зеркало из https://github.com/mozilla/gecko-dev.git
bug 1476231 add ffmpeg floating point real FFT functions to ffvpx when MOZ_LIBAV_FFT is configured r=jya
New files are from ffmpeg n3.4-1-g587fadaef1. MozReview-Commit-ID: E9eLPySCpCz --HG-- rename : media/libav/libavcodec/avfft.c => media/ffvpx/libavcodec/avfft.c rename : media/libav/libavcodec/avfft.h => media/ffvpx/libavcodec/avfft.h rename : media/libav/libavcodec/dct.h => media/ffvpx/libavcodec/dct.h rename : media/libav/libavcodec/fft-internal.h => media/ffvpx/libavcodec/fft-internal.h rename : media/libav/libavcodec/fft.h => media/ffvpx/libavcodec/fft.h rename : media/libav/libavcodec/fft_float.c => media/ffvpx/libavcodec/fft_float.c rename : media/libav/libavcodec/fft_template.c => media/ffvpx/libavcodec/fft_template.c rename : media/libav/libavcodec/rdft.c => media/ffvpx/libavcodec/rdft.c rename : media/libav/libavcodec/rdft.h => media/ffvpx/libavcodec/rdft.h rename : media/libav/libavcodec/x86/fft.asm => media/ffvpx/libavcodec/x86/fft.asm rename : media/libav/libavcodec/x86/fft.h => media/ffvpx/libavcodec/x86/fft.h rename : media/libav/libavcodec/x86/fft_init.c => media/ffvpx/libavcodec/x86/fft_init.c extra : rebase_source : 35a960b3248bb23dbf56c19baad91b665094fbac
This commit is contained in:
Родитель
2ab37ac350
Коммит
b99ba5e660
|
@ -11,4 +11,11 @@
|
|||
#define HAVE_AVX2_EXTERNAL 0
|
||||
#endif
|
||||
|
||||
#ifdef MOZ_LIBAV_FFT
|
||||
#undef CONFIG_FFT
|
||||
#undef CONFIG_RDFT
|
||||
#define CONFIG_FFT 1
|
||||
#define CONFIG_RDFT 1
|
||||
#endif
|
||||
|
||||
#endif
|
||||
|
|
|
@ -53,6 +53,9 @@ av_parser_close
|
|||
av_parser_init
|
||||
av_parser_next
|
||||
av_parser_parse2
|
||||
av_rdft_calc
|
||||
av_rdft_end
|
||||
av_rdft_init
|
||||
av_register_codec_parser
|
||||
av_register_hwaccel
|
||||
av_shrink_packet
|
||||
|
|
|
@ -0,0 +1,145 @@
|
|||
/*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#include "libavutil/attributes.h"
|
||||
#include "libavutil/mem.h"
|
||||
#include "avfft.h"
|
||||
#include "fft.h"
|
||||
#include "rdft.h"
|
||||
#include "dct.h"
|
||||
|
||||
/* FFT */
|
||||
|
||||
FFTContext *av_fft_init(int nbits, int inverse)
|
||||
{
|
||||
FFTContext *s = av_mallocz(sizeof(*s));
|
||||
|
||||
if (s && ff_fft_init(s, nbits, inverse))
|
||||
av_freep(&s);
|
||||
|
||||
return s;
|
||||
}
|
||||
|
||||
void av_fft_permute(FFTContext *s, FFTComplex *z)
|
||||
{
|
||||
s->fft_permute(s, z);
|
||||
}
|
||||
|
||||
void av_fft_calc(FFTContext *s, FFTComplex *z)
|
||||
{
|
||||
s->fft_calc(s, z);
|
||||
}
|
||||
|
||||
av_cold void av_fft_end(FFTContext *s)
|
||||
{
|
||||
if (s) {
|
||||
ff_fft_end(s);
|
||||
av_free(s);
|
||||
}
|
||||
}
|
||||
|
||||
#if CONFIG_MDCT
|
||||
|
||||
FFTContext *av_mdct_init(int nbits, int inverse, double scale)
|
||||
{
|
||||
FFTContext *s = av_malloc(sizeof(*s));
|
||||
|
||||
if (s && ff_mdct_init(s, nbits, inverse, scale))
|
||||
av_freep(&s);
|
||||
|
||||
return s;
|
||||
}
|
||||
|
||||
void av_imdct_calc(FFTContext *s, FFTSample *output, const FFTSample *input)
|
||||
{
|
||||
s->imdct_calc(s, output, input);
|
||||
}
|
||||
|
||||
void av_imdct_half(FFTContext *s, FFTSample *output, const FFTSample *input)
|
||||
{
|
||||
s->imdct_half(s, output, input);
|
||||
}
|
||||
|
||||
void av_mdct_calc(FFTContext *s, FFTSample *output, const FFTSample *input)
|
||||
{
|
||||
s->mdct_calc(s, output, input);
|
||||
}
|
||||
|
||||
av_cold void av_mdct_end(FFTContext *s)
|
||||
{
|
||||
if (s) {
|
||||
ff_mdct_end(s);
|
||||
av_free(s);
|
||||
}
|
||||
}
|
||||
|
||||
#endif /* CONFIG_MDCT */
|
||||
|
||||
#if CONFIG_RDFT
|
||||
|
||||
RDFTContext *av_rdft_init(int nbits, enum RDFTransformType trans)
|
||||
{
|
||||
RDFTContext *s = av_malloc(sizeof(*s));
|
||||
|
||||
if (s && ff_rdft_init(s, nbits, trans))
|
||||
av_freep(&s);
|
||||
|
||||
return s;
|
||||
}
|
||||
|
||||
void av_rdft_calc(RDFTContext *s, FFTSample *data)
|
||||
{
|
||||
s->rdft_calc(s, data);
|
||||
}
|
||||
|
||||
av_cold void av_rdft_end(RDFTContext *s)
|
||||
{
|
||||
if (s) {
|
||||
ff_rdft_end(s);
|
||||
av_free(s);
|
||||
}
|
||||
}
|
||||
|
||||
#endif /* CONFIG_RDFT */
|
||||
|
||||
#if CONFIG_DCT
|
||||
|
||||
DCTContext *av_dct_init(int nbits, enum DCTTransformType inverse)
|
||||
{
|
||||
DCTContext *s = av_malloc(sizeof(*s));
|
||||
|
||||
if (s && ff_dct_init(s, nbits, inverse))
|
||||
av_freep(&s);
|
||||
|
||||
return s;
|
||||
}
|
||||
|
||||
void av_dct_calc(DCTContext *s, FFTSample *data)
|
||||
{
|
||||
s->dct_calc(s, data);
|
||||
}
|
||||
|
||||
av_cold void av_dct_end(DCTContext *s)
|
||||
{
|
||||
if (s) {
|
||||
ff_dct_end(s);
|
||||
av_free(s);
|
||||
}
|
||||
}
|
||||
|
||||
#endif /* CONFIG_DCT */
|
|
@ -0,0 +1,118 @@
|
|||
/*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#ifndef AVCODEC_AVFFT_H
|
||||
#define AVCODEC_AVFFT_H
|
||||
|
||||
/**
|
||||
* @file
|
||||
* @ingroup lavc_fft
|
||||
* FFT functions
|
||||
*/
|
||||
|
||||
/**
|
||||
* @defgroup lavc_fft FFT functions
|
||||
* @ingroup lavc_misc
|
||||
*
|
||||
* @{
|
||||
*/
|
||||
|
||||
typedef float FFTSample;
|
||||
|
||||
typedef struct FFTComplex {
|
||||
FFTSample re, im;
|
||||
} FFTComplex;
|
||||
|
||||
typedef struct FFTContext FFTContext;
|
||||
|
||||
/**
|
||||
* Set up a complex FFT.
