]>
Commit | Line | Data |
---|---|---|
95f15c5b | 1 | `/* Helper function for cshift functions. |
6bc9506f | 2 | Copyright 2008, 2009 Free Software Foundation, Inc. |
95f15c5b | 3 | Contributed by Thomas Koenig <tkoenig@gcc.gnu.org> |
4 | ||
5 | This file is part of the GNU Fortran 95 runtime library (libgfortran). | |
6 | ||
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 | |
6bc9506f | 10 | version 3 of the License, or (at your option) any later version. |
95f15c5b | 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 | |
15 | GNU General Public License for more details. | |
16 | ||
6bc9506f | 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/>. */ | |
95f15c5b | 25 | |
26 | #include "libgfortran.h" | |
27 | #include <stdlib.h> | |
28 | #include <assert.h> | |
29 | #include <string.h>' | |
30 | ||
31 | include(iparm.m4)dnl | |
32 | ||
33 | `#if defined (HAVE_'rtype_name`) | |
34 | ||
35 | void | |
36 | cshift0_'rtype_code` ('rtype` *ret, const 'rtype` *array, ssize_t shift, | |
37 | int which) | |
38 | { | |
39 | /* r.* indicates the return array. */ | |
40 | index_type rstride[GFC_MAX_DIMENSIONS]; | |
41 | index_type rstride0; | |
42 | index_type roffset; | |
43 | 'rtype_name` *rptr; | |
44 | ||
45 | /* s.* indicates the source array. */ | |
46 | index_type sstride[GFC_MAX_DIMENSIONS]; | |
47 | index_type sstride0; | |
48 | index_type soffset; | |
49 | const 'rtype_name` *sptr; | |
50 | ||
51 | index_type count[GFC_MAX_DIMENSIONS]; | |
52 | index_type extent[GFC_MAX_DIMENSIONS]; | |
53 | index_type dim; | |
54 | index_type len; | |
55 | index_type n; | |
56 | ||
57 | which = which - 1; | |
58 | sstride[0] = 0; | |
59 | rstride[0] = 0; | |
60 | ||
61 | extent[0] = 1; | |
62 | count[0] = 0; | |
63 | n = 0; | |
64 | /* Initialized for avoiding compiler warnings. */ | |
65 | roffset = 1; | |
66 | soffset = 1; | |
67 | len = 0; | |
68 | ||
69 | for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++) | |
70 | { | |
71 | if (dim == which) | |
72 | { | |
827aef63 | 73 | roffset = GFC_DESCRIPTOR_STRIDE(ret,dim); |
95f15c5b | 74 | if (roffset == 0) |
75 | roffset = 1; | |
827aef63 | 76 | soffset = GFC_DESCRIPTOR_STRIDE(array,dim); |
95f15c5b | 77 | if (soffset == 0) |
78 | soffset = 1; | |
827aef63 | 79 | len = GFC_DESCRIPTOR_EXTENT(array,dim); |
95f15c5b | 80 | } |
81 | else | |
82 | { | |
83 | count[n] = 0; | |
827aef63 | 84 | extent[n] = GFC_DESCRIPTOR_EXTENT(array,dim); |
85 | rstride[n] = GFC_DESCRIPTOR_STRIDE(ret,dim); | |
86 | sstride[n] = GFC_DESCRIPTOR_STRIDE(array,dim); | |
95f15c5b | 87 | n++; |
88 | } | |
89 | } | |
90 | if (sstride[0] == 0) | |
91 | sstride[0] = 1; | |
92 | if (rstride[0] == 0) | |
93 | rstride[0] = 1; | |
94 | ||
95 | dim = GFC_DESCRIPTOR_RANK (array); | |
96 | rstride0 = rstride[0]; | |
97 | sstride0 = sstride[0]; | |
98 | rptr = ret->data; | |
99 | sptr = array->data; | |
100 | ||
101 | shift = len == 0 ? 0 : shift % (ssize_t)len; | |
102 | if (shift < 0) | |
103 | shift += len; | |
104 | ||
105 | while (rptr) | |
106 | { | |
107 | /* Do the shift for this dimension. */ | |
108 | ||
109 | /* If elements are contiguous, perform the operation | |
110 | in two block moves. */ | |
111 | if (soffset == 1 && roffset == 1) | |
112 | { | |
113 | size_t len1 = shift * sizeof ('rtype_name`); | |
114 | size_t len2 = (len - shift) * sizeof ('rtype_name`); | |
115 | memcpy (rptr, sptr + shift, len2); | |
116 | memcpy (rptr + (len - shift), sptr, len1); | |
117 | } | |
118 | else | |
119 | { | |
120 | /* Otherwise, we will have to perform the copy one element at | |
121 | a time. */ | |
122 | 'rtype_name` *dest = rptr; | |
123 | const 'rtype_name` *src = &sptr[shift * soffset]; | |
124 | ||
125 | for (n = 0; n < len - shift; n++) | |
126 | { | |
127 | *dest = *src; | |
128 | dest += roffset; | |
129 | src += soffset; | |
130 | } | |
131 | for (src = sptr, n = 0; n < shift; n++) | |
132 | { | |
133 | *dest = *src; | |
134 | dest += roffset; | |
135 | src += soffset; | |
136 | } | |
137 | } | |
138 | ||
139 | /* Advance to the next section. */ | |
140 | rptr += rstride0; | |
141 | sptr += sstride0; | |
142 | count[0]++; | |
143 | n = 0; | |
144 | while (count[n] == extent[n]) | |
145 | { | |
146 | /* When we get to the end of a dimension, reset it and increment | |
147 | the next dimension. */ | |
148 | count[n] = 0; | |
149 | /* We could precalculate these products, but this is a less | |
150 | frequently used path so probably not worth it. */ | |
151 | rptr -= rstride[n] * extent[n]; | |
152 | sptr -= sstride[n] * extent[n]; | |
153 | n++; | |
154 | if (n >= dim - 1) | |
155 | { | |
156 | /* Break out of the loop. */ | |
157 | rptr = NULL; | |
158 | break; | |
159 | } | |
160 | else | |
161 | { | |
162 | count[n]++; | |
163 | rptr += rstride[n]; | |
164 | sptr += sstride[n]; | |
165 | } | |
166 | } | |
167 | } | |
168 | ||
169 | return; | |
170 | } | |
171 | ||
172 | #endif' |