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 / MPEG4 part10 codec.
25 * non-MMX i386-specific optimizations for H.264
26 * @author Michael Niedermayer <michaelni@gmx.at>
27 */
28
29 #ifndef AVCODEC_X86_H264_I386_H
30 #define AVCODEC_X86_H264_I386_H
31
32 #include <stddef.h>
33
34 #include "libavcodec/cabac.h"
35 #include "cabac.h"
36
37 #if HAVE_INLINE_ASM
38
39 //FIXME use some macros to avoid duplicating get_cabac (cannot be done yet
40 //as that would make optimization work hard)
41 #if HAVE_7REGS && !BROKEN_COMPILER
42 #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)43 static int decode_significance_x86(CABACContext *c, int max_coeff,
44 uint8_t *significant_coeff_ctx_base,
45 int *index, x86_reg last_off){
46 void *end= significant_coeff_ctx_base + max_coeff - 1;
47 int minusstart= -(intptr_t)significant_coeff_ctx_base;
48 int minusindex= 4-(intptr_t)index;
49 int bit;
50 x86_reg coeff_count;
51
52 #ifdef BROKEN_RELOCATIONS
53 void *tables;
54
55 __asm__ volatile(
56 "lea "MANGLE(ff_h264_cabac_tables)", %0 \n\t"
57 : "=&r"(tables)
58 : NAMED_CONSTRAINTS_ARRAY(ff_h264_cabac_tables)
59 );
60 #endif
61
62 __asm__ volatile(
63 "3: \n\t"
64
65 BRANCHLESS_GET_CABAC("%4", "%q4", "(%1)", "%3", "%w3",
66 "%5", "%q5", "%k0", "%b0",
67 "%c11(%6)", "%c12(%6)",
68 AV_STRINGIFY(H264_NORM_SHIFT_OFFSET),
69 AV_STRINGIFY(H264_LPS_RANGE_OFFSET),
70 AV_STRINGIFY(H264_MLPS_STATE_OFFSET),
71 "%13")
72
73 "test $1, %4 \n\t"
74 " jz 4f \n\t"
75 "add %10, %1 \n\t"
76
77 BRANCHLESS_GET_CABAC("%4", "%q4", "(%1)", "%3", "%w3",
78 "%5", "%q5", "%k0", "%b0",
79 "%c11(%6)", "%c12(%6)",
80 AV_STRINGIFY(H264_NORM_SHIFT_OFFSET),
81 AV_STRINGIFY(H264_LPS_RANGE_OFFSET),
82 AV_STRINGIFY(H264_MLPS_STATE_OFFSET),
83 "%13")
84
85 "sub %10, %1 \n\t"
86 "mov %2, %0 \n\t"
87 "movl %7, %%ecx \n\t"
88 "add %1, %%"REG_c" \n\t"
89 "movl %%ecx, (%0) \n\t"
90
91 "test $1, %4 \n\t"
92 " jnz 5f \n\t"
93
94 "add"OPSIZE" $4, %2 \n\t"
95
96 "4: \n\t"
97 "add $1, %1 \n\t"
98 "cmp %8, %1 \n\t"
99 " jb 3b \n\t"
100 "mov %2, %0 \n\t"
101 "movl %7, %%ecx \n\t"
102 "add %1, %%"REG_c" \n\t"
103 "movl %%ecx, (%0) \n\t"
104 "5: \n\t"
105 "add %9, %k0 \n\t"
106 "shr $2, %k0 \n\t"
107 : "=&q"(coeff_count), "+r"(significant_coeff_ctx_base), "+m"(index),
108 "+&r"(c->low), "=&r"(bit), "+&r"(c->range)
109 : "r"(c), "m"(minusstart), "m"(end), "m"(minusindex), "m"(last_off),
110 "i"(offsetof(CABACContext, bytestream)),
111 "i"(offsetof(CABACContext, bytestream_end))
112 TABLES_ARG
