summaryrefslogtreecommitdiff
path: root/apps/codecs/demac/libdemac/vector_math16_armv5te.h
diff options
context:
space:
mode:
Diffstat (limited to 'apps/codecs/demac/libdemac/vector_math16_armv5te.h')
-rw-r--r--apps/codecs/demac/libdemac/vector_math16_armv5te.h202
1 files changed, 202 insertions, 0 deletions
diff --git a/apps/codecs/demac/libdemac/vector_math16_armv5te.h b/apps/codecs/demac/libdemac/vector_math16_armv5te.h
new file mode 100644
index 0000000000..fef99d9061
--- /dev/null
+++ b/apps/codecs/demac/libdemac/vector_math16_armv5te.h
@@ -0,0 +1,202 @@
1/*
2
3libdemac - A Monkey's Audio decoder
4
5$Id$
6
7Copyright (C) Dave Chapman 2007
8
9ARMv5te vector math copyright (C) 2008 Jens Arnold
10
11This program is free software; you can redistribute it and/or modify
12it under the terms of the GNU General Public License as published by
13the Free Software Foundation; either version 2 of the License, or
14(at your option) any later version.
15
16This program is distributed in the hope that it will be useful,
17but WITHOUT ANY WARRANTY; without even the implied warranty of
18MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
19GNU General Public License for more details.
20
21You should have received a copy of the GNU General Public License
22along with this program; if not, write to the Free Software
23Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110, USA
24
25*/
26
27static inline void vector_add(int16_t* v1, int16_t* v2)
28{
29#if ORDER > 32
30 int order = (ORDER >> 5);
31 while (order--)
32#endif
33 {
34 *v1++ += *v2++;
35 *v1++ += *v2++;
36 *v1++ += *v2++;
37 *v1++ += *v2++;
38 *v1++ += *v2++;
39 *v1++ += *v2++;
40 *v1++ += *v2++;
41 *v1++ += *v2++;
42 *v1++ += *v2++;
43 *v1++ += *v2++;
44 *v1++ += *v2++;
45 *v1++ += *v2++;
46 *v1++ += *v2++;
47 *v1++ += *v2++;
48 *v1++ += *v2++;
49 *v1++ += *v2++;
50#if ORDER > 16
51 *v1++ += *v2++;
52 *v1++ += *v2++;
53 *v1++ += *v2++;
54 *v1++ += *v2++;
55 *v1++ += *v2++;
56 *v1++ += *v2++;
57 *v1++ += *v2++;
58 *v1++ += *v2++;
59 *v1++ += *v2++;
60 *v1++ += *v2++;
61 *v1++ += *v2++;
62 *v1++ += *v2++;
63 *v1++ += *v2++;
64 *v1++ += *v2++;
65 *v1++ += *v2++;
66 *v1++ += *v2++;
67#endif
68 }
69}
70
71static inline void vector_sub(int16_t* v1, int16_t* v2)
72{
73#if ORDER > 32
74 int order = (ORDER >> 5);
75 while (order--)
76#endif
77 {
78 *v1++ -= *v2++;
79 *v1++ -= *v2++;
80 *v1++ -= *v2++;
81 *v1++ -= *v2++;
82 *v1++ -= *v2++;
83 *v1++ -= *v2++;
84 *v1++ -= *v2++;
85 *v1++ -= *v2++;
86 *v1++ -= *v2++;
87 *v1++ -= *v2++;
88 *v1++ -= *v2++;
89 *v1++ -= *v2++;
90 *v1++ -= *v2++;
91 *v1++ -= *v2++;
92 *v1++ -= *v2++;
93 *v1++ -= *v2++;
94#if ORDER > 16
95 *v1++ -= *v2++;
96 *v1++ -= *v2++;
97 *v1++ -= *v2++;
98 *v1++ -= *v2++;
99 *v1++ -= *v2++;
100 *v1++ -= *v2++;
101 *v1++ -= *v2++;
