1 /* Helper function for cshift functions.
2    Copyright (C) 2008-2018 Free Software Foundation, Inc.
3    Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>
4 
5 This file is part of the GNU Fortran 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
10 version 3 of the License, or (at your option) any later version.
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 
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/>.  */
25 
26 #include "libgfortran.h"
27 #include <string.h>
28 
29 
30 #if defined (HAVE_GFC_COMPLEX_4)
31 
32 void
cshift0_c4(gfc_array_c4 * ret,const gfc_array_c4 * array,ptrdiff_t shift,int which)33 cshift0_c4 (gfc_array_c4 *ret, const gfc_array_c4 *array, ptrdiff_t shift,
34 		     int which)
35 {
36   /* r.* indicates the return array.  */
37   index_type rstride[GFC_MAX_DIMENSIONS];
38   index_type rstride0;
39   index_type roffset;
40   GFC_COMPLEX_4 *rptr;
41 
42   /* s.* indicates the source array.  */
43   index_type sstride[GFC_MAX_DIMENSIONS];
44   index_type sstride0;
45   index_type soffset;
46   const GFC_COMPLEX_4 *sptr;
47 
48   index_type count[GFC_MAX_DIMENSIONS];
49   index_type extent[GFC_MAX_DIMENSIONS];
50   index_type dim;
51   index_type len;
52   index_type n;
53 
54   bool do_blocked;
55   index_type r_ex, a_ex;
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   r_ex = 1;
70   a_ex = 1;
71 
72   if (which > 0)
73     {
74       /* Test if both ret and array are contiguous.  */
75       do_blocked = true;
76       dim = GFC_DESCRIPTOR_RANK (array);
77       for (n = 0; n < dim; n ++)
78 	{
79 	  index_type rs, as;
80 	  rs = GFC_DESCRIPTOR_STRIDE (ret, n);
81 	  if (rs != r_ex)
82 	    {
83 	      do_blocked = false;
84 	      break;
85 	    }
86 	  as = GFC_DESCRIPTOR_STRIDE (array, n);
87 	  if (as != a_ex)
88 	    {
89 	      do_blocked = false;
90 	      break;
91 	    }
92 	  r_ex *= GFC_DESCRIPTOR_EXTENT (ret, n);
93 	  a_ex *= GFC_DESCRIPTOR_EXTENT (array, n);
94 	}
95     }
96   else
97     do_blocked = false;
98 
99   n = 0;
100 
101   if (do_blocked)
102     {
103       /* For contiguous arrays, use the relationship that
104 
105          dimension(n1,n2,n3) :: a, b
106 	 b = cshift(a,sh,3)
107 
108          can be dealt with as if
109 
110 	 dimension(n1*n2*n3) :: an, bn
111 	 bn = cshift(a,sh*n1*n2,1)
112 
113 	 we can used a more blocked algorithm for dim>1.  */
114       sstride[0] = 1;
115       rstride[0] = 1;
116       roffset = 1;
117       soffset = 1;
118       len = GFC_DESCRIPTOR_STRIDE(array, which)
119 	* GFC_DESCRIPTOR_EXTENT(array, which);
120       shift *= GFC_DESCRIPTOR_STRIDE(array, which);
121       for (dim = which + 1; dim < GFC_DESCRIPTOR_RANK (array); dim++)
122 	{
123 	  count[n] = 0;
124 	  extent[n] = GFC_DESCRIPTOR_EXTENT(array,dim);
125 	  rstride[n] = GFC_DESCRIPTOR_STRIDE(ret,dim);
126 	  sstride[n] = GFC_DESCRIPTOR_STRIDE(array,dim);
127 	  n++;
128 	}
129       dim = GFC_DESCRIPTOR_RANK (array) - which;
130     }
131   else
132     {
133       for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++)
134 	{
135 	  if (dim == which)
136 	    {
137 	      roffset = GFC_DESCRIPTOR_STRIDE(ret,dim);
138 	      if (roffset == 0)
139 		roffset = 1;
140 	      soffset = GFC_DESCRIPTOR_STRIDE(array,dim);
141 	      if (soffset == 0)
142 		soffset = 1;
143 	      len = GFC_DESCRIPTOR_EXTENT(array,dim);
144 	    }
145 	  else
146 	    {
147 	      count[n] = 0;
148 	      extent[n] = GFC_DESCRIPTOR_EXTENT(array,dim);
149 	      rstride[n] = GFC_DESCRIPTOR_STRIDE(ret,dim);
150 	      sstride[n] = GFC_DESCRIPTOR_STRIDE(array,dim);
151 	      n++;
152 	    }
153 	}
154       if (sstride[0] == 0)
155 	sstride[0] = 1;
156       if (rstride[0] == 0)
157 	rstride[0] = 1;
158 
159       dim = GFC_DESCRIPTOR_RANK (array);
160     }
161 
162   rstride0 = rstride[0];
163   sstride0 = sstride[0];
164   rptr = ret->base_addr;
165   sptr = array->base_addr;
166 
167   /* Avoid the costly modulo for trivially in-bound shifts.  */
168   if (shift < 0 || shift >= len)
169     {
170       shift = len == 0 ? 0 : shift % (ptrdiff_t)len;
171       if (shift < 0)
172 	shift += len;
173     }
174 
175   while (rptr)
176     {
177       /* Do the shift for this dimension.  */
178 
179       /* If elements are contiguous, perform the operation
180 	 in two block moves.  */
181       if (soffset == 1 && roffset == 1)
182 	{
183 	  size_t len1 = shift * sizeof (GFC_COMPLEX_4);
184 	  size_t len2 = (len - shift) * sizeof (GFC_COMPLEX_4);
185 	  memcpy (rptr, sptr + shift, len2);
186 	  memcpy (rptr + (len - shift), sptr, len1);
187 	}
188       else
189 	{
190 	  /* Otherwise, we will have to perform the copy one element at
191 	     a time.  */
192 	  GFC_COMPLEX_4 *dest = rptr;
193 	  const GFC_COMPLEX_4 *src = &sptr[shift * soffset];
194 
195 	  for (n = 0; n < len - shift; n++)
196 	    {
197 	      *dest = *src;
198 	      dest += roffset;
199 	      src += soffset;
200 	    }
201 	  for (src = sptr, n = 0; n < shift; n++)
202 	    {
203 	      *dest = *src;
204 	      dest += roffset;
205 	      src += soffset;
206 	    }
207 	}
208 
209       /* Advance to the next section.  */
210       rptr += rstride0;
211       sptr += sstride0;
212       count[0]++;
213       n = 0;
214       while (count[n] == extent[n])
215         {
216           /* When we get to the end of a dimension, reset it and increment
217              the next dimension.  */
218           count[n] = 0;
219           /* We could precalculate these products, but this is a less
220              frequently used path so probably not worth it.  */
221           rptr -= rstride[n] * extent[n];
222           sptr -= sstride[n] * extent[n];
223           n++;
224           if (n >= dim - 1)
225             {
226               /* Break out of the loop.  */
227               rptr = NULL;
228               break;
229             }
230           else
231             {
232               count[n]++;
233               rptr += rstride[n];
234               sptr += sstride[n];
235             }
236         }
237     }
238 
239   return;
240 }
241 
242 #endif
243