FFmpeg
hevcdsp.c
Go to the documentation of this file.
1 /*
2  * SIMD-optimized IDCT functions for HEVC decoding
3  * Copyright (c) Alexandra Hajkova
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 "config.h"
23 
24 #include "libavutil/attributes.h"
25 #include "libavutil/cpu.h"
26 #include "libavutil/ppc/cpu.h"
28 
29 #include "libavcodec/hevcdsp.h"
30 
31 #if HAVE_ALTIVEC
32 static const vec_s16 trans4[4] = {
33  { 64, 64, 64, 64, 64, 64, 64, 64 },
34  { 83, 36, 83, 36, 83, 36, 83, 36 },
35  { 64, -64, 64, -64, 64, -64, 64, -64 },
36  { 36, -83, 36, -83, 36, -83, 36, -83 },
37 };
38 
39 static const vec_u8 mask[2] = {
40  { 0x00, 0x01, 0x08, 0x09, 0x10, 0x11, 0x18, 0x19, 0x02, 0x03, 0x0A, 0x0B, 0x12, 0x13, 0x1A, 0x1B },
41  { 0x04, 0x05, 0x0C, 0x0D, 0x14, 0x15, 0x1C, 0x1D, 0x06, 0x07, 0x0E, 0x0F, 0x16, 0x17, 0x1E, 0x1F },
42 };
43 
44 static av_always_inline void transform4x4(vec_s16 src_01, vec_s16 src_23,
45  vec_s32 res[4], const int shift,
46  int16_t *coeffs)
47 {
48  vec_s16 src_02, src_13;
49  vec_s32 zero = vec_splat_s32(0);
50  vec_s32 e0, o0, e1, o1;
51  vec_s32 add;
52 
53  src_13 = vec_mergel(src_01, src_23);
54  src_02 = vec_mergeh(src_01, src_23);
55 
56  e0 = vec_msums(src_02, trans4[0], zero);
57  o0 = vec_msums(src_13, trans4[1], zero);
58  e1 = vec_msums(src_02, trans4[2], zero);
59  o1 = vec_msums(src_13, trans4[3], zero);
60 
61  switch(shift) {
62  case 7: add = vec_sl(vec_splat_s32(1), vec_splat_u32( 7 - 1)); break;
63  case 10: add = vec_sl(vec_splat_s32(1), vec_splat_u32(10 - 1)); break;
64  case 12: add = vec_sl(vec_splat_s32(1), vec_splat_u32(12 - 1)); break;
65  default: abort();
66  }
67 
68  e0 = vec_add(e0, add);
69  e1 = vec_add(e1, add);
70 
71  res[0] = vec_add(e0, o0);
72  res[1] = vec_add(e1, o1);
73  res[2] = vec_sub(e1, o1);
74  res[3] = vec_sub(e0, o0);
75 }
76 
77 static av_always_inline void scale(vec_s32 res[4], vec_s16 res_packed[2],
78  const int shift)
79 {
80  int i;
81  vec_u32 v_shift;
82 
83  switch(shift) {
84  case 7: v_shift = vec_splat_u32(7) ; break;
85  case 10: v_shift = vec_splat_u32(10); break;
86  case 12: v_shift = vec_splat_u32(12); break;
87  default: abort();
88  }
89 
90  for (i = 0; i < 4; i++)
91  res[i] = vec_sra(res[i], v_shift);
92 
93  // clip16
94  res_packed[0] = vec_packs(res[0], res[1]);
95  res_packed[1] = vec_packs(res[2], res[3]);
96 }
97 
98 #define FUNCDECL(a, depth) a ## _ ## depth ## _altivec
99 #define FUNC(a, b) FUNCDECL(a, b)
100 
101 #define BIT_DEPTH 8
102 #include "hevcdsp_template.c"
103 #undef BIT_DEPTH
104 
105 #define BIT_DEPTH 10
106 #include "hevcdsp_template.c"
107 #undef BIT_DEPTH
108 #endif /* HAVE_ALTIVEC */
109 
111 {
112 #if HAVE_ALTIVEC
114  return;
115 
116  if (bit_depth == 8)
117  c->idct[0] = ff_hevc_idct_4x4_8_altivec;
118  if (bit_depth == 10)
119  c->idct[0] = ff_hevc_idct_4x4_10_altivec;
120 #endif /* HAVE_ALTIVEC */
121 }
bit_depth
static void bit_depth(AudioStatsContext *s, uint64_t mask, uint64_t imask, AVRational *depth)
Definition: af_astats.c:226
av_get_cpu_flags
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:98
scale
static av_always_inline float scale(float x, float s)
Definition: vf_v360.c:1388
vec_s32
#define vec_s32
Definition: util_altivec.h:39
vec_s16
#define vec_s16
Definition: util_altivec.h:37
av_cold
#define av_cold
Definition: attributes.h:90
mask
static const uint16_t mask[17]
Definition: lzw.c:38
hevcdsp.h
c
Undefined Behavior In the C some operations are like signed integer dereferencing freed accessing outside allocated Undefined Behavior must not occur in a C it is not safe even if the output of undefined operations is unused The unsafety may seem nit picking but Optimizing compilers have in fact optimized code on the assumption that no undefined Behavior occurs Optimizing code based on wrong assumptions can and has in some cases lead to effects beyond the output of computations The signed integer overflow problem in speed critical code Code which is highly optimized and works with signed integers sometimes has the problem that often the output of the computation does not c
Definition: undefined.txt:32
PPC_ALTIVEC
#define PPC_ALTIVEC(flags)
Definition: cpu.h:25
vec_u32
#define vec_u32
Definition: util_altivec.h:38
cpu.h
vec_u8
#define vec_u8
Definition: util_altivec.h:34
HEVCDSPContext
Definition: hevcdsp.h:47
attributes.h
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:271
av_always_inline
#define av_always_inline
Definition: attributes.h:49
shift
static int shift(int a, int b)
Definition: sonic.c:83
zero
#define zero
Definition: regdef.h:64
add
static float add(float src0, float src1)
Definition: dnn_backend_native_layer_mathbinary.c:35
util_altivec.h
ff_hevc_dsp_init_ppc
av_cold void ff_hevc_dsp_init_ppc(HEVCDSPContext *c, const int bit_depth)
Definition: hevcdsp.c:110
cpu.h