|
||||
* @param nbits log2 of the length of the input array
|
||||
* @param inverse if 0 perform the forward transform, if 1 perform the inverse
|
||||
*/
|
||||
FFTContext *av_fft_init(int nbits, int inverse);
|
||||
|
||||
/**
|
||||
* Do the permutation needed BEFORE calling ff_fft_calc().
|
||||
*/
|
||||
void av_fft_permute(FFTContext *s, FFTComplex *z);
|
||||
|
||||
/**
|
||||
* Do a complex FFT with the parameters defined in av_fft_init(). The
|
||||
* input data must be permuted before. No 1.0/sqrt(n) normalization is done.
|
||||
*/
|
||||
void av_fft_calc(FFTContext *s, FFTComplex *z);
|
||||
|
||||
void av_fft_end(FFTContext *s);
|
||||
|
||||
FFTContext *av_mdct_init(int nbits, int inverse, double scale);
|
||||
void av_imdct_calc(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void av_imdct_half(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void av_mdct_calc(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void av_mdct_end(FFTContext *s);
|
||||
|
||||
/* Real Discrete Fourier Transform */
|
||||
|
||||
enum RDFTransformType {
|
||||
DFT_R2C,
|
||||
IDFT_C2R,
|
||||
IDFT_R2C,
|
||||
DFT_C2R,
|
||||
};
|
||||
|
||||
typedef struct RDFTContext RDFTContext;
|
||||
|
||||
/**
|
||||
* Set up a real FFT.
|
||||
* @param nbits log2 of the length of the input array
|
||||
* @param trans the type of transform
|
||||
*/
|
||||
RDFTContext *av_rdft_init(int nbits, enum RDFTransformType trans);
|
||||
void av_rdft_calc(RDFTContext *s, FFTSample *data);
|
||||
void av_rdft_end(RDFTContext *s);
|
||||
|
||||
/* Discrete Cosine Transform */
|
||||
|
||||
typedef struct DCTContext DCTContext;
|
||||
|
||||
enum DCTTransformType {
|
||||
DCT_II = 0,
|
||||
DCT_III,
|
||||
DCT_I,
|
||||
DST_I,
|
||||
};
|
||||
|
||||
/**
|
||||
* Set up DCT.
|
||||
*
|
||||
* @param nbits size of the input array:
|
||||
* (1 << nbits) for DCT-II, DCT-III and DST-I
|
||||
* (1 << nbits) + 1 for DCT-I
|
||||
* @param type the type of transform
|
||||
*
|
||||
* @note the first element of the input of DST-I is ignored
|
||||
*/
|
||||
DCTContext *av_dct_init(int nbits, enum DCTTransformType type);
|
||||
void av_dct_calc(DCTContext *s, FFTSample *data);
|
||||
void av_dct_end (DCTContext *s);
|
||||
|
||||
/**
|
||||
* @}
|
||||
*/
|
||||
|
||||
#endif /* AVCODEC_AVFFT_H */
|
|
@ -0,0 +1,69 @@
|
|||
/*
|
||||
* (I)DCT Transforms
|
||||
* Copyright (c) 2009 Peter Ross <pross@xvid.org>
|
||||
* Copyright (c) 2010 Alex Converse <alex.converse@gmail.com>
|
||||
* Copyright (c) 2010 Vitor Sessak
|
||||
*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#if !defined(AVCODEC_DCT_H) && (!defined(FFT_FLOAT) || FFT_FLOAT)
|
||||
#define AVCODEC_DCT_H
|
||||
|
||||
#include <stddef.h>
|
||||
#include <stdint.h>
|
||||
|
||||
#include "rdft.h"
|
||||
|
||||
struct DCTContext {
|
||||
int nbits;
|
||||
int inverse;
|
||||
RDFTContext rdft;
|
||||
const float *costab;
|
||||
FFTSample *csc2;
|
||||
void (*dct_calc)(struct DCTContext *s, FFTSample *data);
|
||||
void (*dct32)(FFTSample *out, const FFTSample *in);
|
||||
};
|
||||
|
||||
/**
|
||||
* Set up DCT.
|
||||
* @param nbits size of the input array:
|
||||
* (1 << nbits) for DCT-II, DCT-III and DST-I
|
||||
* (1 << nbits) + 1 for DCT-I
|
||||
*
|
||||
* @note the first element of the input of DST-I is ignored
|
||||
*/
|
||||
int ff_dct_init(DCTContext *s, int nbits, enum DCTTransformType type);
|
||||
void ff_dct_end (DCTContext *s);
|
||||
|
||||
void ff_dct_init_x86(DCTContext *s);
|
||||
|
||||
void ff_fdct_ifast(int16_t *data);
|
||||
void ff_fdct_ifast248(int16_t *data);
|
||||
void ff_jpeg_fdct_islow_8(int16_t *data);
|
||||
void ff_jpeg_fdct_islow_10(int16_t *data);
|
||||
void ff_fdct248_islow_8(int16_t *data);
|
||||
void ff_fdct248_islow_10(int16_t *data);
|
||||
|
||||
void ff_j_rev_dct(int16_t *data);
|
||||
void ff_j_rev_dct4(int16_t *data);
|
||||
void ff_j_rev_dct2(int16_t *data);
|
||||
void ff_j_rev_dct1(int16_t *data);
|
||||
void ff_jref_idct_put(uint8_t *dest, ptrdiff_t line_size, int16_t *block);
|
||||
void ff_jref_idct_add(uint8_t *dest, ptrdiff_t line_size, int16_t *block);
|
||||
|
||||
#endif /* AVCODEC_DCT_H */
|
|
@ -6,7 +6,9 @@
|
|||
|
||||
#include "avcodec.h"
|
||||
|
||||
typedef struct FFTContext FFTContext;
|
||||
typedef struct H264PredContext H264PredContext;
|
||||
typedef struct RDFTContext RDFTContext;
|
||||
typedef struct VideoDSPContext VideoDSPContext;
|
||||
typedef struct VP8DSPContext VP8DSPContext;
|
||||
typedef struct VP9DSPContext VP9DSPContext;
|
||||
|
@ -856,6 +858,11 @@ AVBitStreamFilter ff_noise_bsf;
|
|||
AVBitStreamFilter ff_remove_extradata_bsf;
|
||||
AVBitStreamFilter ff_text2movsub_bsf;
|
||||
|
||||
void ff_fft_init_aarch64(FFTContext *s) {}
|
||||
void ff_fft_init_arm(FFTContext *s) {}
|
||||
void ff_fft_init_mips(FFTContext *s) {}
|
||||
void ff_fft_init_ppc(FFTContext *s) {}
|
||||
