1 /*
2 * H.26L/H.264/AVC/JVT/14496-10/... encoder/decoder
3 * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
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 /**
23 * @file
24 * H.264 / AVC / MPEG-4 part10 codec.
25 * non-SIMD x86-specific optimizations for H.264
26 * @author Michael Niedermayer <michaelni@gmx.at>
27 */
28
29 #include <stddef.h>
30
31 #include "libavcodec/cabac.h"
32 #include "cabac.h"
33
34 #if HAVE_INLINE_ASM
35
36 #if ARCH_X86_64
37 #define REG64 "r"
38 #else
39 #define REG64 "m"
40 #endif
41
42 //FIXME use some macros to avoid duplicating get_cabac (cannot be done yet
43 //as that would make optimization work hard)
44 #if HAVE_7REGS && !BROKEN_COMPILER
45 #define decode_significance decode_significance_x86
decode_significance_x86(CABACContext * c,int max_coeff,uint8_t * significant_coeff_ctx_base,int * index,x86_reg last_off)46 static int decode_significance_x86(CABACContext *c, int max_coeff,
47 uint8_t *significant_coeff_ctx_base,
48 int *index, x86_reg last_off){
49 void *end= significant_coeff_ctx_base + max_coeff - 1;
50 int minusstart= -(intptr_t)significant_coeff_ctx_base;
51 int minusindex= 4-(intptr_t)index;
52 int bit;
53 x86_reg coeff_count;
54
55 #ifdef BROKEN_RELOCATIONS
56 void *tables;
57
58 __asm__ volatile(
59 "lea "MANGLE(ff_h264_cabac_tables)", %0 \n\t"
60 : "=&r"(tables)
61 : NAMED_CONSTRAINTS_ARRAY(ff_h264_cabac_tables)
62 );
63 #endif
64
65 __asm__ volatile(
66 "3: \n\t"
67
68 BRANCHLESS_GET_CABAC("%4", "%q4", "(%1)", "%3", "%w3",
69 "%5", "%q5", "%k0", "%b0",
70 "%c11(%6)", "%c12(%6)",
71 AV_STRINGIFY(H264_NORM_SHIFT_OFFSET),
72 AV_STRINGIFY(H264_LPS_RANGE_OFFSET),
73 AV_STRINGIFY(H264_MLPS_STATE_OFFSET),
74 "%13")
75
76 "test $1, %4 \n\t"
77 " jz 4f \n\t"
78 "add %10, %1 \n\t"
79
80 BRANCHLESS_GET_CABAC("%4", "%q4", "(%1)", "%3", "%w3",
81 "%5", "%q5", "%k0", "%b0",
82 "%c11(%6)", "%c12(%6)",
83 AV_STRINGIFY(H264_NORM_SHIFT_OFFSET),
84 AV_STRINGIFY(H264_LPS_RANGE_OFFSET),
85 AV_STRINGIFY(H264_MLPS_STATE_OFFSET),
86 "%13")
87
88 "sub %10, %1 \n\t"
89 "mov %2, %0 \n\t"
90 "movl %7, %%ecx \n\t"
91 "add %1, %%"FF_REG_c" \n\t"
92 "movl %%ecx, (%0) \n\t"
93
94 "test $1, %4 \n\t"
95 " jnz 5f \n\t"
96
97 "add"FF_OPSIZE" $4, %2 \n\t"
98
99 "4: \n\t"
100 "add $1, %1 \n\t"
101 "cmp %8, %1 \n\t"
102 " jb 3b \n\t"
103 "mov %2, %0 \n\t"
104 "movl %7, %%ecx \n\t"
105 "add %1, %%"FF_REG_c" \n\t"
106 "movl %%ecx, (%0) \n\t"
107 "5: \n\t"
108 "add %9, %k0 \n\t"
109 "shr $2, %k0 \n\t"
110 : "=&q"(coeff_count), "+r"(significant_coeff_ctx_base), "+m"(index),
111 "+&r"(c->low), "=&r"(bit), "+&r"(c->range)
112 : "r"(c), "m"(minusstart), "m"(end), "m"(minusindex), "m"(last_off),