102 *v1++ -= *v2++;
103 *v1++ -= *v2++;
104 *v1++ -= *v2++;
105 *v1++ -= *v2++;
106 *v1++ -= *v2++;
107 *v1++ -= *v2++;
108 *v1++ -= *v2++;
109 *v1++ -= *v2++;
110 *v1++ -= *v2++;
111#endif
112 }
113}
114
115/* This version fetches data as 32 bit words, and *requires* v1 to be
116 * 32 bit aligned, otherwise it will result either in a data abort, or
117 * incorrect results (if ARM aligncheck is disabled). */
118static inline int32_t scalarproduct(int16_t* v1, int16_t* v2)
119{
120 int res = 0;
121#if ORDER > 16
122 int cnt = ORDER>>4;
123#endif
124
125 asm volatile (
126 "tst %[v2], #2 \n"
127 "beq 20f \n"
128
129 "10: \n"
130 "ldrh r4, [%[v2]], #2 \n"
131 "mov r4, r4, lsl #16 \n"
132 "1: \n"
133 "ldmia %[v1]!, {r0-r3} \n"
134 "ldmia %[v2]!, {r5-r8} \n"
135 "smlabt %[res], r0, r4, %[res] \n"
136 "smlatb %[res], r0, r5, %[res] \n"
137 "smlabt %[res], r1, r5, %[res] \n"
138 "smlatb %[res], r1, r6, %[res] \n"
139 "smlabt %[res], r2, r6, %[res] \n"
140 "smlatb %[res], r2, r7, %[res] \n"
141 "smlabt %[res], r3, r7, %[res] \n"
142 "smlatb %[res], r3, r8, %[res] \n"
143 "mov r4, r8 \n"
144 "ldmia %[v1]!, {r0-r3} \n"
145 "ldmia %[v2]!, {r5-r8} \n"
146 "smlabt %[res], r0, r4, %[res] \n"
147 "smlatb %[res], r0, r5, %[res] \n"
148 "smlabt %[res], r1, r5, %[res] \n"
149 "smlatb %[res], r1, r6, %[res] \n"
150 "smlabt %[res], r2, r6, %[res] \n"
151 "smlatb %[res], r2, r7, %[res] \n"
152 "smlabt %[res], r3, r7, %[res] \n"
153 "smlatb %[res], r3, r8, %[res] \n"
154#if ORDER > 16
155 "mov r4, r8 \n"
156 "subs %[cnt], %[cnt], #1 \n"
157 "bne 1b \n"
158#endif
159 "b 99f \n"
160
161 "20: \n"
162 "1: \n"
163 "ldmia %[v1]!, {r0-r3} \n"
164 "ldmia %[v2]!, {r4-r7} \n"
165 "smlabb %[res], r0, r4, %[res] \n"
166 "smlatt %[res], r0, r4, %[res] \n"
167 "smlabb %[res], r1, r5, %[res] \n"
168 "smlatt %[res], r1, r5, %[res] \n"
169 "smlabb %[res], r2, r6, %[res] \n"
170 "smlatt %[res], r2, r6, %[res] \n"
171 "smlabb %[res], r3, r7, %[res] \n"
172 "smlatt %[res], r3, r7, %[res] \n"
173 "ldmia %[v1]!, {r0-r3} \n"
174 "ldmia %[v2]!, {r4-r7} \n"
175 "smlabb %[res], r0, r4, %[res] \n"
176 "smlatt %[res], r0, r4, %[res] \n"
177 "smlabb %[res], r1, r5, %[res] \n"
178 "smlatt %[res], r1, r5, %[res] \n"
179 "smlabb %[res], r2, r6, %[res] \n"
180 "smlatt %[res], r2, r6, %[res] \n"
181 "smlabb %[res], r3, r7, %[res] \n"
182 "smlatt %[res], r3, r7, %[res] \n"
183#if ORDER > 16
184 "subs %[cnt], %[cnt], #1 \n"
185 "bne 1b \n"
186#endif
187
188 "99: \n"
189 : /* outputs */
190#if ORDER > 16
191 [cnt]"+r"(cnt),
192#endif
193 [v1] "+r"(v1),
194 [v2] "+r"(v2),
195 [res]"+r"(res)
196 : /* inputs */
197 : /* clobbers */
198 "r0", "r1", "r2", "r3", "r4",
199 "r5", "r6", "r7", "r8"
200 );
201 return res;
202}
generated by cgit v1.2.3 (git 2.39.1) at 2024-09-17 22:57:52 +0000