void ff_rdft_init_arm(RDFTContext *s) {}
|
||||
void ff_h264_pred_init_aarch64(H264PredContext *h, int codec_id,
|
||||
const int bit_depth,
|
||||
const int chroma_format_idc) {}
|
||||
|
|
|
@ -0,0 +1,94 @@
|
|||
/*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#ifndef AVCODEC_FFT_INTERNAL_H
|
||||
#define AVCODEC_FFT_INTERNAL_H
|
||||
|
||||
#if FFT_FLOAT
|
||||
|
||||
#define FIX15(v) (v)
|
||||
#define sqrthalf (float)M_SQRT1_2
|
||||
|
||||
#define BF(x, y, a, b) do { \
|
||||
x = a - b; \
|
||||
y = a + b; \
|
||||
} while (0)
|
||||
|
||||
#define CMUL(dre, dim, are, aim, bre, bim) do { \
|
||||
(dre) = (are) * (bre) - (aim) * (bim); \
|
||||
(dim) = (are) * (bim) + (aim) * (bre); \
|
||||
} while (0)
|
||||
|
||||
#else
|
||||
|
||||
#define SCALE_FLOAT(a, bits) lrint((a) * (double)(1 << (bits)))
|
||||
|
||||
#if FFT_FIXED_32
|
||||
|
||||
#define CMUL(dre, dim, are, aim, bre, bim) do { \
|
||||
int64_t accu; \
|
||||
(accu) = (int64_t)(bre) * (are); \
|
||||
(accu) -= (int64_t)(bim) * (aim); \
|
||||
(dre) = (int)(((accu) + 0x40000000) >> 31); \
|
||||
(accu) = (int64_t)(bre) * (aim); \
|
||||
(accu) += (int64_t)(bim) * (are); \
|
||||
(dim) = (int)(((accu) + 0x40000000) >> 31); \
|
||||
} while (0)
|
||||
|
||||
#define FIX15(a) av_clip(SCALE_FLOAT(a, 31), -2147483647, 2147483647)
|
||||
|
||||
#else /* FFT_FIXED_32 */
|
||||
|
||||
#include "fft.h"
|
||||
#include "mathops.h"
|
||||
|
||||
void ff_mdct_calcw_c(FFTContext *s, FFTDouble *output, const FFTSample *input);
|
||||
|
||||
#define FIX15(a) av_clip(SCALE_FLOAT(a, 15), -32767, 32767)
|
||||
|
||||
#define sqrthalf ((int16_t)((1<<15)*M_SQRT1_2))
|
||||
|
||||
#define BF(x, y, a, b) do { \
|
||||
x = (a - b) >> 1; \
|
||||
y = (a + b) >> 1; \
|
||||
} while (0)
|
||||
|
||||
#define CMULS(dre, dim, are, aim, bre, bim, sh) do { \
|
||||
(dre) = (MUL16(are, bre) - MUL16(aim, bim)) >> sh; \
|
||||
(dim) = (MUL16(are, bim) + MUL16(aim, bre)) >> sh; \
|
||||
} while (0)
|
||||
|
||||
#define CMUL(dre, dim, are, aim, bre, bim) \
|
||||
CMULS(dre, dim, are, aim, bre, bim, 15)
|
||||
|
||||
#define CMULL(dre, dim, are, aim, bre, bim) \
|
||||
CMULS(dre, dim, are, aim, bre, bim, 0)
|
||||
|
||||
#endif /* FFT_FIXED_32 */
|
||||
|
||||
#endif /* FFT_FLOAT */
|
||||
|
||||
#define ff_imdct_calc_c FFT_NAME(ff_imdct_calc_c)
|
||||
#define ff_imdct_half_c FFT_NAME(ff_imdct_half_c)
|
||||
#define ff_mdct_calc_c FFT_NAME(ff_mdct_calc_c)
|
||||
|
||||
void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void ff_imdct_half_c(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void ff_mdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
|
||||
#endif /* AVCODEC_FFT_INTERNAL_H */
|
|
@ -0,0 +1,175 @@
|
|||
/*
|
||||
* Copyright (c) 2000, 2001, 2002 Fabrice Bellard
|
||||
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
|
||||
*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#ifndef AVCODEC_FFT_H
|
||||
#define AVCODEC_FFT_H
|
||||
|
||||
#ifndef FFT_FLOAT
|
||||
#define FFT_FLOAT 1
|
||||
#endif
|
||||
|
||||
#ifndef FFT_FIXED_32
|
||||
#define FFT_FIXED_32 0
|
||||
#endif
|
||||
|
||||
#include <stdint.h>
|
||||
#include "config.h"
|
||||
#include "libavutil/mem.h"
|
||||
|
||||
#if FFT_FLOAT
|
||||
|
||||
#include "avfft.h"
|
||||
|
||||
#define FFT_NAME(x) x
|
||||
|
||||
typedef float FFTDouble;
|
||||
|
||||
#else
|
||||
|
||||
#if FFT_FIXED_32
|
||||
|
||||
#define Q31(x) (int)((x)*2147483648.0 + 0.5)
|
||||
#define FFT_NAME(x) x ## _fixed_32
|
||||
|
||||
typedef int32_t FFTSample;
|
||||
|
||||
#else /* FFT_FIXED_32 */
|
||||
|
||||
#define FFT_NAME(x) x ## _fixed
|
||||
|
||||
typedef int16_t FFTSample;
|
||||
|
||||
#endif /* FFT_FIXED_32 */
|
||||
|
||||
typedef struct FFTComplex {
|
||||
FFTSample re, im;
|
||||
} FFTComplex;
|
||||
|
||||
typedef int FFTDouble;
|
||||
typedef struct FFTContext FFTContext;
|
||||
|
||||
#endif /* FFT_FLOAT */
|
||||
|
||||
typedef struct FFTDComplex {
|
||||
FFTDouble re, im;
|
||||
} FFTDComplex;
|
||||
|
||||
/* FFT computation */
|
||||
|
||||
enum fft_permutation_type {
|
||||
FF_FFT_PERM_DEFAULT,
|
||||
FF_FFT_PERM_SWAP_LSBS,
|
||||
FF_FFT_PERM_AVX,
|
||||
};
|
||||
|
||||
enum mdct_permutation_type {
|
||||
FF_MDCT_PERM_NONE,
|
||||
FF_MDCT_PERM_INTERLEAVE,
|
||||
};
|
||||
|
||||
struct FFTContext {
|
||||
int nbits;
|
||||
int inverse;
|
||||
uint16_t *revtab;
|
||||
FFTComplex *tmp_buf;
|
||||
int mdct_size; /* size of MDCT (i.e. number of input data * 2) */
|
||||
int mdct_bits; /* n = 2^nbits */
|
||||
/* pre/post rotation tables */
|
||||
FFTSample *tcos;
|
||||
FFTSample *tsin;
|
||||
/**
|
||||
* Do the permutation needed BEFORE calling fft_calc().
|
||||
*/
|
||||
void (*fft_permute)(struct FFTContext *s, FFTComplex *z);
|
||||
/**
|
||||
* Do a complex FFT with the parameters defined in ff_fft_init(). The
|
||||
* input data must be permuted before. No 1.0/sqrt(n) normalization is done.
