|
111
|
1 /* Special implementation of the SPREAD intrinsic
|
|
145
|
2 Copyright (C) 2008-2020 Free Software Foundation, Inc.
|
|
111
|
3 Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>, based on
|
|
|
4 spread_generic.c written by Paul Brook <paul@nowt.org>
|
|
|
5
|
|
|
6 This file is part of the GNU Fortran runtime library (libgfortran).
|
|
|
7
|
|
|
8 Libgfortran is free software; you can redistribute it and/or
|
|
|
9 modify it under the terms of the GNU General Public
|
|
|
10 License as published by the Free Software Foundation; either
|
|
|
11 version 3 of the License, or (at your option) any later version.
|
|
|
12
|
|
|
13 Ligbfortran is distributed in the hope that it will be useful,
|
|
|
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
16 GNU General Public License for more details.
|
|
|
17
|
|
|
18 Under Section 7 of GPL version 3, you are granted additional
|
|
|
19 permissions described in the GCC Runtime Library Exception, version
|
|
|
20 3.1, as published by the Free Software Foundation.
|
|
|
21
|
|
|
22 You should have received a copy of the GNU General Public License and
|
|
|
23 a copy of the GCC Runtime Library Exception along with this program;
|
|
|
24 see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
|
|
25 <http://www.gnu.org/licenses/>. */
|
|
|
26
|
|
|
27 #include "libgfortran.h"
|
|
|
28 #include <string.h>
|
|
|
29
|
|
|
30
|
|
|
31 #if defined (HAVE_GFC_INTEGER_4)
|
|
|
32
|
|
|
33 void
|
|
|
34 spread_i4 (gfc_array_i4 *ret, const gfc_array_i4 *source,
|
|
|
35 const index_type along, const index_type pncopies)
|
|
|
36 {
|
|
|
37 /* r.* indicates the return array. */
|
|
|
38 index_type rstride[GFC_MAX_DIMENSIONS];
|
|
|
39 index_type rstride0;
|
|
|
40 index_type rdelta = 0;
|
|
|
41 index_type rrank;
|
|
|
42 index_type rs;
|
|
|
43 GFC_INTEGER_4 *rptr;
|
|
|
44 GFC_INTEGER_4 * restrict dest;
|
|
|
45 /* s.* indicates the source array. */
|
|
|
46 index_type sstride[GFC_MAX_DIMENSIONS];
|
|
|
47 index_type sstride0;
|
|
|
48 index_type srank;
|
|
|
49 const GFC_INTEGER_4 *sptr;
|
|
|
50
|
|
|
51 index_type count[GFC_MAX_DIMENSIONS];
|
|
|
52 index_type extent[GFC_MAX_DIMENSIONS];
|
|
|
53 index_type n;
|
|
|
54 index_type dim;
|
|
|
55 index_type ncopies;
|
|
|
56
|
|
|
57 srank = GFC_DESCRIPTOR_RANK(source);
|
|
|
58
|
|
|
59 rrank = srank + 1;
|
|
|
60 if (rrank > GFC_MAX_DIMENSIONS)
|
|
|
61 runtime_error ("return rank too large in spread()");
|
|
|
62
|
|
|
63 if (along > rrank)
|
|
|
64 runtime_error ("dim outside of rank in spread()");
|
|
|
65
|
|
|
66 ncopies = pncopies;
|
|
|
67
|
|
|
68 if (ret->base_addr == NULL)
|
|
|
69 {
|
|
|
70
|
|
|
71 size_t ub, stride;
|
|
|
72
|
|
|
73 /* The front end has signalled that we need to populate the
|
|
|
