]>
Commit | Line | Data |
---|---|---|
644cb69f FXC |
1 | /* Implementation of the MATMUL intrinsic |
2 | Copyright 2002, 2005 Free Software Foundation, Inc. | |
3 | Contributed by Paul Brook <paul@nowt.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 | |
10 | version 2 of the License, or (at your option) any later version. | |
11 | ||
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 | |
19 | executable.) | |
20 | ||
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. | |
25 | ||
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. */ | |
30 | ||
31 | #include "config.h" | |
32 | #include <stdlib.h> | |
33 | #include <string.h> | |
34 | #include <assert.h> | |
35 | #include "libgfortran.h" | |
36 | ||
37 | #if defined (HAVE_GFC_COMPLEX_10) | |
38 | ||
39 | /* This is a C version of the following fortran pseudo-code. The key | |
40 | point is the loop order -- we access all arrays column-first, which | |
41 | improves the performance enough to boost galgel spec score by 50%. | |
42 | ||
43 | DIMENSION A(M,COUNT), B(COUNT,N), C(M,N) | |
44 | C = 0 | |
45 | DO J=1,N | |
46 | DO K=1,COUNT | |
47 | DO I=1,M | |
48 | C(I,J) = C(I,J)+A(I,K)*B(K,J) | |
49 | */ | |
50 | ||
51 | extern void matmul_c10 (gfc_array_c10 * retarray, gfc_array_c10 * a, gfc_array_c10 * b); | |
52 | export_proto(matmul_c10); | |
53 | ||
54 | void | |
55 | matmul_c10 (gfc_array_c10 * retarray, gfc_array_c10 * a, gfc_array_c10 * b) | |
56 | { | |
57 | GFC_COMPLEX_10 *abase; | |
58 | GFC_COMPLEX_10 *bbase; | |
59 | GFC_COMPLEX_10 *dest; | |
60 | ||
61 | index_type rxstride, rystride, axstride, aystride, bxstride, bystride; | |
62 | index_type x, y, n, count, xcount, ycount; | |
63 | ||
64 | assert (GFC_DESCRIPTOR_RANK (a) == 2 | |
65 | || GFC_DESCRIPTOR_RANK (b) == 2); | |
66 | ||
67 | /* C[xcount,ycount] = A[xcount, count] * B[count,ycount] | |
68 | ||
69 | Either A or B (but not both) can be rank 1: | |
70 | ||
71 | o One-dimensional argument A is implicitly treated as a row matrix | |
72 | dimensioned [1,count], so xcount=1. | |
73 | ||
74 | o One-dimensional argument B is implicitly treated as a column matrix | |
75 | dimensioned [count, 1], so ycount=1. | |
76 | */ | |
77 | ||
78 | if (retarray->data == NULL) | |
79 | { | |
80 | if (GFC_DESCRIPTOR_RANK (a) == 1) | |
81 | { | |
82 | retarray->dim[0].lbound = 0; | |
83 | retarray->dim[0].ubound = b->dim[1].ubound - b->dim[1].lbound; | |
84 | retarray->dim[0].stride = 1; | |
85 | } | |
86 | else if (GFC_DESCRIPTOR_RANK (b) == 1) | |
87 | { | |
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; | |
91 | } | |
92 | else | |
93 | { | |
94 | retarray->dim[0].lbound = 0; | |
95 | retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound; | |
96 | retarray->dim[0].stride = 1; | |
97 | ||
98 | retarray->dim[1].lbound = 0; | |
99 | retarray->dim[1].ubound = b->dim[1].ubound - b->dim[1].lbound; | |
100 | retarray->dim[1].stride = retarray->dim[0].ubound+1; | |
101 | } | |
102 | ||
103 | retarray->data | |
104 | = internal_malloc_size (sizeof (GFC_COMPLEX_10) * size0 ((array_t *) retarray)); | |
105 | retarray->offset = 0; | |
106 | } | |
107 | ||
108 | abase = a->data; | |
109 | bbase = b->data; | |
110 | dest = retarray->data; | |
111 | ||
112 | if (retarray->dim[0].stride == 0) | |
113 | retarray->dim[0].stride = 1; | |
