FFmpeg
vc1dsp_init.c
Go to the documentation of this file.
1 /*
2  * VC-1 and WMV3 - DSP functions MMX-optimized
3  * Copyright (c) 2007 Christophe GISQUET <christophe.gisquet@free.fr>
4  *
5  * Permission is hereby granted, free of charge, to any person
6  * obtaining a copy of this software and associated documentation
7  * files (the "Software"), to deal in the Software without
8  * restriction, including without limitation the rights to use,
9  * copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the
11  * Software is furnished to do so, subject to the following
12  * conditions:
13  *
14  * The above copyright notice and this permission notice shall be
15  * included in all copies or substantial portions of the Software.
16  *
17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
18  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
19  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
20  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
21  * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
22  * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
23  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
24  * OTHER DEALINGS IN THE SOFTWARE.
25  */
26 
27 #include "libavutil/attributes.h"
28 #include "libavutil/cpu.h"
29 #include "libavutil/x86/cpu.h"
30 #include "libavutil/x86/asm.h"
31 #include "libavcodec/vc1dsp.h"
32 #include "fpel.h"
33 #include "vc1dsp.h"
34 #include "config.h"
35 
36 #define LOOP_FILTER4(EXT) \
37 void ff_vc1_v_loop_filter4_ ## EXT(uint8_t *src, ptrdiff_t stride, int pq); \
38 void ff_vc1_h_loop_filter4_ ## EXT(uint8_t *src, ptrdiff_t stride, int pq);
39 #define LOOP_FILTER816(EXT) \
40 void ff_vc1_v_loop_filter8_ ## EXT(uint8_t *src, ptrdiff_t stride, int pq); \
41 void ff_vc1_h_loop_filter8_ ## EXT(uint8_t *src, ptrdiff_t stride, int pq); \
42 \
43 static void vc1_v_loop_filter16_ ## EXT(uint8_t *src, ptrdiff_t stride, int pq) \
44 { \
45  ff_vc1_v_loop_filter8_ ## EXT(src, stride, pq); \
46  ff_vc1_v_loop_filter8_ ## EXT(src+8, stride, pq); \
47 } \
48 \
49 static void vc1_h_loop_filter16_ ## EXT(uint8_t *src, ptrdiff_t stride, int pq) \
50 { \
51  ff_vc1_h_loop_filter8_ ## EXT(src, stride, pq); \
52  ff_vc1_h_loop_filter8_ ## EXT(src+8*stride, stride, pq); \
53 }
54 
55 LOOP_FILTER816(sse2)
56 LOOP_FILTER4(ssse3)
57 LOOP_FILTER816(ssse3)
58 
59 void ff_vc1_h_loop_filter8_sse4(uint8_t *src, ptrdiff_t stride, int pq);
60 
61 static void vc1_h_loop_filter16_sse4(uint8_t *src, ptrdiff_t stride, int pq)
62 {
63  ff_vc1_h_loop_filter8_sse4(src, stride, pq);
64  ff_vc1_h_loop_filter8_sse4(src+8*stride, stride, pq);
65 }
66 
67 #define DECLARE_FUNCTION(OP, DEPTH, INSN) \
68  static void OP##vc1_mspel_mc00_##DEPTH##INSN(uint8_t *dst, \
