FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
vp3dsp_altivec.c
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2009 David Conrad
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 <string.h>
22 
23 #include "config.h"
24 #include "libavutil/attributes.h"
25 #include "libavutil/cpu.h"
26 #include "libavutil/ppc/cpu.h"
29 #include "libavcodec/vp3dsp.h"
30 
31 #if HAVE_ALTIVEC
32 
33 static const vec_s16 constants =
34  {0, 64277, 60547, 54491, 46341, 36410, 25080, 12785};
35 #if HAVE_BIGENDIAN
36 static const vec_u8 interleave_high =
37  {0, 1, 16, 17, 4, 5, 20, 21, 8, 9, 24, 25, 12, 13, 28, 29};
38 #else
39 static const vec_u8 interleave_high =
40  {2, 3, 18, 19, 6, 7, 22, 23, 10, 11, 26, 27, 14, 15, 30, 31};
41 #endif
42 
43 #define IDCT_START \
44  vec_s16 A, B, C, D, Ad, Bd, Cd, Dd, E, F, G, H;\
45  vec_s16 Ed, Gd, Add, Bdd, Fd, Hd;\
46  vec_s16 eight = vec_splat_s16(8);\
47  vec_u16 four = vec_splat_u16(4);\
48 \
49  vec_s16 C1 = vec_splat(constants, 1);\
50  vec_s16 C2 = vec_splat(constants, 2);\
51  vec_s16 C3 = vec_splat(constants, 3);\
52  vec_s16 C4 = vec_splat(constants, 4);\
53  vec_s16 C5 = vec_splat(constants, 5);\
54  vec_s16 C6 = vec_splat(constants, 6);\
55  vec_s16 C7 = vec_splat(constants, 7);\
56 \
57  vec_s16 b0 = vec_ld(0x00, block);\
58  vec_s16 b1 = vec_ld(0x10, block);\
59  vec_s16 b2 = vec_ld(0x20, block);\
60  vec_s16 b3 = vec_ld(0x30, block);\
61  vec_s16 b4 = vec_ld(0x40, block);\
62  vec_s16 b5 = vec_ld(0x50, block);\
63  vec_s16 b6 = vec_ld(0x60, block);\
64  vec_s16 b7 = vec_ld(0x70, block);
65 
66 // these functions do (a*C)>>16
67 // things are tricky because a is signed, but C unsigned.
68 // M15 is used if C fits in 15 bit unsigned (C6,C7)
69 // M16 is used if C requires 16 bits unsigned
70 static inline vec_s16 M15(vec_s16 a, vec_s16 C)
71 {
72  return (vec_s16)vec_perm(vec_mule(a,C), vec_mulo(a,C), interleave_high);
73 }
74 static inline vec_s16 M16(vec_s16 a, vec_s16 C)
75 {
76  return vec_add(a, M15(a, C));
77 }
78 
79 #define IDCT_1D(ADD, SHIFT)\
80  A = vec_add(M16(b1, C1), M15(b7, C7));\
81  B = vec_sub(M15(b1, C7), M16(b7, C1));\
82  C = vec_add(M16(b3, C3), M16(b5, C5));\
83  D = vec_sub(M16(b5, C3), M16(b3, C5));\
84 \
85  Ad = M16(vec_sub(A, C), C4);\
86  Bd = M16(vec_sub(B, D), C4);\
87 \
88  Cd = vec_add(A, C);\
89  Dd = vec_add(B, D);\
90 \
91  E = ADD(M16(vec_add(b0, b4), C4));\
92  F = ADD(M16(vec_sub(b0, b4), C4));\
93 \
94  G = vec_add(M16(b2, C2), M15(b6, C6));\
95  H = vec_sub(M15(b2, C6), M16(b6, C2));\
96 \
97  Ed = vec_sub(E, G);\
98  Gd = vec_add(E, G);\
99 \
100  Add = vec_add(F, Ad);\
101  Bdd = vec_sub(Bd, H);\
102 \
103  Fd = vec_sub(F, Ad);\
104  Hd = vec_add(Bd, H);\
105 \
106  b0 = SHIFT(vec_add(Gd, Cd));\
107  b7 = SHIFT(vec_sub(Gd, Cd));\
108 \
109  b1 = SHIFT(vec_add(Add, Hd));\
110  b2 = SHIFT(vec_sub(Add, Hd));\
111 \
112  b3 = SHIFT(vec_add(Ed, Dd));\
113  b4 = SHIFT(vec_sub(Ed, Dd));\
114 \
115  b5 = SHIFT(vec_add(Fd, Bdd));\
116  b6 = SHIFT(vec_sub(Fd, Bdd));
117 
118 #define NOP(a) a
119 #define ADD8(a) vec_add(a, eight)