|
||||
*/
|
||||
void (*fft_calc)(struct FFTContext *s, FFTComplex *z);
|
||||
void (*imdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void (*imdct_half)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void (*mdct_calc)(struct FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void (*mdct_calcw)(struct FFTContext *s, FFTDouble *output, const FFTSample *input);
|
||||
enum fft_permutation_type fft_permutation;
|
||||
enum mdct_permutation_type mdct_permutation;
|
||||
uint32_t *revtab32;
|
||||
};
|
||||
|
||||
#if CONFIG_HARDCODED_TABLES
|
||||
#define COSTABLE_CONST const
|
||||
#else
|
||||
#define COSTABLE_CONST
|
||||
#endif
|
||||
|
||||
#define COSTABLE(size) \
|
||||
COSTABLE_CONST DECLARE_ALIGNED(32, FFTSample, FFT_NAME(ff_cos_##size))[size/2]
|
||||
|
||||
extern COSTABLE(16);
|
||||
extern COSTABLE(32);
|
||||
extern COSTABLE(64);
|
||||
extern COSTABLE(128);
|
||||
extern COSTABLE(256);
|
||||
extern COSTABLE(512);
|
||||
extern COSTABLE(1024);
|
||||
extern COSTABLE(2048);
|
||||
extern COSTABLE(4096);
|
||||
extern COSTABLE(8192);
|
||||
extern COSTABLE(16384);
|
||||
extern COSTABLE(32768);
|
||||
extern COSTABLE(65536);
|
||||
extern COSTABLE(131072);
|
||||
extern COSTABLE_CONST FFTSample* const FFT_NAME(ff_cos_tabs)[18];
|
||||
|
||||
#define ff_init_ff_cos_tabs FFT_NAME(ff_init_ff_cos_tabs)
|
||||
|
||||
/**
|
||||
* Initialize the cosine table in ff_cos_tabs[index]
|
||||
* @param index index in ff_cos_tabs array of the table to initialize
|
||||
*/
|
||||
void ff_init_ff_cos_tabs(int index);
|
||||
|
||||
#define ff_fft_init FFT_NAME(ff_fft_init)
|
||||
#define ff_fft_end FFT_NAME(ff_fft_end)
|
||||
|
||||
/**
|
||||
* Set up a complex FFT.
|
||||
* @param nbits log2 of the length of the input array
|
||||
* @param inverse if 0 perform the forward transform, if 1 perform the inverse
|
||||
*/
|
||||
int ff_fft_init(FFTContext *s, int nbits, int inverse);
|
||||
|
||||
void ff_fft_init_aarch64(FFTContext *s);
|
||||
void ff_fft_init_x86(FFTContext *s);
|
||||
void ff_fft_init_arm(FFTContext *s);
|
||||
void ff_fft_init_mips(FFTContext *s);
|
||||
void ff_fft_init_ppc(FFTContext *s);
|
||||
|
||||
void ff_fft_fixed_init_arm(FFTContext *s);
|
||||
|
||||
void ff_fft_end(FFTContext *s);
|
||||
|
||||
#define ff_mdct_init FFT_NAME(ff_mdct_init)
|
||||
#define ff_mdct_end FFT_NAME(ff_mdct_end)
|
||||
|
||||
int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale);
|
||||
void ff_mdct_end(FFTContext *s);
|
||||
|
||||
#endif /* AVCODEC_FFT_H */
|
|
@ -0,0 +1,21 @@
|
|||
/*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#define FFT_FLOAT 1
|
||||
#define FFT_FIXED_32 0
|
||||
#include "fft_template.c"
|
|
@ -0,0 +1,613 @@
|
|||
/*
|
||||
* FFT/IFFT transforms
|
||||
* Copyright (c) 2008 Loren Merritt
|
||||
* Copyright (c) 2002 Fabrice Bellard
|
||||
* Partly based on libdjbfft by D. J. Bernstein
|
||||
*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
/**
|
||||
* @file
|
||||
* FFT/IFFT transforms.
|
||||
*/
|
||||
|
||||
#include <stdlib.h>
|
||||
#include <string.h>
|
||||
#include "libavutil/mathematics.h"
|
||||
#include "libavutil/thread.h"
|
||||
#include "fft.h"
|
||||
#include "fft-internal.h"
|
||||
|
||||
#if FFT_FIXED_32
|
||||
#include "fft_table.h"
|
||||
|
||||
static void av_cold fft_lut_init(void)
|
||||
{
|
||||
int n = 0;
|
||||
ff_fft_lut_init(ff_fft_offsets_lut, 0, 1 << 17, &n);
|
||||
}
|
||||
|
||||
#else /* FFT_FIXED_32 */
|
||||
|
||||
/* cos(2*pi*x/n) for 0<=x<=n/4, followed by its reverse */
|
||||
#if !CONFIG_HARDCODED_TABLES
|
||||
COSTABLE(16);
|
||||
COSTABLE(32);
|
||||
COSTABLE(64);
|
||||
COSTABLE(128);
|
||||
COSTABLE(256);
|
||||
COSTABLE(512);
|
||||
COSTABLE(1024);
|
||||
COSTABLE(2048);
|
||||
COSTABLE(4096);
|
||||
COSTABLE(8192);
|
||||
COSTABLE(16384);
|
||||
COSTABLE(32768);
|
||||
COSTABLE(65536);
|
||||
COSTABLE(131072);
|
||||
|
||||
static av_cold void init_ff_cos_tabs(int index)
|
||||
{
|
||||
int i;
|
||||
int m = 1<<index;
|
||||
double freq = 2*M_PI/m;
|
||||
FFTSample *tab = FFT_NAME(ff_cos_tabs)[index];
|
||||
for(i=0; i<=m/4; i++)
|
||||
tab[i] = FIX15(cos(i*freq));
|
||||
for(i=1; i<m/4; i++)
|
||||
tab[m/2-i] = tab[i];
|
||||
}
|
||||
|
||||
typedef struct CosTabsInitOnce {
|
||||
void (*func)(void);
|
||||
AVOnce control;
