FFmpeg
vp9dsp_init_16bpp_aarch64_template.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2017 Google Inc.
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 #include <stdint.h>
22 
23 #include "libavutil/attributes.h"
24 #include "libavutil/internal.h"
25 #include "libavutil/mem_internal.h"
26 #include "libavutil/aarch64/cpu.h"
27 #include "vp9dsp_init.h"
28 
29 #define declare_fpel(type, sz, suffix) \
30 void ff_vp9_##type##sz##suffix##_neon(uint8_t *dst, ptrdiff_t dst_stride, \
31  const uint8_t *src, ptrdiff_t src_stride, \
32  int h, int mx, int my)
33 
34 #define decl_mc_func(op, filter, dir, sz, bpp) \
35 void ff_vp9_##op##_##filter##sz##_##dir##_##bpp##_neon(uint8_t *dst, ptrdiff_t dst_stride, \
36  const uint8_t *src, ptrdiff_t src_stride, \
37  int h, int mx, int my)
38 
39 #define define_8tap_2d_fn(op, filter, sz, bpp) \
40 static void op##_##filter##sz##_hv_##bpp##_neon(uint8_t *dst, ptrdiff_t dst_stride, \
41  const uint8_t *src, \
42  ptrdiff_t src_stride, \
43  int h, int mx, int my) \
44 { \
45  LOCAL_ALIGNED_16(uint8_t, temp, [((1 + (sz < 64)) * sz + 8) * sz * 2]); \
46  /* We only need h + 7 lines, but the horizontal filter assumes an \
47  * even number of rows, so filter h + 8 lines here. */ \
48  ff_vp9_put_##filter##sz##_h_##bpp##_neon(temp, 2 * sz, \
49  src - 3 * src_stride, src_stride, \
50  h + 8, mx, 0); \
51  ff_vp9_##op##_##filter##sz##_v_##bpp##_neon(dst, dst_stride, \
52  temp + 3 * 2 * sz, 2 * sz, \
53  h, 0, my); \
54 }
55 
56 #define decl_filter_funcs(op, dir, sz, bpp) \
57  decl_mc_func(op, regular, dir, sz, bpp); \
58  decl_mc_func(op, sharp, dir, sz, bpp); \
59  decl_mc_func(op, smooth, dir, sz, bpp)
60 
61 #define decl_mc_funcs(sz, bpp) \
62  decl_filter_funcs(put, h, sz, bpp); \
63  decl_filter_funcs(avg, h, sz, bpp); \
64  decl_filter_funcs(put, v, sz, bpp); \
65  decl_filter_funcs(avg, v, sz, bpp); \
66  decl_filter_funcs(put, hv, sz, bpp); \
67  decl_filter_funcs(avg, hv, sz, bpp)
68 
69 #define ff_vp9_copy32_neon ff_vp9_copy32_aarch64
70 #define ff_vp9_copy64_neon ff_vp9_copy64_aarch64
71 #define ff_vp9_copy128_neon ff_vp9_copy128_aarch64
72 
73 declare_fpel(copy, 128, );
74 declare_fpel(copy, 64, );
75 declare_fpel(copy, 32, );
76 declare_fpel(copy, 16, );
77 declare_fpel(copy, 8, );
78 declare_fpel(avg, 64, _16);
79 declare_fpel(avg, 32, _16);
80 declare_fpel(avg, 16, _16);
81 declare_fpel(avg, 8, _16);
82 declare_fpel(avg, 4, _16);
83 
84 decl_mc_funcs(64, BPP);
85 decl_mc_funcs(32, BPP);
86 decl_mc_funcs(16, BPP);
87 decl_mc_funcs(8, BPP);
88 decl_mc_funcs(4, BPP);
89 
90 #define define_8tap_2d_funcs(sz, bpp) \
91  define_8tap_2d_fn(put, regular, sz, bpp) \
92  define_8tap_2d_fn(put, sharp, sz, bpp) \
93  define_8tap_2d_fn(put, smooth, sz, bpp) \
94  define_8tap_2d_fn(avg, regular, sz, bpp) \
