ac3dsp_init.c 4.07 KB
Newer Older
1 2 3 4
/*
 * x86-optimized AC-3 DSP utils
 * Copyright (c) 2011 Justin Ruggles
 *
5
 * This file is part of Libav.
6
 *
7
 * Libav is free software; you can redistribute it and/or
8 9 10 11
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
12
 * Libav is distributed in the hope that it will be useful,
13 14 15 16 17
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with Libav; if not, write to the Free Software
19 20 21
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

22
#include "libavutil/x86/asm.h"
23 24 25 26 27 28 29
#include "dsputil_mmx.h"
#include "libavcodec/ac3dsp.h"

extern void ff_ac3_exponent_min_mmx   (uint8_t *exp, int num_reuse_blocks, int nb_coefs);
extern void ff_ac3_exponent_min_mmxext(uint8_t *exp, int num_reuse_blocks, int nb_coefs);
extern void ff_ac3_exponent_min_sse2  (uint8_t *exp, int num_reuse_blocks, int nb_coefs);

30 31 32 33
extern int ff_ac3_max_msb_abs_int16_mmx  (const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_mmx2 (const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_sse2 (const int16_t *src, int len);
extern int ff_ac3_max_msb_abs_int16_ssse3(const int16_t *src, int len);
34

35 36 37 38 39 40
extern void ff_ac3_lshift_int16_mmx (int16_t *src, unsigned int len, unsigned int shift);
extern void ff_ac3_lshift_int16_sse2(int16_t *src, unsigned int len, unsigned int shift);

extern void ff_ac3_rshift_int32_mmx (int32_t *src, unsigned int len, unsigned int shift);
extern void ff_ac3_rshift_int32_sse2(int32_t *src, unsigned int len, unsigned int shift);

41 42 43 44
extern void ff_float_to_fixed24_3dnow(int32_t *dst, const float *src, unsigned int len);
extern void ff_float_to_fixed24_sse  (int32_t *dst, const float *src, unsigned int len);
extern void ff_float_to_fixed24_sse2 (int32_t *dst, const float *src, unsigned int len);

45 46
extern int ff_ac3_compute_mantissa_size_sse2(uint16_t mant_cnt[6][16]);

47 48 49 50
extern void ff_ac3_extract_exponents_3dnow(uint8_t *exp, int32_t *coef, int nb_coefs);
extern void ff_ac3_extract_exponents_sse2 (uint8_t *exp, int32_t *coef, int nb_coefs);
extern void ff_ac3_extract_exponents_ssse3(uint8_t *exp, int32_t *coef, int nb_coefs);

51
av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact)
52
{
53
#if HAVE_YASM
54 55 56 57
    int mm_flags = av_get_cpu_flags();

    if (mm_flags & AV_CPU_FLAG_MMX) {
        c->ac3_exponent_min = ff_ac3_exponent_min_mmx;
58
        c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx;
59 60
        c->ac3_lshift_int16 = ff_ac3_lshift_int16_mmx;
        c->ac3_rshift_int32 = ff_ac3_rshift_int32_mmx;
61
    }
62
    if (mm_flags & AV_CPU_FLAG_3DNOW && HAVE_AMD3DNOW) {
63
        c->extract_exponents = ff_ac3_extract_exponents_3dnow;
64 65 66 67
        if (!bit_exact) {
            c->float_to_fixed24 = ff_float_to_fixed24_3dnow;
        }
    }
68
    if (mm_flags & AV_CPU_FLAG_MMXEXT && HAVE_MMXEXT) {
69
        c->ac3_exponent_min = ff_ac3_exponent_min_mmxext;
70
        c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_mmx2;
71
    }
72 73 74
    if (mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE) {
        c->float_to_fixed24 = ff_float_to_fixed24_sse;
    }
75 76
    if (mm_flags & AV_CPU_FLAG_SSE2 && HAVE_SSE) {
        c->ac3_exponent_min = ff_ac3_exponent_min_sse2;
77
        c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_sse2;
78
        c->float_to_fixed24 = ff_float_to_fixed24_sse2;
79
        c->compute_mantissa_size = ff_ac3_compute_mantissa_size_sse2;
80
        c->extract_exponents = ff_ac3_extract_exponents_sse2;
81 82 83 84
        if (!(mm_flags & AV_CPU_FLAG_SSE2SLOW)) {
            c->ac3_lshift_int16 = ff_ac3_lshift_int16_sse2;
            c->ac3_rshift_int32 = ff_ac3_rshift_int32_sse2;
        }
85 86 87
    }
    if (mm_flags & AV_CPU_FLAG_SSSE3 && HAVE_SSSE3) {
        c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_ssse3;
88 89 90
        if (!(mm_flags & AV_CPU_FLAG_ATOM)) {
            c->extract_exponents = ff_ac3_extract_exponents_ssse3;
        }
91 92 93
    }
#endif
}