|
||||
} CosTabsInitOnce;
|
||||
|
||||
#define INIT_FF_COS_TABS_FUNC(index, size) \
|
||||
static av_cold void init_ff_cos_tabs_ ## size (void)\
|
||||
{ \
|
||||
init_ff_cos_tabs(index); \
|
||||
}
|
||||
|
||||
INIT_FF_COS_TABS_FUNC(4, 16)
|
||||
INIT_FF_COS_TABS_FUNC(5, 32)
|
||||
INIT_FF_COS_TABS_FUNC(6, 64)
|
||||
INIT_FF_COS_TABS_FUNC(7, 128)
|
||||
INIT_FF_COS_TABS_FUNC(8, 256)
|
||||
INIT_FF_COS_TABS_FUNC(9, 512)
|
||||
INIT_FF_COS_TABS_FUNC(10, 1024)
|
||||
INIT_FF_COS_TABS_FUNC(11, 2048)
|
||||
INIT_FF_COS_TABS_FUNC(12, 4096)
|
||||
INIT_FF_COS_TABS_FUNC(13, 8192)
|
||||
INIT_FF_COS_TABS_FUNC(14, 16384)
|
||||
INIT_FF_COS_TABS_FUNC(15, 32768)
|
||||
INIT_FF_COS_TABS_FUNC(16, 65536)
|
||||
INIT_FF_COS_TABS_FUNC(17, 131072)
|
||||
|
||||
static CosTabsInitOnce cos_tabs_init_once[] = {
|
||||
{ NULL },
|
||||
{ NULL },
|
||||
{ NULL },
|
||||
{ NULL },
|
||||
{ init_ff_cos_tabs_16, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_32, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_64, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_128, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_256, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_512, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_1024, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_2048, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_4096, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_8192, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_16384, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_32768, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_65536, AV_ONCE_INIT },
|
||||
{ init_ff_cos_tabs_131072, AV_ONCE_INIT },
|
||||
};
|
||||
|
||||
#endif
|
||||
COSTABLE_CONST FFTSample * const FFT_NAME(ff_cos_tabs)[] = {
|
||||
NULL, NULL, NULL, NULL,
|
||||
FFT_NAME(ff_cos_16),
|
||||
FFT_NAME(ff_cos_32),
|
||||
FFT_NAME(ff_cos_64),
|
||||
FFT_NAME(ff_cos_128),
|
||||
FFT_NAME(ff_cos_256),
|
||||
FFT_NAME(ff_cos_512),
|
||||
FFT_NAME(ff_cos_1024),
|
||||
FFT_NAME(ff_cos_2048),
|
||||
FFT_NAME(ff_cos_4096),
|
||||
FFT_NAME(ff_cos_8192),
|
||||
FFT_NAME(ff_cos_16384),
|
||||
FFT_NAME(ff_cos_32768),
|
||||
FFT_NAME(ff_cos_65536),
|
||||
FFT_NAME(ff_cos_131072),
|
||||
};
|
||||
|
||||
#endif /* FFT_FIXED_32 */
|
||||
|
||||
static void fft_permute_c(FFTContext *s, FFTComplex *z);
|
||||
static void fft_calc_c(FFTContext *s, FFTComplex *z);
|
||||
|
||||
static int split_radix_permutation(int i, int n, int inverse)
|
||||
{
|
||||
int m;
|
||||
if(n <= 2) return i&1;
|
||||
m = n >> 1;
|
||||
if(!(i&m)) return split_radix_permutation(i, m, inverse)*2;
|
||||
m >>= 1;
|
||||
if(inverse == !(i&m)) return split_radix_permutation(i, m, inverse)*4 + 1;
|
||||
else return split_radix_permutation(i, m, inverse)*4 - 1;
|
||||
}
|
||||
|
||||
av_cold void ff_init_ff_cos_tabs(int index)
|
||||
{
|
||||
#if (!CONFIG_HARDCODED_TABLES) && (!FFT_FIXED_32)
|
||||
ff_thread_once(&cos_tabs_init_once[index].control, cos_tabs_init_once[index].func);
|
||||
#endif
|
||||
}
|
||||
|
||||
static const int avx_tab[] = {
|
||||
0, 4, 1, 5, 8, 12, 9, 13, 2, 6, 3, 7, 10, 14, 11, 15
|
||||
};
|
||||
|
||||
static int is_second_half_of_fft32(int i, int n)
|
||||
{
|
||||
if (n <= 32)
|
||||
return i >= 16;
|
||||
else if (i < n/2)
|
||||
return is_second_half_of_fft32(i, n/2);
|
||||
else if (i < 3*n/4)
|
||||
return is_second_half_of_fft32(i - n/2, n/4);
|
||||
else
|
||||
return is_second_half_of_fft32(i - 3*n/4, n/4);
|
||||
}
|
||||
|
||||
static av_cold void fft_perm_avx(FFTContext *s)
|
||||
{
|
||||
int i;
|
||||
int n = 1 << s->nbits;
|
||||
|
||||
for (i = 0; i < n; i += 16) {
|
||||
int k;
|
||||
if (is_second_half_of_fft32(i, n)) {
|
||||
for (k = 0; k < 16; k++)
|
||||
s->revtab[-split_radix_permutation(i + k, n, s->inverse) & (n - 1)] =
|
||||
i + avx_tab[k];
|
||||
|
||||
} else {
|
||||
for (k = 0; k < 16; k++) {
|
||||
int j = i + k;
|
||||
j = (j & ~7) | ((j >> 1) & 3) | ((j << 2) & 4);
|
||||
s->revtab[-split_radix_permutation(i + k, n, s->inverse) & (n - 1)] = j;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
|
||||
{
|
||||
int i, j, n;
|
||||
|
||||
s->revtab = NULL;
|
||||
s->revtab32 = NULL;
|
||||
|
||||
if (nbits < 2 || nbits > 17)
|
||||
goto fail;
|
||||
s->nbits = nbits;
|
||||
n = 1 << nbits;