95  define_8tap_2d_fn(avg, sharp, sz, bpp) \
96  define_8tap_2d_fn(avg, smooth, sz, bpp)
97 
103 
104 static av_cold void vp9dsp_mc_init_aarch64(VP9DSPContext *dsp)
105 {
106  int cpu_flags = av_get_cpu_flags();
107 
108 #define init_fpel(idx1, idx2, sz, type, suffix) \
109  dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] = \
110  dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = \
111  dsp->mc[idx1][FILTER_8TAP_SHARP ][idx2][0][0] = \
112  dsp->mc[idx1][FILTER_BILINEAR ][idx2][0][0] = ff_vp9_##type##sz##suffix
113 
114 #define init_copy(idx, sz, suffix) \
115  init_fpel(idx, 0, sz, copy, suffix)
116 
117 #define init_avg(idx, sz, suffix) \
118  init_fpel(idx, 1, sz, avg, suffix)
119 
120 #define init_copy_avg(idx, sz1, sz2) \
121  init_copy(idx, sz2, _neon); \
122  init_avg (idx, sz1, _16_neon)
123 
124  if (have_armv8(cpu_flags)) {
125  init_copy(0, 128, _aarch64);
126  init_copy(1, 64, _aarch64);
127  init_copy(2, 32, _aarch64);
128  }
129 
130  if (have_neon(cpu_flags)) {
131 #define init_mc_func(idx1, idx2, op, filter, fname, dir, mx, my, sz, pfx, bpp) \
132  dsp->mc[idx1][filter][idx2][mx][my] = pfx##op##_##fname##sz##_##dir##_##bpp##_neon
133 
134 #define init_mc_funcs(idx, dir, mx, my, sz, pfx, bpp) \
135  init_mc_func(idx, 0, put, FILTER_8TAP_REGULAR, regular, dir, mx, my, sz, pfx, bpp); \
136  init_mc_func(idx, 0, put, FILTER_8TAP_SHARP, sharp, dir, mx, my, sz, pfx, bpp); \
137  init_mc_func(idx, 0, put, FILTER_8TAP_SMOOTH, smooth, dir, mx, my, sz, pfx, bpp); \
138  init_mc_func(idx, 1, avg, FILTER_8TAP_REGULAR, regular, dir, mx, my, sz, pfx, bpp); \
139  init_mc_func(idx, 1, avg, FILTER_8TAP_SHARP, sharp, dir, mx, my, sz, pfx, bpp); \
140  init_mc_func(idx, 1, avg, FILTER_8TAP_SMOOTH, smooth, dir, mx, my, sz, pfx, bpp)
141 
142 #define init_mc_funcs_dirs(idx, sz, bpp) \
143  init_mc_funcs(idx, v, 0, 1, sz, ff_vp9_, bpp); \
144  init_mc_funcs(idx, h, 1, 0, sz, ff_vp9_, bpp); \
145  init_mc_funcs(idx, hv, 1, 1, sz, , bpp)
146 
147 
148  init_avg(0, 64, _16_neon);
149  init_avg(1, 32, _16_neon);
150  init_avg(2, 16, _16_neon);
151  init_copy_avg(3, 8, 16);
152  init_copy_avg(4, 4, 8);
153 
154  init_mc_funcs_dirs(0, 64, BPP);
155  init_mc_funcs_dirs(1, 32, BPP);
156  init_mc_funcs_dirs(2, 16, BPP);
157  init_mc_funcs_dirs(3, 8, BPP);
158  init_mc_funcs_dirs(4, 4, BPP);
159  }
160 }
161 
162 #define define_itxfm2(type_a, type_b, sz, bpp) \
163 void ff_vp9_##type_a##_##type_b##_##sz##x##sz##_add_##bpp##_neon(uint8_t *_dst, \
164  ptrdiff_t stride, \
165  int16_t *_block, int eob)
166 #define define_itxfm(type_a, type_b, sz, bpp) define_itxfm2(type_a, type_b, sz, bpp)
167 
168 #define define_itxfm_funcs(sz, bpp) \
169  define_itxfm(idct, idct, sz, bpp); \
170  define_itxfm(iadst, idct, sz, bpp); \