69  const uint8_t *src, ptrdiff_t stride, int rnd) \
70  { \
71  ff_ ## OP ## pixels ## DEPTH ## INSN(dst, src, stride, DEPTH); \
72  }
73 
74 DECLARE_FUNCTION(put_, 8, _sse2)
75 DECLARE_FUNCTION(avg_, 8, _sse2)
76 DECLARE_FUNCTION(put_, 16, _sse2)
77 DECLARE_FUNCTION(avg_, 16, _sse2)
78 
79 void ff_put_vc1_chroma_mc8_nornd_ssse3(uint8_t *dst, const uint8_t *src,
80  ptrdiff_t stride, int h, int x, int y);
81 void ff_avg_vc1_chroma_mc8_nornd_ssse3(uint8_t *dst, const uint8_t *src,
82  ptrdiff_t stride, int h, int x, int y);
83 void ff_vc1_inv_trans_4x4_dc_mmxext(uint8_t *dest, ptrdiff_t linesize,
84  int16_t *block);
85 void ff_vc1_inv_trans_4x8_dc_mmxext(uint8_t *dest, ptrdiff_t linesize,
86  int16_t *block);
87 void ff_vc1_inv_trans_8x4_dc_mmxext(uint8_t *dest, ptrdiff_t linesize,
88  int16_t *block);
89 void ff_vc1_inv_trans_8x8_dc_mmxext(uint8_t *dest, ptrdiff_t linesize,
90  int16_t *block);
91 
92 #define MSPEL_FUNC(OP, X, Y, SIZE, XMM) \
93  void ff_vc1_ ## OP ## _mspel_mc ## X ## Y ## _ ## SIZE ##_ ## XMM \
94  (uint8_t *dst, const uint8_t *src, ptrdiff_t stride, int rnd); \
95  dsp->OP ## _vc1_mspel_pixels_tab[SIZE == 8][X + 4 * Y] = \
96  ff_vc1_ ## OP ## _mspel_mc ## X ## Y## _ ## SIZE ##_ ## XMM
97 #define MSPEL_FUNCS_SIZE(X, Y, SIZE, XMM) \
98  MSPEL_FUNC(put, X, Y, SIZE, XMM); \
99  MSPEL_FUNC(avg, X, Y, SIZE, XMM)
100 #define MSPEL_FUNCS(X, Y, XMM) \
101  MSPEL_FUNCS_SIZE(X, Y, 8, XMM); \
102  MSPEL_FUNCS_SIZE(X, Y, 16, XMM)
103 
104 av_cold void ff_vc1dsp_init_x86(VC1DSPContext *dsp)
105 {
106  int cpu_flags = av_get_cpu_flags();
107 
108 #if HAVE_6REGS && HAVE_INLINE_ASM && HAVE_MMX_EXTERNAL
109  if (HAVE_6REGS && INLINE_MMX(cpu_flags))
110  if (EXTERNAL_MMX(cpu_flags))
111  ff_vc1dsp_init_mmx(dsp);
112 
113  if (HAVE_6REGS && INLINE_MMXEXT(cpu_flags))
114  if (EXTERNAL_MMXEXT(cpu_flags))
115  ff_vc1dsp_init_mmxext(dsp);
116 #endif /* HAVE_6REGS && HAVE_INLINE_ASM && HAVE_MMX_EXTERNAL */
117 
118 #define ASSIGN_LF4(EXT) \
119  dsp->vc1_v_loop_filter4 = ff_vc1_v_loop_filter4_ ## EXT; \
120  dsp->vc1_h_loop_filter4 = ff_vc1_h_loop_filter4_ ## EXT
121 #define ASSIGN_LF816(EXT) \
122  dsp->vc1_v_loop_filter8 = ff_vc1_v_loop_filter8_ ## EXT; \
123  dsp->vc1_h_loop_filter8 = ff_vc1_h_loop_filter8_ ## EXT; \
124  dsp->vc1_v_loop_filter16 = vc1_v_loop_filter16_ ## EXT; \
125  dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_ ## EXT
126 
127  if (EXTERNAL_MMXEXT(cpu_flags)) {
128  dsp->vc1_inv_trans_8x8_dc = ff_vc1_inv_trans_8x8_dc_mmxext;
129  dsp->vc1_inv_trans_4x8_dc = ff_vc1_inv_trans_4x8_dc_mmxext;
130  dsp->vc1_inv_trans_8x4_dc = ff_vc1_inv_trans_8x4_dc_mmxext;
131  dsp->vc1_inv_trans_4x4_dc = ff_vc1_inv_trans_4x4_dc_mmxext;
132  }
133  if (EXTERNAL_SSE2(cpu_flags)) {