|
||||
|
||||
if (nbits <= 16) {
|
||||
s->revtab = av_malloc(n * sizeof(uint16_t));
|
||||
if (!s->revtab)
|
||||
goto fail;
|
||||
} else {
|
||||
s->revtab32 = av_malloc(n * sizeof(uint32_t));
|
||||
if (!s->revtab32)
|
||||
goto fail;
|
||||
}
|
||||
s->tmp_buf = av_malloc(n * sizeof(FFTComplex));
|
||||
if (!s->tmp_buf)
|
||||
goto fail;
|
||||
s->inverse = inverse;
|
||||
s->fft_permutation = FF_FFT_PERM_DEFAULT;
|
||||
|
||||
s->fft_permute = fft_permute_c;
|
||||
s->fft_calc = fft_calc_c;
|
||||
#if CONFIG_MDCT
|
||||
s->imdct_calc = ff_imdct_calc_c;
|
||||
s->imdct_half = ff_imdct_half_c;
|
||||
s->mdct_calc = ff_mdct_calc_c;
|
||||
#endif
|
||||
|
||||
#if FFT_FIXED_32
|
||||
{
|
||||
static AVOnce control = AV_ONCE_INIT;
|
||||
ff_thread_once(&control, fft_lut_init);
|
||||
}
|
||||
#else /* FFT_FIXED_32 */
|
||||
#if FFT_FLOAT
|
||||
if (ARCH_AARCH64) ff_fft_init_aarch64(s);
|
||||
if (ARCH_ARM) ff_fft_init_arm(s);
|
||||
if (ARCH_PPC) ff_fft_init_ppc(s);
|
||||
if (ARCH_X86) ff_fft_init_x86(s);
|
||||
if (CONFIG_MDCT) s->mdct_calcw = s->mdct_calc;
|
||||
if (HAVE_MIPSFPU) ff_fft_init_mips(s);
|
||||
#else
|
||||
if (CONFIG_MDCT) s->mdct_calcw = ff_mdct_calcw_c;
|
||||
if (ARCH_ARM) ff_fft_fixed_init_arm(s);
|
||||
#endif
|
||||
for(j=4; j<=nbits; j++) {
|
||||
ff_init_ff_cos_tabs(j);
|
||||
}
|
||||
#endif /* FFT_FIXED_32 */
|
||||
|
||||
|
||||
if (s->fft_permutation == FF_FFT_PERM_AVX) {
|
||||
fft_perm_avx(s);
|
||||
} else {
|
||||
for(i=0; i<n; i++) {
|
||||
int k;
|
||||
j = i;
|
||||
if (s->fft_permutation == FF_FFT_PERM_SWAP_LSBS)
|
||||
j = (j&~3) | ((j>>1)&1) | ((j<<1)&2);
|
||||
k = -split_radix_permutation(i, n, s->inverse) & (n-1);
|
||||
if (s->revtab)
|
||||
s->revtab[k] = j;
|
||||
if (s->revtab32)
|
||||
s->revtab32[k] = j;
|
||||
}
|
||||
}
|
||||
|
||||
return 0;
|
||||
fail:
|
||||
av_freep(&s->revtab);
|
||||
av_freep(&s->revtab32);
|
||||
av_freep(&s->tmp_buf);
|
||||
return -1;
|
||||
}
|
||||
|
||||
static void fft_permute_c(FFTContext *s, FFTComplex *z)
|
||||
{
|
||||
int j, np;
|
||||
const uint16_t *revtab = s->revtab;
|
||||
const uint32_t *revtab32 = s->revtab32;
|
||||
np = 1 << s->nbits;
|
||||
/* TODO: handle split-radix permute in a more optimal way, probably in-place */
|
||||
if (revtab) {
|
||||
for(j=0;j<np;j++) s->tmp_buf[revtab[j]] = z[j];
|
||||
} else
|
||||
for(j=0;j<np;j++) s->tmp_buf[revtab32[j]] = z[j];
|
||||
|
||||
memcpy(z, s->tmp_buf, np * sizeof(FFTComplex));
|
||||
}
|
||||
|
||||
av_cold void ff_fft_end(FFTContext *s)
|
||||
{
|
||||
av_freep(&s->revtab);
|
||||
av_freep(&s->revtab32);
|
||||
av_freep(&s->tmp_buf);
|
||||
}
|
||||
|
||||
#if FFT_FIXED_32
|
||||
|
||||
static void fft_calc_c(FFTContext *s, FFTComplex *z) {
|
||||
|
||||
int nbits, i, n, num_transforms, offset, step;
|
||||
int n4, n2, n34;
|
||||
unsigned tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;
|
||||
FFTComplex *tmpz;
|
||||
const int fft_size = (1 << s->nbits);
|
||||
int64_t accu;
|
||||
|
||||
num_transforms = (0x2aab >> (16 - s->nbits)) | 1;
|
||||
|
||||
for (n=0; n<num_transforms; n++){
|
||||
offset = ff_fft_offsets_lut[n] << 2;
|
||||
tmpz = z + offset;
|
||||
|
||||
tmp1 = tmpz[0].re + (unsigned)tmpz[1].re;
|
||||
tmp5 = tmpz[2].re + (unsigned)tmpz[3].re;
|
||||
tmp2 = tmpz[0].im + (unsigned)tmpz[1].im;
|
||||
tmp6 = tmpz[2].im + (unsigned)tmpz[3].im;
|
||||
tmp3 = tmpz[0].re - (unsigned)tmpz[1].re;
|
||||
tmp8 = tmpz[2].im - (unsigned)tmpz[3].im;
|
||||
tmp4 = tmpz[0].im - (unsigned)tmpz[1].im;
|
||||
tmp7 = tmpz[2].re - (unsigned)tmpz[3].re;
|
||||
|
||||
tmpz[0].re = tmp1 + tmp5;
|
||||
tmpz[2].re = tmp1 - tmp5;
|
||||
tmpz[0].im = tmp2 + tmp6;
|
||||
tmpz[2].im = tmp2 - tmp6;
|
||||
tmpz[1].re = tmp3 + tmp8;
|
||||
tmpz[3].re = tmp3 - tmp8;
|
||||
tmpz[1].im = tmp4 - tmp7;
|
||||
tmpz[3].im = tmp4 + tmp7;
|
||||
}
|
||||
|
||||
if (fft_size < 8)
|
||||
return;
|
||||
|
||||
num_transforms = (num_transforms >> 1) | 1;
|
||||
|
||||
for (n=0; n<num_transforms; n++){
|
||||
offset = ff_fft_offsets_lut[n] << 3;
|
||||
tmpz = z + offset;
|
||||
|
||||
tmp1 = tmpz[4].re + (unsigned)tmpz[5].re;
|
||||
tmp3 = tmpz[6].re + (unsigned)tmpz[7].re;
|
||||
tmp2 = tmpz[4].im + (unsigned)tmpz[5].im;
|
||||
tmp4 = tmpz[6].im + (unsigned)tmpz[7].im;
|
||||
tmp5 = tmp1 + tmp3;
