1/*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19#include <stdint.h>
20
21#include "libavutil/attributes.h"
22#include "libavutil/intreadwrite.h"
23#include "libavcodec/vc1dsp.h"
24#include "vc1dsp.h"
25
26void ff_vc1_inv_trans_8x8_neon(int16_t *block);
27void ff_vc1_inv_trans_4x8_neon(uint8_t *dest, ptrdiff_t stride, int16_t *block);
28void ff_vc1_inv_trans_8x4_neon(uint8_t *dest, ptrdiff_t stride, int16_t *block);
29void ff_vc1_inv_trans_4x4_neon(uint8_t *dest, ptrdiff_t stride, int16_t *block);
30
31void ff_vc1_inv_trans_8x8_dc_neon(uint8_t *dest, ptrdiff_t stride, int16_t *block);
32void ff_vc1_inv_trans_4x8_dc_neon(uint8_t *dest, ptrdiff_t stride, int16_t *block);
33void ff_vc1_inv_trans_8x4_dc_neon(uint8_t *dest, ptrdiff_t stride, int16_t *block);
34void ff_vc1_inv_trans_4x4_dc_neon(uint8_t *dest, ptrdiff_t stride, int16_t *block);
35
36void ff_vc1_v_loop_filter4_neon(uint8_t *src, int stride, int pq);
37void ff_vc1_h_loop_filter4_neon(uint8_t *src, int stride, int pq);
38void ff_vc1_v_loop_filter8_neon(uint8_t *src, int stride, int pq);
39void ff_vc1_h_loop_filter8_neon(uint8_t *src, int stride, int pq);
40void ff_vc1_v_loop_filter16_neon(uint8_t *src, int stride, int pq);
41void ff_vc1_h_loop_filter16_neon(uint8_t *src, int stride, int pq);
42
43void ff_put_pixels8x8_neon(uint8_t *block, const uint8_t *pixels,
44                           ptrdiff_t line_size, int rnd);
45
46#define DECL_PUT(X, Y) \
47void ff_put_vc1_mspel_mc##X##Y##_neon(uint8_t *dst, const uint8_t *src, \
48                                      ptrdiff_t stride, int rnd); \
49static void ff_put_vc1_mspel_mc##X##Y##_16_neon(uint8_t *dst, const uint8_t *src, \
50                                         ptrdiff_t stride, int rnd) \
51{ \
52  ff_put_vc1_mspel_mc##X##Y##_neon(dst+0, src+0, stride, rnd); \
53  ff_put_vc1_mspel_mc##X##Y##_neon(dst+8, src+8, stride, rnd); \
54  dst += 8*stride; src += 8*stride; \
55  ff_put_vc1_mspel_mc##X##Y##_neon(dst+0, src+0, stride, rnd); \
56  ff_put_vc1_mspel_mc##X##Y##_neon(dst+8, src+8, stride, rnd); \
57}
58
59DECL_PUT(1, 0)
60DECL_PUT(2, 0)
61DECL_PUT(3, 0)
62
63DECL_PUT(0, 1)
64DECL_PUT(0, 2)
65DECL_PUT(0, 3)
66
67DECL_PUT(1, 1)
68DECL_PUT(1, 2)
69DECL_PUT(1, 3)
70
71DECL_PUT(2, 1)
72DECL_PUT(2, 2)
73DECL_PUT(2, 3)
74
75DECL_PUT(3, 1)
76DECL_PUT(3, 2)
77DECL_PUT(3, 3)
78
79void ff_put_vc1_chroma_mc8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
80                                int h, int x, int y);
81void ff_avg_vc1_chroma_mc8_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
82                                int h, int x, int y);
83void ff_put_vc1_chroma_mc4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
84                                int h, int x, int y);
85void ff_avg_vc1_chroma_mc4_neon(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
86                                int h, int x, int y);
87
88int ff_vc1_unescape_buffer_helper_neon(const uint8_t *src, int size, uint8_t *dst);
89
90static int vc1_unescape_buffer_neon(const uint8_t *src, int size, uint8_t *dst)
91{
92    /* Dealing with starting and stopping, and removing escape bytes, are
93     * comparatively less time-sensitive, so are more clearly expressed using
94     * a C wrapper around the assembly inner loop. Note that we assume a
