x86: Move some variable declarations below the appropriat #ifdef.
[libav.git] / libavcodec / x86 / ac3dsp_mmx.c
CommitLineData
dda3f0ef
JR
1/*
2 * x86-optimized AC-3 DSP utils
3 * Copyright (c) 2011 Justin Ruggles
4 *
2912e87a 5 * This file is part of Libav.
dda3f0ef 6 *
2912e87a 7 * Libav is free software; you can redistribute it and/or
dda3f0ef
JR
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 *
2912e87a 12 * Libav is distributed in the hope that it will be useful,
dda3f0ef
JR
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
2912e87a 18 * License along with Libav; if not, write to the Free Software
dda3f0ef
JR
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22#include "libavutil/x86_cpu.h"
23#include "dsputil_mmx.h"
24#include "libavcodec/ac3dsp.h"
25
26extern void ff_ac3_exponent_min_mmx (uint8_t *exp, int num_reuse_blocks, int nb_coefs);
27extern void ff_ac3_exponent_min_mmxext(uint8_t *exp, int num_reuse_blocks, int nb_coefs);
28extern void ff_ac3_exponent_min_sse2 (uint8_t *exp, int num_reuse_blocks, int nb_coefs);
29
fbb6b49d
JR
30extern int ff_ac3_max_msb_abs_int16_mmx (const int16_t *src, int len);
31extern int ff_ac3_max_msb_abs_int16_mmxext(const int16_t *src, int len);
32extern int ff_ac3_max_msb_abs_int16_sse2 (const int16_t *src, int len);
33extern int ff_ac3_max_msb_abs_int16_ssse3 (const int16_t *src, int len);
34
f1efbca5
JR
35extern void ff_ac3_lshift_int16_mmx (int16_t *src, unsigned int len, unsigned int shift);
36extern void ff_ac3_lshift_int16_sse2(int16_t *src, unsigned int len, unsigned int shift);
37
38extern void ff_ac3_rshift_int32_mmx (int32_t *src, unsigned int len, unsigned int shift);
39extern void ff_ac3_rshift_int32_sse2(int32_t *src, unsigned int len, unsigned int shift);
40
0f999cfd
JR
41extern void ff_float_to_fixed24_3dnow(int32_t *dst, const float *src, unsigned int len);
42extern void ff_float_to_fixed24_sse (int32_t *dst, const float *src, unsigned int len);
43extern void ff_float_to_fixed24_sse2 (int32_t *dst, const float *src, unsigned int len);
44
6ca23db9
JR
45extern int ff_ac3_compute_mantissa_size_sse2(uint16_t mant_cnt[6][16]);
46
f99a5ef9
JR
47extern void ff_ac3_extract_exponents_3dnow(uint8_t *exp, int32_t *coef, int nb_coefs);
48extern void ff_ac3_extract_exponents_sse2 (uint8_t *exp, int32_t *coef, int nb_coefs);
49extern void ff_ac3_extract_exponents_ssse3(uint8_t *exp, int32_t *coef, int nb_coefs);
50
0f999cfd 51av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact)
dda3f0ef 52{
265980da 53#if HAVE_YASM
dda3f0ef
JR
54 int mm_flags = av_get_cpu_flags();
55
dda3f0ef
JR
56 if (mm_flags & AV_CPU_FLAG_MMX) {
57 c->ac3_exponent_min = ff_ac3_exponent_min_mmx;
fbb6b49d 58 c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx;
f1efbca5
JR
59 c->ac3_lshift_int16 = ff_ac3_lshift_int16_mmx;
60 c->ac3_rshift_int32 = ff_ac3_rshift_int32_mmx;
dda3f0ef 61 }
0f999cfd 62 if (mm_flags & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW) {
f99a5ef9 63 c->extract_exponents = ff_ac3_extract_exponents_3dnow;
0f999cfd
JR
64 if (!bit_exact) {
65 c->float_to_fixed24 = ff_float_to_fixed24_3dnow;
66 }
67 }
dda3f0ef
JR
68 if (mm_flags & AV_CPU_FLAG_MMX2 && HAVE_MMX2) {
69 c->ac3_exponent_min = ff_ac3_exponent_min_mmxext;
fbb6b49d 70 c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmxext;
dda3f0ef 71 }
0f999cfd
JR
72 if (mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE) {
73 c->float_to_fixed24 = ff_float_to_fixed24_sse;
74 }
dda3f0ef
JR
75 if (mm_flags & AV_CPU_FLAG_SSE2 && HAVE_SSE) {
76 c->ac3_exponent_min = ff_ac3_exponent_min_sse2;
fbb6b49d 77 c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_sse2;
0f999cfd 78 c->float_to_fixed24 = ff_float_to_fixed24_sse2;
6ca23db9 79 c->compute_mantissa_size = ff_ac3_compute_mantissa_size_sse2;
f99a5ef9 80 c->extract_exponents = ff_ac3_extract_exponents_sse2;
f1efbca5
JR
81 if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
82 c->ac3_lshift_int16 = ff_ac3_lshift_int16_sse2;
83 c->ac3_rshift_int32 = ff_ac3_rshift_int32_sse2;
84 }
fbb6b49d
JR
85 }
86 if (mm_flags & AV_CPU_FLAG_SSSE3 && HAVE_SSSE3) {
87 c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_ssse3;
f99a5ef9
JR
88 if (!(mm_flags & AV_CPU_FLAG_ATOM)) {
89 c->extract_exponents = ff_ac3_extract_exponents_ssse3;
90 }
dda3f0ef
JR
91 }
92#endif
93}