|
||||
tmp7 = tmp1 - tmp3;
|
||||
tmp6 = tmp2 + tmp4;
|
||||
tmp8 = tmp2 - tmp4;
|
||||
|
||||
tmp1 = tmpz[4].re - (unsigned)tmpz[5].re;
|
||||
tmp2 = tmpz[4].im - (unsigned)tmpz[5].im;
|
||||
tmp3 = tmpz[6].re - (unsigned)tmpz[7].re;
|
||||
tmp4 = tmpz[6].im - (unsigned)tmpz[7].im;
|
||||
|
||||
tmpz[4].re = tmpz[0].re - tmp5;
|
||||
tmpz[0].re = tmpz[0].re + tmp5;
|
||||
tmpz[4].im = tmpz[0].im - tmp6;
|
||||
tmpz[0].im = tmpz[0].im + tmp6;
|
||||
tmpz[6].re = tmpz[2].re - tmp8;
|
||||
tmpz[2].re = tmpz[2].re + tmp8;
|
||||
tmpz[6].im = tmpz[2].im + tmp7;
|
||||
tmpz[2].im = tmpz[2].im - tmp7;
|
||||
|
||||
accu = (int64_t)Q31(M_SQRT1_2)*(int)(tmp1 + tmp2);
|
||||
tmp5 = (int32_t)((accu + 0x40000000) >> 31);
|
||||
accu = (int64_t)Q31(M_SQRT1_2)*(int)(tmp3 - tmp4);
|
||||
tmp7 = (int32_t)((accu + 0x40000000) >> 31);
|
||||
accu = (int64_t)Q31(M_SQRT1_2)*(int)(tmp2 - tmp1);
|
||||
tmp6 = (int32_t)((accu + 0x40000000) >> 31);
|
||||
accu = (int64_t)Q31(M_SQRT1_2)*(int)(tmp3 + tmp4);
|
||||
tmp8 = (int32_t)((accu + 0x40000000) >> 31);
|
||||
tmp1 = tmp5 + tmp7;
|
||||
tmp3 = tmp5 - tmp7;
|
||||
tmp2 = tmp6 + tmp8;
|
||||
tmp4 = tmp6 - tmp8;
|
||||
|
||||
tmpz[5].re = tmpz[1].re - tmp1;
|
||||
tmpz[1].re = tmpz[1].re + tmp1;
|
||||
tmpz[5].im = tmpz[1].im - tmp2;
|
||||
tmpz[1].im = tmpz[1].im + tmp2;
|
||||
tmpz[7].re = tmpz[3].re - tmp4;
|
||||
tmpz[3].re = tmpz[3].re + tmp4;
|
||||
tmpz[7].im = tmpz[3].im + tmp3;
|
||||
tmpz[3].im = tmpz[3].im - tmp3;
|
||||
}
|
||||
|
||||
step = 1 << ((MAX_LOG2_NFFT-4) - 4);
|
||||
n4 = 4;
|
||||
|
||||
for (nbits=4; nbits<=s->nbits; nbits++){
|
||||
n2 = 2*n4;
|
||||
n34 = 3*n4;
|
||||
num_transforms = (num_transforms >> 1) | 1;
|
||||
|
||||
for (n=0; n<num_transforms; n++){
|
||||
const FFTSample *w_re_ptr = ff_w_tab_sr + step;
|
||||
const FFTSample *w_im_ptr = ff_w_tab_sr + MAX_FFT_SIZE/(4*16) - step;
|
||||
offset = ff_fft_offsets_lut[n] << nbits;
|
||||
tmpz = z + offset;
|
||||
|
||||
tmp5 = tmpz[ n2].re + (unsigned)tmpz[n34].re;
|
||||
tmp1 = tmpz[ n2].re - (unsigned)tmpz[n34].re;
|
||||
tmp6 = tmpz[ n2].im + (unsigned)tmpz[n34].im;
|
||||
tmp2 = tmpz[ n2].im - (unsigned)tmpz[n34].im;
|
||||
|
||||
tmpz[ n2].re = tmpz[ 0].re - tmp5;
|
||||
tmpz[ 0].re = tmpz[ 0].re + tmp5;
|
||||
tmpz[ n2].im = tmpz[ 0].im - tmp6;
|
||||
tmpz[ 0].im = tmpz[ 0].im + tmp6;
|
||||
tmpz[n34].re = tmpz[n4].re - tmp2;
|
||||
tmpz[ n4].re = tmpz[n4].re + tmp2;
|
||||
tmpz[n34].im = tmpz[n4].im + tmp1;
|
||||
tmpz[ n4].im = tmpz[n4].im - tmp1;
|
||||
|
||||
for (i=1; i<n4; i++){
|
||||
FFTSample w_re = w_re_ptr[0];
|
||||
FFTSample w_im = w_im_ptr[0];
|
||||
accu = (int64_t)w_re*tmpz[ n2+i].re;
|
||||
accu += (int64_t)w_im*tmpz[ n2+i].im;
|
||||
tmp1 = (int32_t)((accu + 0x40000000) >> 31);
|
||||
accu = (int64_t)w_re*tmpz[ n2+i].im;
|
||||
accu -= (int64_t)w_im*tmpz[ n2+i].re;
|
||||
tmp2 = (int32_t)((accu + 0x40000000) >> 31);
|
||||
accu = (int64_t)w_re*tmpz[n34+i].re;
|
||||
accu -= (int64_t)w_im*tmpz[n34+i].im;
|
||||
tmp3 = (int32_t)((accu + 0x40000000) >> 31);
|
||||
accu = (int64_t)w_re*tmpz[n34+i].im;
|
||||
accu += (int64_t)w_im*tmpz[n34+i].re;
|
||||
tmp4 = (int32_t)((accu + 0x40000000) >> 31);
|
||||
|
||||
tmp5 = tmp1 + tmp3;
|
||||
tmp1 = tmp1 - tmp3;
|
||||
tmp6 = tmp2 + tmp4;
|
||||
tmp2 = tmp2 - tmp4;
|
||||
|
||||
tmpz[ n2+i].re = tmpz[ i].re - tmp5;
|
||||
tmpz[ i].re = tmpz[ i].re + tmp5;
|
||||
tmpz[ n2+i].im = tmpz[ i].im - tmp6;
|
||||
tmpz[ i].im = tmpz[ i].im + tmp6;
|
||||
tmpz[n34+i].re = tmpz[n4+i].re - tmp2;
|
||||
tmpz[ n4+i].re = tmpz[n4+i].re + tmp2;
|
||||
tmpz[n34+i].im = tmpz[n4+i].im + tmp1;
|
||||
tmpz[ n4+i].im = tmpz[n4+i].im - tmp1;
|
||||
|
||||
w_re_ptr += step;
|
||||
w_im_ptr -= step;
|
||||
}
|
||||
}
|
||||
step >>= 1;
|
||||
n4 <<= 1;
|
||||
}
|
||||
}
|
||||
|
||||
#else /* FFT_FIXED_32 */
|
||||
|
||||
#define BUTTERFLIES(a0,a1,a2,a3) {\
|
||||
BF(t3, t5, t5, t1);\
|
||||
BF(a2.re, a0.re, a0.re, t5);\
|
||||
BF(a3.im, a1.im, a1.im, t3);\
|
||||
BF(t4, t6, t2, t6);\
|
||||
BF(a3.re, a1.re, a1.re, t4);\
|
||||
BF(a2.im, a0.im, a0.im, t6);\
|
||||
}
|
||||
|
||||
// force loading all the inputs before storing any.
|
||||
// this is slightly slower for small data, but avoids store->load aliasing
|
||||
// for addresses separated by large powers of 2.
