101d3ca808e4d2f7bb330a0320be52176eab8e8c
[libav.git] / libavcodec / armv4l / dsputil_iwmmxt.c
1 /*
2 * iWMMXt optimized DSP utils
3 * Copyright (c) 2004 AGAWA Koji
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg 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 GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #include "libavcodec/dsputil.h"
23
24 #define DEF(x, y) x ## _no_rnd_ ## y ##_iwmmxt
25 #define SET_RND(regd) __asm__ volatile ("mov r12, #1 \n\t tbcsth " #regd ", r12":::"r12");
26 #define WAVG2B "wavg2b"
27 #include "dsputil_iwmmxt_rnd_template.c"
28 #undef DEF
29 #undef SET_RND
30 #undef WAVG2B
31
32 #define DEF(x, y) x ## _ ## y ##_iwmmxt
33 #define SET_RND(regd) __asm__ volatile ("mov r12, #2 \n\t tbcsth " #regd ", r12":::"r12");
34 #define WAVG2B "wavg2br"
35 #include "dsputil_iwmmxt_rnd_template.c"
36 #undef DEF
37 #undef SET_RND
38 #undef WAVG2BR
39
40 // need scheduling
41 #define OP(AVG) \
42 __asm__ volatile ( \
43 /* alignment */ \
44 "and r12, %[pixels], #7 \n\t" \
45 "bic %[pixels], %[pixels], #7 \n\t" \
46 "tmcr wcgr1, r12 \n\t" \
47 \
48 "wldrd wr0, [%[pixels]] \n\t" \
49 "wldrd wr1, [%[pixels], #8] \n\t" \
50 "add %[pixels], %[pixels], %[line_size] \n\t" \
51 "walignr1 wr4, wr0, wr1 \n\t" \
52 \
53 "1: \n\t" \
54 \
55 "wldrd wr2, [%[pixels]] \n\t" \
56 "wldrd wr3, [%[pixels], #8] \n\t" \
57 "add %[pixels], %[pixels], %[line_size] \n\t" \
58 "pld [%[pixels]] \n\t" \
59 "walignr1 wr5, wr2, wr3 \n\t" \
60 AVG " wr6, wr4, wr5 \n\t" \
61 "wstrd wr6, [%[block]] \n\t" \
62 "add %[block], %[block], %[line_size] \n\t" \
63 \
64 "wldrd wr0, [%[pixels]] \n\t" \
65 "wldrd wr1, [%[pixels], #8] \n\t" \
66 "add %[pixels], %[pixels], %[line_size] \n\t" \
67 "walignr1 wr4, wr0, wr1 \n\t" \
68 "pld [%[pixels]] \n\t" \
69 AVG " wr6, wr4, wr5 \n\t" \
70 "wstrd wr6, [%[block]] \n\t" \
71 "add %[block], %[block], %[line_size] \n\t" \
72 \
73 "subs %[h], %[h], #2 \n\t" \
74 "bne 1b \n\t" \
75 : [block]"+r"(block), [pixels]"+r"(pixels), [h]"+r"(h) \
76 : [line_size]"r"(line_size) \
77 : "memory", "r12");
78 void put_pixels8_y2_iwmmxt(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
79 {
80 OP("wavg2br");
81 }
82 void put_no_rnd_pixels8_y2_iwmmxt(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
83 {
84 OP("wavg2b");
85 }
86 #undef OP
87
88 void add_pixels_clamped_iwmmxt(const DCTELEM *block, uint8_t *pixels, int line_size)
89 {
90 uint8_t *pixels2 = pixels + line_size;
91
92 __asm__ volatile (
93 "mov r12, #4 \n\t"
94 "1: \n\t"
95 "pld [%[pixels], %[line_size2]] \n\t"
96 "pld [%[pixels2], %[line_size2]] \n\t"
97 "wldrd wr4, [%[pixels]] \n\t"
98 "wldrd wr5, [%[pixels2]] \n\t"
99 "pld [%[block], #32] \n\t"
100 "wunpckelub wr6, wr4 \n\t"
101 "wldrd wr0, [%[block]] \n\t"
102 "wunpckehub wr7, wr4 \n\t"
103 "wldrd wr1, [%[block], #8] \n\t"
104 "wunpckelub wr8, wr5 \n\t"
105 "wldrd wr2, [%[block], #16] \n\t"
106 "wunpckehub wr9, wr5 \n\t"
107 "wldrd wr3, [%[block], #24] \n\t"
108 "add %[block], %[block], #32 \n\t"
109 "waddhss wr10, wr0, wr6 \n\t"
110 "waddhss wr11, wr1, wr7 \n\t"
111 "waddhss wr12, wr2, wr8 \n\t"
112 "waddhss wr13, wr3, wr9 \n\t"
113 "wpackhus wr14, wr10, wr11 \n\t"
114 "wpackhus wr15, wr12, wr13 \n\t"
115 "wstrd wr14, [%[pixels]] \n\t"
116 "add %[pixels], %[pixels], %[line_size2] \n\t"