120 #define SHIFT4(a) vec_sra(a, four)
121 
122 static void vp3_idct_put_altivec(uint8_t *dst, int stride, int16_t block[64])
123 {
124  vec_u8 t;
125  IDCT_START
126 
127  // pixels are signed; so add 128*16 in addition to the normal 8
128  vec_s16 v2048 = vec_sl(vec_splat_s16(1), vec_splat_u16(11));
129  eight = vec_add(eight, v2048);
130 
131  IDCT_1D(NOP, NOP)
132  TRANSPOSE8(b0, b1, b2, b3, b4, b5, b6, b7);
133  IDCT_1D(ADD8, SHIFT4)
134 
135 #define PUT(a)\
136  t = vec_packsu(a, a);\
137  vec_ste((vec_u32)t, 0, (unsigned int *)dst);\
138  vec_ste((vec_u32)t, 4, (unsigned int *)dst);
139 
140  PUT(b0) dst += stride;
141  PUT(b1) dst += stride;
142  PUT(b2) dst += stride;
143  PUT(b3) dst += stride;
144  PUT(b4) dst += stride;
145  PUT(b5) dst += stride;
146  PUT(b6) dst += stride;
147  PUT(b7)
148  memset(block, 0, sizeof(*block) * 64);
149 }
150 
151 static void vp3_idct_add_altivec(uint8_t *dst, int stride, int16_t block[64])
152 {
153  LOAD_ZERO;
154  vec_u8 t, vdst;
155  vec_s16 vdst_16;
156  vec_u8 vdst_mask = vec_mergeh(vec_splat_u8(-1), vec_lvsl(0, dst));
157 
158  IDCT_START
159 
160  IDCT_1D(NOP, NOP)
161  TRANSPOSE8(b0, b1, b2, b3, b4, b5, b6, b7);
162  IDCT_1D(ADD8, SHIFT4)
163 
164 #if HAVE_BIGENDIAN
165 #define GET_VDST16\
166  vdst = vec_ld(0, dst);\
167  vdst_16 = (vec_s16)vec_perm(vdst, zero_u8v, vdst_mask);
168 #else
169 #define GET_VDST16\
170  vdst = vec_vsx_ld(0,dst);\
171  vdst_16 = (vec_s16)vec_mergeh(vdst, zero_u8v);
172 #endif
173 
174 #define ADD(a)\
175  GET_VDST16;\
176  vdst_16 = vec_adds(a, vdst_16);\
177  t = vec_packsu(vdst_16, vdst_16);\
178  vec_ste((vec_u32)t, 0, (unsigned int *)dst);\
179  vec_ste((vec_u32)t, 4, (unsigned int *)dst);
180 
181  ADD(b0) dst += stride;
182  ADD(b1) dst += stride;
183  ADD(b2) dst += stride;
184  ADD(b3) dst += stride;
185  ADD(b4) dst += stride;
186  ADD(b5) dst += stride;
187  ADD(b6) dst += stride;
188  ADD(b7)
189  memset(block, 0, sizeof(*block) * 64);
190 }
191 
192 #endif /* HAVE_ALTIVEC */
193 
195 {
196 #if HAVE_ALTIVEC
198  return;
199 
200  c->idct_put = vp3_idct_put_altivec;
201  c->idct_add = vp3_idct_add_altivec;
202 #endif
203 }
void(* idct_put)(uint8_t *dest, int line_size, int16_t *block)
Definition: vp3dsp.h:41
#define C
Macro definitions for various function/variable attributes.
static const struct @195 constants[]
uint8_t
#define av_cold
Definition: attributes.h:82
#define vec_s16
Definition: types_altivec.h:30
#define ADD(a, b)
#define PPC_ALTIVEC(flags)
Definition: cpu.h:26
#define LOAD_ZERO
Definition: types_altivec.h:38
void(* idct_add)(uint8_t *dest, int line_size, int16_t *block)
Definition: vp3dsp.h:42
#define vec_u8
Definition: types_altivec.h:27
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:76
Contains misc utility macros and inline functions.
static int flags
Definition: cpu.c:47
GLint GLenum GLboolean GLsizei stride
Definition: opengl_enc.c:105
static double c[64]
av_cold void ff_vp3dsp_init_ppc(VP3DSPContext *c, int flags)
static int16_t block[64]
Definition: dct-test.c:112