171  define_itxfm(idct, iadst, sz, bpp); \
172  define_itxfm(iadst, iadst, sz, bpp)
173 
177 define_itxfm(idct, idct, 32, BPP);
178 define_itxfm(iwht, iwht, 4, BPP);
179 
180 
182 {
183  int cpu_flags = av_get_cpu_flags();
184 
185  if (have_neon(cpu_flags)) {
186 #define init_itxfm2(tx, sz, bpp) \
187  dsp->itxfm_add[tx][DCT_DCT] = ff_vp9_idct_idct_##sz##_add_##bpp##_neon; \
188  dsp->itxfm_add[tx][DCT_ADST] = ff_vp9_iadst_idct_##sz##_add_##bpp##_neon; \
189  dsp->itxfm_add[tx][ADST_DCT] = ff_vp9_idct_iadst_##sz##_add_##bpp##_neon; \
190  dsp->itxfm_add[tx][ADST_ADST] = ff_vp9_iadst_iadst_##sz##_add_##bpp##_neon
191 #define init_itxfm(tx, sz, bpp) init_itxfm2(tx, sz, bpp)
192 
193 #define init_idct2(tx, nm, bpp) \
194  dsp->itxfm_add[tx][DCT_DCT] = \
195  dsp->itxfm_add[tx][ADST_DCT] = \
196  dsp->itxfm_add[tx][DCT_ADST] = \
197  dsp->itxfm_add[tx][ADST_ADST] = ff_vp9_##nm##_add_##bpp##_neon
198 #define init_idct(tx, nm, bpp) init_idct2(tx, nm, bpp)
199 
200  init_itxfm(TX_4X4, 4x4, BPP);
201  init_itxfm(TX_8X8, 8x8, BPP);
202  init_itxfm(TX_16X16, 16x16, BPP);
203  init_idct(TX_32X32, idct_idct_32x32, BPP);
204  init_idct(4, iwht_iwht_4x4, BPP);
205  }
206 }
207 
208 #define define_loop_filter(dir, wd, size, bpp) \
209 void ff_vp9_loop_filter_##dir##_##wd##_##size##_##bpp##_neon(uint8_t *dst, ptrdiff_t stride, int E, int I, int H)
210 
211 #define define_loop_filters(wd, size, bpp) \
212  define_loop_filter(h, wd, size, bpp); \
213  define_loop_filter(v, wd, size, bpp)
214 
215 define_loop_filters(4, 8, BPP);
216 define_loop_filters(8, 8, BPP);
217 define_loop_filters(16, 8, BPP);
218 
219 define_loop_filters(16, 16, BPP);
220 
221 define_loop_filters(44, 16, BPP);
222 define_loop_filters(48, 16, BPP);
223 define_loop_filters(84, 16, BPP);
224 define_loop_filters(88, 16, BPP);
225 
227 {
228  int cpu_flags = av_get_cpu_flags();
229 
230  if (have_neon(cpu_flags)) {
231 #define init_lpf_func_8(idx1, idx2, dir, wd, bpp) \
232  dsp->loop_filter_8[idx1][idx2] = ff_vp9_loop_filter_##dir##_##wd##_8_##bpp##_neon
233 
234 #define init_lpf_func_16(idx, dir, bpp) \
235  dsp->loop_filter_16[idx] = ff_vp9_loop_filter_##dir##_16_16_##bpp##_neon
236 
237 #define init_lpf_func_mix2(idx1, idx2, idx3, dir, wd, bpp) \
238  dsp->loop_filter_mix2[idx1][idx2][idx3] = ff_vp9_loop_filter_##dir##_##wd##_16_##bpp##_neon
239 
240 #define init_lpf_funcs_8_wd(idx, wd, bpp) \
241  init_lpf_func_8(idx, 0, h, wd, bpp); \
242  init_lpf_func_8(idx, 1, v, wd, bpp)
243 
244 #define init_lpf_funcs_16(bpp) \
245  init_lpf_func_16(0, h, bpp); \
246  init_lpf_func_16(1, v, bpp)
247 
248 #define init_lpf_funcs_mix2_wd(idx1, idx2, wd, bpp) \
249  init_lpf_func_mix2(idx1, idx2, 0, h, wd, bpp); \
250  init_lpf_func_mix2(idx1, idx2, 1, v, wd, bpp)
251 
