FFmpeg
mpegvideo.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2010 Mans Rullgard
3  * Copyright (c) 2014 James Yu <james.yu@linaro.org>
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 <arm_neon.h>
23 
24 #include "config.h"
25 
26 #include "libavutil/attributes.h"
27 #include "libavutil/cpu.h"
28 #if ARCH_AARCH64
29 # include "libavutil/aarch64/cpu.h"
30 #elif ARCH_ARM
31 # include "libavutil/arm/cpu.h"
32 #endif
33 
34 #include "libavcodec/mpegvideo.h"
35 
36 static void inline ff_dct_unquantize_h263_neon(int qscale, int qadd, int nCoeffs,
37  int16_t *block)
38 {
39  int16x8_t q0s16, q2s16, q3s16, q8s16, q10s16, q11s16, q13s16;
40  int16x8_t q14s16, q15s16, qzs16;
41  int16x4_t d0s16, d2s16, d3s16, dzs16;
42  uint16x8_t q1u16, q9u16;
43  uint16x4_t d1u16;
44 
45  dzs16 = vdup_n_s16(0);
46  qzs16 = vdupq_n_s16(0);
47 
48  q15s16 = vdupq_n_s16(qscale << 1);
49  q14s16 = vdupq_n_s16(qadd);
50  q13s16 = vnegq_s16(q14s16);
51 
52  if (nCoeffs > 4) {
53  for (; nCoeffs > 8; nCoeffs -= 16, block += 16) {
54  q0s16 = vld1q_s16(block);
55  q3s16 = vreinterpretq_s16_u16(vcltq_s16(q0s16, qzs16));
56  q8s16 = vld1q_s16(block + 8);
57  q1u16 = vceqq_s16(q0s16, qzs16);
58  q2s16 = vmulq_s16(q0s16, q15s16);
59  q11s16 = vreinterpretq_s16_u16(vcltq_s16(q8s16, qzs16));
60  q10s16 = vmulq_s16(q8s16, q15s16);
61  q3s16 = vbslq_s16(vreinterpretq_u16_s16(q3s16), q13s16, q14s16);
62  q11s16 = vbslq_s16(vreinterpretq_u16_s16(q11s16), q13s16, q14s16);
63  q2s16 = vaddq_s16(q2s16, q3s16);
64  q9u16 = vceqq_s16(q8s16, qzs16);
65  q10s16 = vaddq_s16(q10s16, q11s16);
66  q0s16 = vbslq_s16(q1u16, q0s16, q2s16);
67  q8s16 = vbslq_s16(q9u16, q8s16, q10s16);
68  vst1q_s16(block, q0s16);
69  vst1q_s16(block + 8, q8s16);
70  }
71  }
72  if (nCoeffs <= 0)
73  return;
74 
75  d0s16 = vld1_s16(block);
76  d3s16 = vreinterpret_s16_u16(vclt_s16(d0s16, dzs16));
77  d1u16 = vceq_s16(d0s16, dzs16);
78  d2s16 = vmul_s16(d0s16, vget_high_s16(q15s16));
79  d3s16 = vbsl_s16(vreinterpret_u16_s16(d3s16),
80  vget_high_s16(q13s16), vget_high_s16(q14s16));
81  d2s16 = vadd_s16(d2s16, d3s16);
82  d0s16 = vbsl_s16(d1u16, d0s16, d2s16);
83  vst1_s16(block, d0s16);
84 }
85 
87  int n, int qscale)
88 {
89  int nCoeffs = s->inter_scantable.raster_end[s->block_last_index[n]];
90  int qadd = (qscale - 1) | 1;
91 
92  ff_dct_unquantize_h263_neon(qscale, qadd, nCoeffs + 1, block);
93 }
94 
96  int n, int qscale)
97 {
98  int qadd;
99  int nCoeffs, blk0;
100 
101  if (!s->h263_aic) {
102  if (n < 4)
103  block[0] *= s->y_dc_scale;
104  else
105  block[0] *= s->c_dc_scale;
106  qadd = (qscale - 1) | 1;
107  } else {
108  qadd = 0;
109  }
110 
111  if (s->ac_pred) {
112  nCoeffs = 63;
113  } else {
114  nCoeffs = s->inter_scantable.raster_end[s->block_last_index[n]];
115  if (nCoeffs <= 0)
116  return;
117  }
118 
119  blk0 = block[0];
120 
121  ff_dct_unquantize_h263_neon(qscale, qadd, nCoeffs + 1, block);
122 
123  block[0] = blk0;
124 }
125 
126 
128 {
129  int cpu_flags = av_get_cpu_flags();
130 
131  if (have_neon(cpu_flags)) {
132  s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_neon;
133  s->dct_unquantize_h263_inter = dct_unquantize_h263_inter_neon;
134  }
135 }
dct_unquantize_h263_inter_neon
static void dct_unquantize_h263_inter_neon(MpegEncContext *s, int16_t *block, int n, int qscale)
Definition: mpegvideo.c:86
mpegvideo.h
av_get_cpu_flags
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:101
cpu_flags
static atomic_int cpu_flags
Definition: cpu.c:52
av_cold
#define av_cold
Definition: attributes.h:90
s
#define s(width, name)
Definition: cbs_vp9.c:256
cpu.h
cpu.h
have_neon
#define have_neon(flags)
Definition: cpu.h:26
attributes.h
dct_unquantize_h263_intra_neon
static void dct_unquantize_h263_intra_neon(MpegEncContext *s, int16_t *block, int n, int qscale)
Definition: mpegvideo.c:95
blk0
#define blk0(i)
Definition: sha.c:54
block
The exact code depends on how similar the blocks are and how related they are to the block
Definition: filter_design.txt:207
ff_dct_unquantize_h263_neon
static void ff_dct_unquantize_h263_neon(int qscale, int qadd, int nCoeffs, int16_t *block)
Definition: mpegvideo.c:36
cpu.h
ff_mpv_common_init_neon
av_cold void ff_mpv_common_init_neon(MpegEncContext *s)
Definition: mpegvideo.c:127
MpegEncContext
MpegEncContext.
Definition: mpegvideo.h:62