95     * little-endian machine that supports unaligned loads. */
96    int dsize = 0;
97    while (size >= 4)
98    {
99        int found = 0;
100        while (!found && (((uintptr_t) dst) & 7) && size >= 4)
101        {
102            found = (AV_RL32(src) &~ 0x03000000) == 0x00030000;
103            if (!found)
104            {
105                *dst++ = *src++;
106                --size;
107                ++dsize;
108            }
109        }
110        if (!found)
111        {
112            int skip = size - ff_vc1_unescape_buffer_helper_neon(src, size, dst);
113            dst += skip;
114            src += skip;
115            size -= skip;
116            dsize += skip;
117            while (!found && size >= 4)
118            {
119                found = (AV_RL32(src) &~ 0x03000000) == 0x00030000;
120                if (!found)
121                {
122                    *dst++ = *src++;
123                    --size;
124                    ++dsize;
125                }
126            }
127        }
128        if (found)
129        {
130            *dst++ = *src++;
131            *dst++ = *src++;
132            ++src;
133            size -= 3;
134            dsize += 2;
135        }
136    }
137    while (size > 0)
138    {
139        *dst++ = *src++;
140        --size;
141        ++dsize;
142    }
143    return dsize;
144}
145
146#define FN_ASSIGN(X, Y) \
147    dsp->put_vc1_mspel_pixels_tab[0][X+4*Y] = ff_put_vc1_mspel_mc##X##Y##_16_neon; \
148    dsp->put_vc1_mspel_pixels_tab[1][X+4*Y] = ff_put_vc1_mspel_mc##X##Y##_neon
149
150av_cold void ff_vc1dsp_init_neon(VC1DSPContext *dsp)
151{
152    dsp->vc1_inv_trans_8x8 = ff_vc1_inv_trans_8x8_neon;
153    dsp->vc1_inv_trans_4x8 = ff_vc1_inv_trans_4x8_neon;
154    dsp->vc1_inv_trans_8x4 = ff_vc1_inv_trans_8x4_neon;
155    dsp->vc1_inv_trans_4x4 = ff_vc1_inv_trans_4x4_neon;
156    dsp->vc1_inv_trans_8x8_dc = ff_vc1_inv_trans_8x8_dc_neon;
157    dsp->vc1_inv_trans_4x8_dc = ff_vc1_inv_trans_4x8_dc_neon;
158    dsp->vc1_inv_trans_8x4_dc = ff_vc1_inv_trans_8x4_dc_neon;
159    dsp->vc1_inv_trans_4x4_dc = ff_vc1_inv_trans_4x4_dc_neon;
160
161    dsp->vc1_v_loop_filter4  = ff_vc1_v_loop_filter4_neon;
162    dsp->vc1_h_loop_filter4  = ff_vc1_h_loop_filter4_neon;
163    dsp->vc1_v_loop_filter8  = ff_vc1_v_loop_filter8_neon;
164    dsp->vc1_h_loop_filter8  = ff_vc1_h_loop_filter8_neon;
165    dsp->vc1_v_loop_filter16 = ff_vc1_v_loop_filter16_neon;
166    dsp->vc1_h_loop_filter16 = ff_vc1_h_loop_filter16_neon;
167
168    dsp->put_vc1_mspel_pixels_tab[1][ 0] = ff_put_pixels8x8_neon;
169    FN_ASSIGN(1, 0);
170    FN_ASSIGN(2, 0);
171    FN_ASSIGN(3, 0);
172
173    FN_ASSIGN(0, 1);
174    FN_ASSIGN(1, 1);
175    FN_ASSIGN(2, 1);
176    FN_ASSIGN(3, 1);
177
178    FN_ASSIGN(0, 2);
179    FN_ASSIGN(1, 2);
180    FN_ASSIGN(2, 2);
181    FN_ASSIGN(3, 2);
182
183    FN_ASSIGN(0, 3);
184    FN_ASSIGN(1, 3);
185    FN_ASSIGN(2, 3);
186    FN_ASSIGN(3, 3);
187
188    dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = ff_put_vc1_chroma_mc8_neon;
189    dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_neon;
190    dsp->put_no_rnd_vc1_chroma_pixels_tab[1] = ff_put_vc1_chroma_mc4_neon;
191    dsp->avg_no_rnd_vc1_chroma_pixels_tab[1] = ff_avg_vc1_chroma_mc4_neon;
192
193    dsp->vc1_unescape_buffer = vc1_unescape_buffer_neon;
194}
195