252 #define init_lpf_funcs_8(bpp) \
253  init_lpf_funcs_8_wd(0, 4, bpp); \
254  init_lpf_funcs_8_wd(1, 8, bpp); \
255  init_lpf_funcs_8_wd(2, 16, bpp)
256 
257 #define init_lpf_funcs_mix2(bpp) \
258  init_lpf_funcs_mix2_wd(0, 0, 44, bpp); \
259  init_lpf_funcs_mix2_wd(0, 1, 48, bpp); \
260  init_lpf_funcs_mix2_wd(1, 0, 84, bpp); \
261  init_lpf_funcs_mix2_wd(1, 1, 88, bpp)
262 
266  }
267 }
268 
270 {
271  vp9dsp_mc_init_aarch64(dsp);
274 }
vp9dsp_loopfilter_init_aarch64
static av_cold void vp9dsp_loopfilter_init_aarch64(VP9DSPContext *dsp)
Definition: vp9dsp_init_16bpp_aarch64_template.c:225
mem_internal.h
VP9DSPContext
Definition: vp9dsp.h:40
define_loop_filters
#define define_loop_filters(wd, size, bpp)
Definition: vp9dsp_init_16bpp_aarch64_template.c:210
av_get_cpu_flags
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:103
cpu_flags
static atomic_int cpu_flags
Definition: cpu.c:52
INIT_FUNC
av_cold void INIT_FUNC(VP9DSPContext *dsp)
Definition: vp9dsp_init_16bpp_aarch64_template.c:268
idct
static void idct(int16_t block[64])
Definition: 4xm.c:166
av_cold
#define av_cold
Definition: attributes.h:90
decl_mc_funcs
#define decl_mc_funcs(sz, bpp)
Definition: vp9dsp_init_16bpp_aarch64_template.c:60
init_lpf_funcs_mix2
#define init_lpf_funcs_mix2(bpp)
init_avg
#define init_avg(idx, sz, suffix)
init_lpf_funcs_16
#define init_lpf_funcs_16(bpp)
BPP
#define BPP
Definition: vp9dsp_init_10bpp_aarch64.c:21
TX_8X8
@ TX_8X8
Definition: vp9.h:29
define_itxfm_funcs
#define define_itxfm_funcs(sz, bpp)
Definition: vp9dsp_init_16bpp_aarch64_template.c:167
TX_16X16
@ TX_16X16
Definition: vp9.h:30
init_idct
#define init_idct(tx, nm, bpp)
copy
static void copy(const float *p1, float *p2, const int length)
Definition: vf_vaguedenoiser.c:185
have_armv8
#define have_armv8(flags)
Definition: cpu.h:25
avg
#define avg(a, b, c, d)
Definition: colorspacedsp_template.c:28
TX_4X4
@ TX_4X4
Definition: vp9.h:28
have_neon
#define have_neon(flags)
Definition: cpu.h:26
init_mc_funcs_dirs
#define init_mc_funcs_dirs(idx, sz, bpp)
attributes.h
vp9dsp_init.h
define_itxfm
#define define_itxfm(type_a, type_b, sz, bpp)
Definition: vp9dsp_init_16bpp_aarch64_template.c:165
init_lpf_funcs_8
#define init_lpf_funcs_8(bpp)
internal.h
vp9dsp_itxfm_init_aarch64
static av_cold void vp9dsp_itxfm_init_aarch64(VP9DSPContext *dsp)
Definition: vp9dsp_init_16bpp_aarch64_template.c:180
define_8tap_2d_funcs
#define define_8tap_2d_funcs(sz, bpp)
Definition: vp9dsp_init_16bpp_aarch64_template.c:89
init_itxfm
#define init_itxfm(tx, sz, bpp)
init_copy_avg
#define init_copy_avg(idx, sz1, sz2)
TX_32X32
@ TX_32X32
Definition: vp9.h:31
init_copy
#define init_copy(idx, sz, suffix)
declare_fpel
#define declare_fpel(type, sz, suffix)
Definition: vp9dsp_init_16bpp_aarch64_template.c:29
cpu.h