134  ASSIGN_LF816(sse2);
135 
136  dsp->put_vc1_mspel_pixels_tab[0][0] = put_vc1_mspel_mc00_16_sse2;
137  dsp->put_vc1_mspel_pixels_tab[1][0] = put_vc1_mspel_mc00_8_sse2;
138  dsp->avg_vc1_mspel_pixels_tab[0][0] = avg_vc1_mspel_mc00_16_sse2;
139  dsp->avg_vc1_mspel_pixels_tab[1][0] = avg_vc1_mspel_mc00_8_sse2;
140  }
141  if (EXTERNAL_SSSE3(cpu_flags)) {
142  ASSIGN_LF4(ssse3);
143  ASSIGN_LF816(ssse3);
144  dsp->put_no_rnd_vc1_chroma_pixels_tab[0] = ff_put_vc1_chroma_mc8_nornd_ssse3;
145  dsp->avg_no_rnd_vc1_chroma_pixels_tab[0] = ff_avg_vc1_chroma_mc8_nornd_ssse3;
146 
147  MSPEL_FUNCS(0, 1, ssse3);
148  MSPEL_FUNCS(0, 2, ssse3);
149  MSPEL_FUNCS(0, 3, ssse3);
150  MSPEL_FUNCS(1, 0, ssse3);
151  MSPEL_FUNCS(2, 0, ssse3);
152  MSPEL_FUNCS(3, 0, ssse3);
153  }
154  if (EXTERNAL_SSE4(cpu_flags)) {
155  dsp->vc1_h_loop_filter8 = ff_vc1_h_loop_filter8_sse4;
156  dsp->vc1_h_loop_filter16 = vc1_h_loop_filter16_sse4;
157  }
158 }
ff_vc1_inv_trans_8x8_dc_mmxext
void ff_vc1_inv_trans_8x8_dc_mmxext(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
INLINE_MMXEXT
#define INLINE_MMXEXT(flags)
Definition: cpu.h:81
LOOP_FILTER4
#define LOOP_FILTER4(EXT)
Definition: vc1dsp_init.c:36
vc1dsp.h
VC1DSPContext::avg_vc1_mspel_pixels_tab
vc1op_pixels_func avg_vc1_mspel_pixels_tab[2][16]
Definition: vc1dsp.h:62
av_cold
#define av_cold
Definition: attributes.h:119
VC1DSPContext::avg_no_rnd_vc1_chroma_pixels_tab
h264_chroma_mc_func avg_no_rnd_vc1_chroma_pixels_tab[3]
Definition: vc1dsp.h:66
ASSIGN_LF816
#define ASSIGN_LF816(EXT)
ASSIGN_LF4
#define ASSIGN_LF4(EXT)
VC1DSPContext::vc1_inv_trans_8x8_dc
void(* vc1_inv_trans_8x8_dc)(uint8_t *dest, ptrdiff_t stride, int16_t *block)
Definition: vc1dsp.h:43
VC1DSPContext::put_no_rnd_vc1_chroma_pixels_tab
h264_chroma_mc_func put_no_rnd_vc1_chroma_pixels_tab[3]
Definition: vc1dsp.h:65
VC1DSPContext::vc1_inv_trans_4x4_dc
void(* vc1_inv_trans_4x4_dc)(uint8_t *dest, ptrdiff_t stride, int16_t *block)
Definition: vc1dsp.h:46
LOOP_FILTER816
#define LOOP_FILTER816(EXT)
Definition: vc1dsp_init.c:39
cpu.h
EXTERNAL_SSSE3
#define EXTERNAL_SSSE3(flags)
Definition: cpu.h:59
VC1DSPContext::vc1_inv_trans_8x4_dc
void(* vc1_inv_trans_8x4_dc)(uint8_t *dest, ptrdiff_t stride, int16_t *block)
Definition: vc1dsp.h:44
ff_put_vc1_chroma_mc8_nornd_ssse3
void ff_put_vc1_chroma_mc8_nornd_ssse3(uint8_t *dst, const uint8_t *src, ptrdiff_t stride, int h, int x, int y)
VC1DSPContext::vc1_h_loop_filter16
void(* vc1_h_loop_filter16)(uint8_t *src, ptrdiff_t stride, int pq)
Definition: vc1dsp.h:56
MSPEL_FUNCS
#define MSPEL_FUNCS(X, Y, XMM)
Definition: vc1dsp_init.c:100
av_get_cpu_flags
int av_get_cpu_flags(void)
Return the flags which specify extensions supported by the CPU.