117 "subs r12, r12, #1 \n\t"
118 "wstrd wr15, [%[pixels2]] \n\t"
119 "add %[pixels2], %[pixels2], %[line_size2] \n\t"
120 "bne 1b \n\t"
121 : [block]"+r"(block), [pixels]"+r"(pixels), [pixels2]"+r"(pixels2)
122 : [line_size2]"r"(line_size << 1)
123 : "cc", "memory", "r12");
124 }
125
126 static void clear_blocks_iwmmxt(DCTELEM *blocks)
127 {
128 __asm__ volatile(
129 "wzero wr0 \n\t"
130 "mov r1, #(128 * 6 / 32) \n\t"
131 "1: \n\t"
132 "wstrd wr0, [%0] \n\t"
133 "wstrd wr0, [%0, #8] \n\t"
134 "wstrd wr0, [%0, #16] \n\t"
135 "wstrd wr0, [%0, #24] \n\t"
136 "subs r1, r1, #1 \n\t"
137 "add %0, %0, #32 \n\t"
138 "bne 1b \n\t"
139 : "+r"(blocks)
140 :
141 : "r1"
142 );
143 }
144
145 static void nop(uint8_t *block, const uint8_t *pixels, int line_size, int h)
146 {
147 return;
148 }
149
150 /* A run time test is not simple. If this file is compiled in
151 * then we should install the functions
152 */
153 int mm_flags = MM_IWMMXT; /* multimedia extension flags */
154
155 void dsputil_init_iwmmxt(DSPContext* c, AVCodecContext *avctx)
156 {
157 if (avctx->dsp_mask) {
158 if (avctx->dsp_mask & FF_MM_FORCE)
159 mm_flags |= (avctx->dsp_mask & 0xffff);
160 else
161 mm_flags &= ~(avctx->dsp_mask & 0xffff);
162 }
163
164 if (!(mm_flags & MM_IWMMXT)) return;
165
166 c->add_pixels_clamped = add_pixels_clamped_iwmmxt;
167
168 c->clear_blocks = clear_blocks_iwmmxt;
169
170 c->put_pixels_tab[0][0] = put_pixels16_iwmmxt;
171 c->put_pixels_tab[0][1] = put_pixels16_x2_iwmmxt;
172 c->put_pixels_tab[0][2] = put_pixels16_y2_iwmmxt;
173 c->put_pixels_tab[0][3] = put_pixels16_xy2_iwmmxt;
174 c->put_no_rnd_pixels_tab[0][0] = put_pixels16_iwmmxt;
175 c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_iwmmxt;
176 c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_iwmmxt;
177 c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_iwmmxt;
178
179 c->put_pixels_tab[1][0] = put_pixels8_iwmmxt;
180 c->put_pixels_tab[1][1] = put_pixels8_x2_iwmmxt;
181 c->put_pixels_tab[1][2] = put_pixels8_y2_iwmmxt;
182 c->put_pixels_tab[1][3] = put_pixels8_xy2_iwmmxt;
183 c->put_no_rnd_pixels_tab[1][0] = put_pixels8_iwmmxt;
184 c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_iwmmxt;
185 c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_iwmmxt;
186 c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy2_iwmmxt;
187
188 c->avg_pixels_tab[0][0] = avg_pixels16_iwmmxt;
189 c->avg_pixels_tab[0][1] = avg_pixels16_x2_iwmmxt;
190 c->avg_pixels_tab[0][2] = avg_pixels16_y2_iwmmxt;
191 c->avg_pixels_tab[0][3] = avg_pixels16_xy2_iwmmxt;
192 c->avg_no_rnd_pixels_tab[0][0] = avg_pixels16_iwmmxt;
193 c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x2_iwmmxt;
194 c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y2_iwmmxt;
195 c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy2_iwmmxt;
196
197 c->avg_pixels_tab[1][0] = avg_pixels8_iwmmxt;
198 c->avg_pixels_tab[1][1] = avg_pixels8_x2_iwmmxt;
199 c->avg_pixels_tab[1][2] = avg_pixels8_y2_iwmmxt;
200 c->avg_pixels_tab[1][3] = avg_pixels8_xy2_iwmmxt;
201 c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_iwmmxt;
202 c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x2_iwmmxt;
203 c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y2_iwmmxt;
204 c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy2_iwmmxt;
205 }