113 "i"(offsetof(CABACContext, bytestream)),
114 "i"(offsetof(CABACContext, bytestream_end))
115 TABLES_ARG
116 : "%"FF_REG_c, "memory"
117 );
118 return coeff_count;
119 }
120
121 #define decode_significance_8x8 decode_significance_8x8_x86
decode_significance_8x8_x86(CABACContext * c,uint8_t * significant_coeff_ctx_base,int * index,uint8_t * last_coeff_ctx_base,const uint8_t * sig_off)122 static int decode_significance_8x8_x86(CABACContext *c,
123 uint8_t *significant_coeff_ctx_base,
124 int *index, uint8_t *last_coeff_ctx_base, const uint8_t *sig_off){
125 int minusindex= 4-(intptr_t)index;
126 int bit;
127 x86_reg coeff_count;
128 x86_reg last=0;
129 x86_reg state;
130
131 #ifdef BROKEN_RELOCATIONS
132 void *tables;
133
134 __asm__ volatile(
135 "lea "MANGLE(ff_h264_cabac_tables)", %0 \n\t"
136 : "=&r"(tables)
137 : NAMED_CONSTRAINTS_ARRAY(ff_h264_cabac_tables)
138 );
139 #endif
140
141 __asm__ volatile(
142 "mov %1, %6 \n\t"
143 "3: \n\t"
144
145 "mov %10, %0 \n\t"
146 "movzb (%0, %6), %6 \n\t"
147 "add %9, %6 \n\t"
148
149 BRANCHLESS_GET_CABAC("%4", "%q4", "(%6)", "%3", "%w3",
150 "%5", "%q5", "%k0", "%b0",
151 "%c12(%7)", "%c13(%7)",
152 AV_STRINGIFY(H264_NORM_SHIFT_OFFSET),
153 AV_STRINGIFY(H264_LPS_RANGE_OFFSET),
154 AV_STRINGIFY(H264_MLPS_STATE_OFFSET),
155 "%15")
156
157 "mov %1, %6 \n\t"
158 "test $1, %4 \n\t"
159 " jz 4f \n\t"
160
161 #ifdef BROKEN_RELOCATIONS
162 "movzb %c14(%15, %q6), %6\n\t"
163 #else
164 "movzb "MANGLE(ff_h264_cabac_tables)"+%c14(%6), %6\n\t"
165 #endif
166 "add %11, %6 \n\t"
167
168 BRANCHLESS_GET_CABAC("%4", "%q4", "(%6)", "%3", "%w3",
169 "%5", "%q5", "%k0", "%b0",
170 "%c12(%7)", "%c13(%7)",
171 AV_STRINGIFY(H264_NORM_SHIFT_OFFSET),
172 AV_STRINGIFY(H264_LPS_RANGE_OFFSET),
173 AV_STRINGIFY(H264_MLPS_STATE_OFFSET),
174 "%15")
175
176 "mov %2, %0 \n\t"
177 "mov %1, %6 \n\t"
178 "mov %k6, (%0) \n\t"
179
180 "test $1, %4 \n\t"
181 " jnz 5f \n\t"
182
183 "add"FF_OPSIZE" $4, %2 \n\t"
184
185 "4: \n\t"
186 "add $1, %6 \n\t"
187 "mov %6, %1 \n\t"
188 "cmp $63, %6 \n\t"
189 " jb 3b \n\t"
190 "mov %2, %0 \n\t"
191 "mov %k6, (%0) \n\t"
192 "5: \n\t"
193 "addl %8, %k0 \n\t"
194 "shr $2, %k0 \n\t"
195 : "=&q"(coeff_count), "+"REG64(last), "+"REG64(index), "+&r"(c->low),
196 "=&r"(bit), "+&r"(c->range), "=&r"(state)
197 : "r"(c), "m"(minusindex), "m"(significant_coeff_ctx_base),
198 REG64(sig_off), REG64(last_coeff_ctx_base),
199 "i"(offsetof(CABACContext, bytestream)),
200 "i"(offsetof(CABACContext, bytestream_end)),
201 "i"(H264_LAST_COEFF_FLAG_OFFSET_8x8_OFFSET) TABLES_ARG
202 : "%"FF_REG_c, "memory"
203 );
204 return coeff_count;
205 }
206 #endif /* HAVE_7REGS && BROKEN_COMPILER */
207
208 #endif /* HAVE_INLINE_ASM */
209