74 return array descriptor. */
|
|
131
|
75 ret->dtype.rank = rrank;
|
|
|
76
|
|
111
|
77 dim = 0;
|
|
|
78 rs = 1;
|
|
|
79 for (n = 0; n < rrank; n++)
|
|
|
80 {
|
|
|
81 stride = rs;
|
|
|
82 if (n == along - 1)
|
|
|
83 {
|
|
|
84 ub = ncopies - 1;
|
|
|
85 rdelta = rs;
|
|
|
86 rs *= ncopies;
|
|
|
87 }
|
|
|
88 else
|
|
|
89 {
|
|
|
90 count[dim] = 0;
|
|
|
91 extent[dim] = GFC_DESCRIPTOR_EXTENT(source,dim);
|
|
|
92 sstride[dim] = GFC_DESCRIPTOR_STRIDE(source,dim);
|
|
|
93 rstride[dim] = rs;
|
|
|
94
|
|
|
95 ub = extent[dim] - 1;
|
|
|
96 rs *= extent[dim];
|
|
|
97 dim++;
|
|
|
98 }
|
|
|
99 GFC_DIMENSION_SET(ret->dim[n], 0, ub, stride);
|
|
|
100 }
|
|
|
101 ret->offset = 0;
|
|
|
102
|
|
|
103 /* xmallocarray allocates a single byte for zero size. */
|
|
|
104 ret->base_addr = xmallocarray (rs, sizeof(GFC_INTEGER_4));
|
|
|
105 if (rs <= 0)
|
|
|
106 return;
|
|
|
107 }
|
|
|
108 else
|
|
|
109 {
|
|
|
110 int zero_sized;
|
|
|
111
|
|
|
112 zero_sized = 0;
|
|
|
113
|
|
|
114 dim = 0;
|
|
|
115 if (GFC_DESCRIPTOR_RANK(ret) != rrank)
|
|
|
116 runtime_error ("rank mismatch in spread()");
|
|
|
117
|
|
|
118 if (unlikely (compile_options.bounds_check))
|
|
|
119 {
|
|
|
120 for (n = 0; n < rrank; n++)
|
|
|
121 {
|
|
|
122 index_type ret_extent;
|
|
|
123
|
|
|
124 ret_extent = GFC_DESCRIPTOR_EXTENT(ret,n);
|
|
|
125 if (n == along - 1)
|
|
|
126 {
|
|
|
127 rdelta = GFC_DESCRIPTOR_STRIDE(ret,n);
|
|
|
128
|
|
|
129 if (ret_extent != ncopies)
|
|
|
130 runtime_error("Incorrect extent in return value of SPREAD"
|
|
|
131 " intrinsic in dimension %ld: is %ld,"
|
|
|
132 " should be %ld", (long int) n+1,
|
|
|
133 (long int) ret_extent, (long int) ncopies);
|
|
|
134 }
|
|
|
135 else
|
|
|
136 {
|
|
|
137 count[dim] = 0;
|
|
|
138 extent[dim] = GFC_DESCRIPTOR_EXTENT(source,dim);
|
|
|
139 if (ret_extent != extent[dim])
|
|
|
140 runtime_error("Incorrect extent in return value of SPREAD"
|
|
|
141 " intrinsic in dimension %ld: is %ld,"
|
|
|
142 " should be %ld", (long int) n+1,
|
|
|
143 (long int) ret_extent,
|
|
|
144 (long int) extent[dim]);
|
|
|
145
|
|
|
146 if (extent[dim] <= 0)
|
|
|
147 zero_sized = 1;
|
|
|
148 sstride[dim] = GFC_DESCRIPTOR_STRIDE(source,dim);
|
|
|
149 rstride[dim] = GFC_DESCRIPTOR_STRIDE(ret,n);
|
|
|
150 dim++;
|
|
|
151 }
|
|
|
152 }
|
|
|
153 }
|
|
|
154 else
|
|
|
155 {
|
|
|
156 for (n = 0; n < rrank; n++)
|
|
|
157 {
|
|
|
158 if (n == along - 1)
|
|
|
159 {
|
|
|
160 rdelta = GFC_DESCRIPTOR_STRIDE(ret,n);
|
|
|
161 }
|
|
|
162 else
|
|
|
163 {
|
|
|
164 count[dim] = 0;
|
|
|
165 extent[dim] = GFC_DESCRIPTOR_EXTENT(source,dim);
|
|
|
166 if (extent[dim] <= 0)
|
|
|
167 zero_sized = 1;
|
|
|
168 sstride[dim] = GFC_DESCRIPTOR_STRIDE(source,dim);
|
|
|
169 rstride[dim] = GFC_DESCRIPTOR_STRIDE(ret,n);