113 : "%"REG_c, "memory"
114 );
115 return coeff_count;
116 }
117
118 #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)119 static int decode_significance_8x8_x86(CABACContext *c,
120 uint8_t *significant_coeff_ctx_base,
121 int *index, uint8_t *last_coeff_ctx_base, const uint8_t *sig_off){
122 int minusindex= 4-(intptr_t)index;
123 int bit;
124 x86_reg coeff_count;
125 x86_reg last=0;
126 x86_reg state;
127
128 #ifdef BROKEN_RELOCATIONS
129 void *tables;
130
131 __asm__ volatile(
132 "lea "MANGLE(ff_h264_cabac_tables)", %0 \n\t"
133 : "=&r"(tables)
134 : NAMED_CONSTRAINTS_ARRAY(ff_h264_cabac_tables)
135 );
136 #endif
137
138 __asm__ volatile(
139 "mov %1, %6 \n\t"
140 "3: \n\t"
141
142 "mov %10, %0 \n\t"
143 "movzbl (%0, %6), %k6 \n\t"
144 "add %9, %6 \n\t"
145
146 BRANCHLESS_GET_CABAC("%4", "%q4", "(%6)", "%3", "%w3",
147 "%5", "%q5", "%k0", "%b0",
148 "%c12(%7)", "%c13(%7)",
149 AV_STRINGIFY(H264_NORM_SHIFT_OFFSET),
150 AV_STRINGIFY(H264_LPS_RANGE_OFFSET),
151 AV_STRINGIFY(H264_MLPS_STATE_OFFSET),
152 "%15")
153
154 "mov %1, %k6 \n\t"
155 "test $1, %4 \n\t"
156 " jz 4f \n\t"
157
158 #ifdef BROKEN_RELOCATIONS
159 "movzbl %c14(%15, %q6), %k6\n\t"
160 #else
161 "movzbl "MANGLE(ff_h264_cabac_tables)"+%c14(%k6), %k6\n\t"
162 #endif
163 "add %11, %6 \n\t"
164
165 BRANCHLESS_GET_CABAC("%4", "%q4", "(%6)", "%3", "%w3",
166 "%5", "%q5", "%k0", "%b0",
167 "%c12(%7)", "%c13(%7)",
168 AV_STRINGIFY(H264_NORM_SHIFT_OFFSET),
169 AV_STRINGIFY(H264_LPS_RANGE_OFFSET),
170 AV_STRINGIFY(H264_MLPS_STATE_OFFSET),
171 "%15")
172
173 "mov %2, %0 \n\t"
174 "mov %1, %k6 \n\t"
175 "movl %k6, (%0) \n\t"
176
177 "test $1, %4 \n\t"
178 " jnz 5f \n\t"
179
180 "add"OPSIZE" $4, %2 \n\t"
181
182 "4: \n\t"
183 "addl $1, %k6 \n\t"
184 "mov %k6, %1 \n\t"
185 "cmpl $63, %k6 \n\t"
186 " jb 3b \n\t"
187 "mov %2, %0 \n\t"
188 "movl %k6, (%0) \n\t"
189 "5: \n\t"
190 "addl %8, %k0 \n\t"
191 "shr $2, %k0 \n\t"
192 : "=&q"(coeff_count), "+m"(last), "+m"(index), "+&r"(c->low),
193 "=&r"(bit), "+&r"(c->range), "=&r"(state)
194 : "r"(c), "m"(minusindex), "m"(significant_coeff_ctx_base),
195 "m"(sig_off), "m"(last_coeff_ctx_base),
196 "i"(offsetof(CABACContext, bytestream)),
197 "i"(offsetof(CABACContext, bytestream_end)),
198 "i"(H264_LAST_COEFF_FLAG_OFFSET_8x8_OFFSET) TABLES_ARG
199 : "%"REG_c, "memory"
200 );
201 return coeff_count;
202 }
203 #endif /* HAVE_7REGS && BROKEN_COMPILER */
204
205 #endif /* HAVE_INLINE_ASM */
206 #endif /* AVCODEC_X86_H264_I386_H */
207