EXTENT(a,0); |
119 | ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,0); | |
9ad13e91 TK |
120 | if (arg_extent != ret_extent) |
121 | runtime_error ("Incorrect extent in return array in" | |
122 | " MATMUL intrinsic for dimension 1:" | |
123 | " is %ld, should be %ld", | |
124 | (long int) ret_extent, (long int) arg_extent); | |
125 | ||
dfb55fdc TK |
126 | arg_extent = GFC_DESCRIPTOR_EXTENT(b,1); |
127 | ret_extent = GFC_DESCRIPTOR_EXTENT(retarray,1); | |
9ad13e91 TK |
128 | if (arg_extent != ret_extent) |
129 | runtime_error ("Incorrect extent in return array in" | |
130 | " MATMUL intrinsic for dimension 2:" | |
131 | " is %ld, should be %ld", | |
132 | (long int) ret_extent, (long int) arg_extent); | |
133 | } | |
134 | } | |
883c9d4d | 135 | |
21d1335b | 136 | abase = a->base_addr; |
28dc6b33 TK |
137 | a_kind = GFC_DESCRIPTOR_SIZE (a); |
138 | ||
139 | if (a_kind == 1 || a_kind == 2 || a_kind == 4 || a_kind == 8 | |
140 | #ifdef HAVE_GFC_LOGICAL_16 | |
141 | || a_kind == 16 | |
142 | #endif | |
143 | ) | |
144 | abase = GFOR_POINTER_TO_L1 (abase, a_kind); | |
145 | else | |
146 | internal_error (NULL, "Funny sized logical array"); | |
147 | ||
21d1335b | 148 | bbase = b->base_addr; |
28dc6b33 TK |
149 | b_kind = GFC_DESCRIPTOR_SIZE (b); |
150 | ||
151 | if (b_kind == 1 || b_kind == 2 || b_kind == 4 || b_kind == 8 | |
152 | #ifdef HAVE_GFC_LOGICAL_16 | |
153 | || b_kind == 16 | |
154 | #endif | |
155 | ) | |
156 | bbase = GFOR_POINTER_TO_L1 (bbase, b_kind); | |
157 | else | |
158 | internal_error (NULL, "Funny sized logical array"); | |
159 | ||
21d1335b | 160 | dest = retarray->base_addr; |
6de9cd9a | 161 | |
6de9cd9a DN |
162 | |
163 | if (GFC_DESCRIPTOR_RANK (retarray) == 1) | |
164 | { | |
dfb55fdc | 165 | rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); |
6de9cd9a DN |
166 | rystride = rxstride; |
167 | } | |
168 | else | |
169 | { | |
dfb55fdc TK |
170 | rxstride = GFC_DESCRIPTOR_STRIDE(retarray,0); |
171 | rystride = GFC_DESCRIPTOR_STRIDE(retarray,1); | |
6de9cd9a DN |
172 | } |
173 | ||
174 | /* If we have rank 1 parameters, zero the absent stride, and set the size to | |
175 | one. */ | |
176 | if (GFC_DESCRIPTOR_RANK (a) == 1) | |
177 | { | |
dfb55fdc TK |
178 | astride = GFC_DESCRIPTOR_STRIDE_BYTES(a,0); |
179 | count = GFC_DESCRIPTOR_EXTENT(a,0); | |
6de9cd9a DN |
180 | xstride = 0; |
181 | rxstride = 0; | |
182 | xcount = 1; | |
183 | } | |
184 | else | |
185 | { | |
dfb55fdc TK |
186 | astride = GFC_DESCRIPTOR_STRIDE_BYTES(a,1); |
187 | count = GFC_DESCRIPTOR_EXTENT(a,1); | |
188 | xstride = GFC_DESCRIPTOR_STRIDE_BYTES(a,0); | |
189 | xcount = GFC_DESCRIPTOR_EXTENT(a,0); | |
6de9cd9a DN |
190 | } |
191 | if (GFC_DESCRIPTOR_RANK (b) == 1) | |
192 | { | |
dfb55fdc TK |
193 | bstride = GFC_DESCRIPTOR_STRIDE_BYTES(b,0); |
194 | assert(count == GFC_DESCRIPTOR_EXTENT(b,0)); | |
6de9cd9a DN |
195 | ystride = 0; |
196 | rystride = 0; | |
197 | ycount = 1; | |
198 | } | |
199 | else | |
200 | { | |
dfb55fdc TK |
201 | bstride = GFC_DESCRIPTOR_STRIDE_BYTES(b,0); |
202 | assert(count == GFC_DESCRIPTOR_EXTENT(b,0)); | |
203 | ystride = GFC_DESCRIPTOR_STRIDE_BYTES(b,1); | |
204 | ycount = GFC_DESCRIPTOR_EXTENT(b,1); | |
6de9cd9a DN |
205 | } |
206 | ||
207 | for (y = 0; y < ycount; y++) | |
208 | { | |
209 | for (x = 0; x < xcount; x++) | |
210 | { | |
211 | /* Do the summation for this element. For real and integer types | |
212 | this is the same as DOT_PRODUCT. For complex types we use do | |
213 | a*b, not conjg(a)*b. */ | |
214 | pa = abase; | |
215 | pb = bbase; | |
216 | *dest = 0; | |
217 | ||
218 | for (n = 0; n < count; n++) | |
219 | { | |
220 | if (*pa && *pb) | |
221 | { | |
222 | *dest = 1; | |
223 | break; | |
224 | } | |
225 | pa += astride; | |
226 | pb += bstride; | |
227 | } | |
228 | ||
229 | dest += rxstride; | |
230 | abase += xstride; | |
231 | } | |
232 | abase -= xstride * xcount; | |
233 | bbase += ystride; | |
234 | dest += rystride - (rxstride * xcount); | |
235 | } | |
236 | } | |
644cb69f FXC |
237 | |
238 | #endif | |
28dc6b33 | 239 |