|
|
|
170 dim++;
|
|
|
171 }
|
|
|
172 }
|
|
|
173 }
|
|
|
174
|
|
|
175 if (zero_sized)
|
|
|
176 return;
|
|
|
177
|
|
|
178 if (sstride[0] == 0)
|
|
|
179 sstride[0] = 1;
|
|
|
180 }
|
|
|
181 sstride0 = sstride[0];
|
|
|
182 rstride0 = rstride[0];
|
|
|
183 rptr = ret->base_addr;
|
|
|
184 sptr = source->base_addr;
|
|
|
185
|
|
|
186 while (sptr)
|
|
|
187 {
|
|
|
188 /* Spread this element. */
|
|
|
189 dest = rptr;
|
|
|
190 for (n = 0; n < ncopies; n++)
|
|
|
191 {
|
|
|
192 *dest = *sptr;
|
|
|
193 dest += rdelta;
|
|
|
194 }
|
|
|
195 /* Advance to the next element. */
|
|
|
196 sptr += sstride0;
|
|
|
197 rptr += rstride0;
|
|
|
198 count[0]++;
|
|
|
199 n = 0;
|
|
|
200 while (count[n] == extent[n])
|
|
|
201 {
|
|
|
202 /* When we get to the end of a dimension, reset it and increment
|
|
|
203 the next dimension. */
|
|
|
204 count[n] = 0;
|
|
|
205 /* We could precalculate these products, but this is a less
|
|
|
206 frequently used path so probably not worth it. */
|
|
|
207 sptr -= sstride[n] * extent[n];
|
|
|
208 rptr -= rstride[n] * extent[n];
|
|
|
209 n++;
|
|
|
210 if (n >= srank)
|
|
|
211 {
|
|
|
212 /* Break out of the loop. */
|
|
|
213 sptr = NULL;
|
|
|
214 break;
|
|
|
215 }
|
|
|
216 else
|
|
|
217 {
|
|
|
218 count[n]++;
|
|
|
219 sptr += sstride[n];
|
|
|
220 rptr += rstride[n];
|
|
|
221 }
|
|
|
222 }
|
|
|
223 }
|
|
|
224 }
|
|
|
225
|
|
|
226 /* This version of spread_internal treats the special case of a scalar
|
|
|
227 source. This is much simpler than the more general case above. */
|
|
|
228
|
|
|
229 void
|
|
|
230 spread_scalar_i4 (gfc_array_i4 *ret, const GFC_INTEGER_4 *source,
|
|
131
|
231 const index_type along, const index_type ncopies)
|
|
111
|
232 {
|
|
|
233 GFC_INTEGER_4 * restrict dest;
|
|
|
234 index_type stride;
|
|
|
235
|
|
|
236 if (GFC_DESCRIPTOR_RANK (ret) != 1)
|
|
|
237 runtime_error ("incorrect destination rank in spread()");
|
|
|
238
|
|
|
239 if (along > 1)
|
|
|
240 runtime_error ("dim outside of rank in spread()");
|
|
|
241
|
|
|
242 if (ret->base_addr == NULL)
|
|
|
243 {
|
|
|
244 ret->base_addr = xmallocarray (ncopies, sizeof (GFC_INTEGER_4));
|
|
|
245 ret->offset = 0;
|
|
|
246 GFC_DIMENSION_SET(ret->dim[0], 0, ncopies - 1, 1);
|
|
|
247 }
|
|
|
248 else
|
|
|
249 {
|
|
|
250 if (ncopies - 1 > (GFC_DESCRIPTOR_EXTENT(ret,0) - 1)
|
|
|
251 / GFC_DESCRIPTOR_STRIDE(ret,0))
|
|
|
252 runtime_error ("dim too large in spread()");
|
|
|
253 }
|
|
|
254
|
|
|
255 dest = ret->base_addr;
|
|
|
256 stride = GFC_DESCRIPTOR_STRIDE(ret,0);
|
|
|
257
|
|
131
|
258 for (index_type n = 0; n < ncopies; n++)
|
|
111
|
259 {
|
|
|
260 *dest = *source;
|
|
|
261 dest += stride;
|
|
|
262 }
|
|
|
263 }
|
|
|
264
|
|
|
265 #endif
|
|
|
266
|