114 | if (a->dim[0].stride == 0) | |
115 | a->dim[0].stride = 1; | |
116 | if (b->dim[0].stride == 0) | |
117 | b->dim[0].stride = 1; | |
118 | ||
119 | ||
120 | if (GFC_DESCRIPTOR_RANK (retarray) == 1) | |
121 | { | |
122 | /* One-dimensional result may be addressed in the code below | |
123 | either as a row or a column matrix. We want both cases to | |
124 | work. */ | |
125 | rxstride = rystride = retarray->dim[0].stride; | |
126 | } | |
127 | else | |
128 | { | |
129 | rxstride = retarray->dim[0].stride; | |
130 | rystride = retarray->dim[1].stride; | |
131 | } | |
132 | ||
133 | ||
134 | if (GFC_DESCRIPTOR_RANK (a) == 1) | |
135 | { | |
136 | /* Treat it as a a row matrix A[1,count]. */ | |
137 | axstride = a->dim[0].stride; | |
138 | aystride = 1; | |
139 | ||
140 | xcount = 1; | |
141 | count = a->dim[0].ubound + 1 - a->dim[0].lbound; | |
142 | } | |
143 | else | |
144 | { | |
145 | axstride = a->dim[0].stride; | |
146 | aystride = a->dim[1].stride; | |
147 | ||
148 | count = a->dim[1].ubound + 1 - a->dim[1].lbound; | |
149 | xcount = a->dim[0].ubound + 1 - a->dim[0].lbound; | |
150 | } | |
151 | ||
152 | assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound); | |
153 | ||
154 | if (GFC_DESCRIPTOR_RANK (b) == 1) | |
155 | { | |
156 | /* Treat it as a column matrix B[count,1] */ | |
157 | bxstride = b->dim[0].stride; | |
158 | ||
159 | /* bystride should never be used for 1-dimensional b. | |
160 | in case it is we want it to cause a segfault, rather than | |
161 | an incorrect result. */ | |
162 | bystride = 0xDEADBEEF; | |
163 | ycount = 1; | |
164 | } | |
165 | else | |
166 | { | |
167 | bxstride = b->dim[0].stride; | |
168 | bystride = b->dim[1].stride; | |
169 | ycount = b->dim[1].ubound + 1 - b->dim[1].lbound; | |
170 | } | |
171 | ||
172 | abase = a->data; | |
173 | bbase = b->data; | |
174 | dest = retarray->data; | |
175 | ||
176 | if (rxstride == 1 && axstride == 1 && bxstride == 1) | |
177 | { | |
178 | GFC_COMPLEX_10 *bbase_y; | |
179 | GFC_COMPLEX_10 *dest_y; | |
180 | GFC_COMPLEX_10 *abase_n; | |
181 | GFC_COMPLEX_10 bbase_yn; | |
182 | ||
183 | if (rystride == ycount) | |
184 | memset (dest, 0, (sizeof (GFC_COMPLEX_10) * size0((array_t *) retarray))); | |
185 | else | |
186 | { | |
187 | for (y = 0; y < ycount; y++) | |
188 | for (x = 0; x < xcount; x++) | |
189 | dest[x + y*rystride] = (GFC_COMPLEX_10)0; | |
190 | } | |
191 | ||
192 | for (y = 0; y < ycount; y++) | |
193 | { | |
194 | bbase_y = bbase + y*bystride; | |
195 | dest_y = dest + y*rystride; | |
196 | for (n = 0; n < count; n++) | |
197 | { | |
198 | abase_n = abase + n*aystride; | |
199 | bbase_yn = bbase_y[n]; | |
200 | for (x = 0; x < xcount; x++) | |
201 | { | |
202 | dest_y[x] += abase_n[x] * bbase_yn; | |
203 | } | |
204 | } | |
205 | } | |
206 | } | |
207 | else | |
208 | { | |
209 | for (y = 0; y < ycount; y++) | |
210 | for (x = 0; x < xcount; x++) | |
211 | dest[x*rxstride + y*rystride] = (GFC_COMPLEX_10)0; | |
212 | ||
213 | for (y = 0; y < ycount; y++) | |
214 | for (n = 0; n < count; n++) | |
215 | for (x = 0; x < xcount; x++) | |
216 | /* dest[x,y] += a[x,n] * b[n,y] */ | |
217 | dest[x*rxstride + y*rystride] += abase[x*axstride + n*aystride] * bbase[n*bxstride + y*bystride]; | |
218 | } | |
219 | } | |
220 | ||
221 | #endif |