Definition: cpu.c:109
EXTERNAL_MMX
#define EXTERNAL_MMX(flags)
Definition: cpu.h:50
ff_vc1_inv_trans_4x8_dc_mmxext
void ff_vc1_inv_trans_4x8_dc_mmxext(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
asm.h
vc1_h_loop_filter16_sse4
static void vc1_h_loop_filter16_sse4(uint8_t *src, ptrdiff_t stride, int pq)
Definition: vc1dsp_init.c:61
attributes.h
DECLARE_FUNCTION
#define DECLARE_FUNCTION(OP, DEPTH, INSN)
Definition: vc1dsp_init.c:67
VC1DSPContext::vc1_inv_trans_4x8_dc
void(* vc1_inv_trans_4x8_dc)(uint8_t *dest, ptrdiff_t stride, int16_t *block)
Definition: vc1dsp.h:45
dst
uint8_t ptrdiff_t const uint8_t ptrdiff_t int intptr_t intptr_t int int16_t * dst
Definition: dsp.h:87
cpu_flags
CheckasmCpu cpu_flags
Definition: checkasm.c:84
EXTERNAL_SSE2
#define EXTERNAL_SSE2(flags)
Definition: cpu.h:53
EXTERNAL_SSE4
#define EXTERNAL_SSE4(flags)
Definition: cpu.h:62
ff_vc1_inv_trans_8x4_dc_mmxext
void ff_vc1_inv_trans_8x4_dc_mmxext(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
INLINE_MMX
#define INLINE_MMX(flags)
Definition: cpu.h:80
VC1DSPContext
Definition: vc1dsp.h:37
VC1DSPContext::vc1_h_loop_filter8
void(* vc1_h_loop_filter8)(uint8_t *src, ptrdiff_t stride, int pq)
Definition: vc1dsp.h:54
VC1DSPContext::put_vc1_mspel_pixels_tab
vc1op_pixels_func put_vc1_mspel_pixels_tab[2][16]
Definition: vc1dsp.h:61
ff_avg_vc1_chroma_mc8_nornd_ssse3
void ff_avg_vc1_chroma_mc8_nornd_ssse3(uint8_t *dst, const uint8_t *src, ptrdiff_t stride, int h, int x, int y)
HAVE_6REGS
#define HAVE_6REGS
Definition: asm.h:75
ff_vc1_inv_trans_4x4_dc_mmxext
void ff_vc1_inv_trans_4x4_dc_mmxext(uint8_t *dest, ptrdiff_t linesize, int16_t *block)
ff_vc1_h_loop_filter8_sse4
void ff_vc1_h_loop_filter8_sse4(uint8_t *src, ptrdiff_t stride, int pq)
fpel.h
EXTERNAL_MMXEXT
#define EXTERNAL_MMXEXT(flags)
Definition: cpu.h:51
cpu.h
ff_vc1dsp_init_x86
av_cold void ff_vc1dsp_init_x86(VC1DSPContext *dsp)
Definition: vc1dsp_init.c:104
block
The exact code depends on how similar the blocks are and how related they are to the block
Definition: filter_design.txt:207
h
h
Definition: vp9dsp_template.c:2070
stride
#define stride
Definition: h264pred_template.c:536
src
#define src
Definition: vp8dsp.c:248
Generated on Thu Jul 16 2026 19:23:01 for FFmpeg by   doxygen 1.8.17