|
||||
#define BUTTERFLIES_BIG(a0,a1,a2,a3) {\
|
||||
FFTSample r0=a0.re, i0=a0.im, r1=a1.re, i1=a1.im;\
|
||||
BF(t3, t5, t5, t1);\
|
||||
BF(a2.re, a0.re, r0, t5);\
|
||||
BF(a3.im, a1.im, i1, t3);\
|
||||
BF(t4, t6, t2, t6);\
|
||||
BF(a3.re, a1.re, r1, t4);\
|
||||
BF(a2.im, a0.im, i0, t6);\
|
||||
}
|
||||
|
||||
#define TRANSFORM(a0,a1,a2,a3,wre,wim) {\
|
||||
CMUL(t1, t2, a2.re, a2.im, wre, -wim);\
|
||||
CMUL(t5, t6, a3.re, a3.im, wre, wim);\
|
||||
BUTTERFLIES(a0,a1,a2,a3)\
|
||||
}
|
||||
|
||||
#define TRANSFORM_ZERO(a0,a1,a2,a3) {\
|
||||
t1 = a2.re;\
|
||||
t2 = a2.im;\
|
||||
t5 = a3.re;\
|
||||
t6 = a3.im;\
|
||||
BUTTERFLIES(a0,a1,a2,a3)\
|
||||
}
|
||||
|
||||
/* z[0...8n-1], w[1...2n-1] */
|
||||
#define PASS(name)\
|
||||
static void name(FFTComplex *z, const FFTSample *wre, unsigned int n)\
|
||||
{\
|
||||
FFTDouble t1, t2, t3, t4, t5, t6;\
|
||||
int o1 = 2*n;\
|
||||
int o2 = 4*n;\
|
||||
int o3 = 6*n;\
|
||||
const FFTSample *wim = wre+o1;\
|
||||
n--;\
|
||||
\
|
||||
TRANSFORM_ZERO(z[0],z[o1],z[o2],z[o3]);\
|
||||
TRANSFORM(z[1],z[o1+1],z[o2+1],z[o3+1],wre[1],wim[-1]);\
|
||||
do {\
|
||||
z += 2;\
|
||||
wre += 2;\
|
||||
wim -= 2;\
|
||||
TRANSFORM(z[0],z[o1],z[o2],z[o3],wre[0],wim[0]);\
|
||||
TRANSFORM(z[1],z[o1+1],z[o2+1],z[o3+1],wre[1],wim[-1]);\
|
||||
} while(--n);\
|
||||
}
|
||||
|
||||
PASS(pass)
|
||||
#undef BUTTERFLIES
|
||||
#define BUTTERFLIES BUTTERFLIES_BIG
|
||||
PASS(pass_big)
|
||||
|
||||
#define DECL_FFT(n,n2,n4)\
|
||||
static void fft##n(FFTComplex *z)\
|
||||
{\
|
||||
fft##n2(z);\
|
||||
fft##n4(z+n4*2);\
|
||||
fft##n4(z+n4*3);\
|
||||
pass(z,FFT_NAME(ff_cos_##n),n4/2);\
|
||||
}
|
||||
|
||||
static void fft4(FFTComplex *z)
|
||||
{
|
||||
FFTDouble t1, t2, t3, t4, t5, t6, t7, t8;
|
||||
|
||||
BF(t3, t1, z[0].re, z[1].re);
|
||||
BF(t8, t6, z[3].re, z[2].re);
|
||||
BF(z[2].re, z[0].re, t1, t6);
|
||||
BF(t4, t2, z[0].im, z[1].im);
|
||||
BF(t7, t5, z[2].im, z[3].im);
|
||||
BF(z[3].im, z[1].im, t4, t8);
|
||||
BF(z[3].re, z[1].re, t3, t7);
|
||||
BF(z[2].im, z[0].im, t2, t5);
|
||||
}
|
||||
|
||||
static void fft8(FFTComplex *z)
|
||||
{
|
||||
FFTDouble t1, t2, t3, t4, t5, t6;
|
||||
|
||||
fft4(z);
|
||||
|
||||
BF(t1, z[5].re, z[4].re, -z[5].re);
|
||||
BF(t2, z[5].im, z[4].im, -z[5].im);
|
||||
BF(t5, z[7].re, z[6].re, -z[7].re);
|
||||
BF(t6, z[7].im, z[6].im, -z[7].im);
|
||||
|
||||
BUTTERFLIES(z[0],z[2],z[4],z[6]);
|
||||
TRANSFORM(z[1],z[3],z[5],z[7],sqrthalf,sqrthalf);
|
||||
}
|
||||
|
||||
#if !CONFIG_SMALL
|
||||
static void fft16(FFTComplex *z)
|
||||
{
|
||||
FFTDouble t1, t2, t3, t4, t5, t6;
|
||||
FFTSample cos_16_1 = FFT_NAME(ff_cos_16)[1];
|
||||
FFTSample cos_16_3 = FFT_NAME(ff_cos_16)[3];
|
||||
|
||||
fft8(z);
|
||||
fft4(z+8);
|
||||
fft4(z+12);
|
||||
|
||||
TRANSFORM_ZERO(z[0],z[4],z[8],z[12]);
|
||||
TRANSFORM(z[2],z[6],z[10],z[14],sqrthalf,sqrthalf);
|
||||
TRANSFORM(z[1],z[5],z[9],z[13],cos_16_1,cos_16_3);
|
||||
TRANSFORM(z[3],z[7],z[11],z[15],cos_16_3,cos_16_1);
|
||||
}
|
||||
#else
|
||||
DECL_FFT(16,8,4)
|
||||
#endif
|
||||
DECL_FFT(32,16,8)
|
||||
DECL_FFT(64,32,16)
|
||||
DECL_FFT(128,64,32)
|
||||
DECL_FFT(256,128,64)
|
||||
DECL_FFT(512,256,128)
|
||||
#if !CONFIG_SMALL
|
||||
#define pass pass_big
|
||||
#endif
|
||||
DECL_FFT(1024,512,256)
|
||||
DECL_FFT(2048,1024,512)
|
||||
DECL_FFT(4096,2048,1024)
|
||||
DECL_FFT(8192,4096,2048)
|
||||
DECL_FFT(16384,8192,4096)
|
||||
DECL_FFT(32768,16384,8192)
|
||||
DECL_FFT(65536,32768,16384)
|
||||
DECL_FFT(131072,65536,32768)
|
||||
|
||||
static void (* const fft_dispatch[])(FFTComplex*) = {
|
||||
fft4, fft8, fft16, fft32, fft64, fft128, fft256, fft512, fft1024,
|
||||
fft2048, fft4096, fft8192, fft16384, fft32768, fft65536, fft131072
|
||||
};
|
||||
|
||||
static void fft_calc_c(FFTContext *s, FFTComplex *z)
|
||||
{
|
||||
fft_dispatch[s->nbits-2](z);
|
||||
}
|
||||
#endif /* FFT_FIXED_32 */
|
|
@ -72,6 +72,13 @@ if not CONFIG['MOZ_FFVPX_FLACONLY']:
|
|||
'vp9recon.c'
|
||||
]
|
||||
|
||||
if CONFIG['MOZ_LIBAV_FFT']:
|
||||
SOURCES += [
|
||||
'avfft.c',
|
||||
'fft_float.c',
|
||||
'rdft.c',
|
||||
]
|
||||
|
||||
SYMBOLS_FILE = 'avcodec.symbols'
|
||||
NoVisibilityFlags()
|
||||
|
||||
|
|
|
@ -0,0 +1,117 @@
|
|||
/*
|
||||
* (I)RDFT transforms
|
||||
* Copyright (c) 2009 Alex Converse <alex dot converse at gmail dot com>
|
||||
*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
#include <stdlib.h>
|
||||
#include <math.h>
|
||||
#include "libavutil/mathematics.h"
|
||||
#include "rdft.h"
|
||||
|
||||
/**
|
||||
* @file
|
||||
* (Inverse) Real Discrete Fourier Transforms.
|
||||
*/
|
||||
|
||||
/** Map one real FFT into two parallel real even and odd FFTs. Then interleave
|
||||
* the two real FFTs into one complex FFT. Unmangle the results.
|
||||
* ref: http://www.engineeringproductivitytools.com/stuff/T0001/PT10.HTM
|
||||
*/
|
||||
static void rdft_calc_c(RDFTContext *s, FFTSample *data)
|
||||
{
|
||||
int i, i1, i2;
|
||||
FFTComplex ev, od, odsum;
|
||||
const int n = 1 << s->nbits;
|
||||
const float k1 = 0.5;
|
||||
const float k2 = 0.5 - s->inverse;
|
||||
const FFTSample *tcos = s->tcos;
|
||||
const FFTSample *tsin = s->tsin;
|
||||
|
||||
if (!s->inverse) {
|
||||
s->fft.fft_permute(&s->fft, (FFTComplex*)data);
|
||||
s->fft.fft_calc(&s->fft, (FFTComplex*)data);
|
||||
}
|
||||
/* i=0 is a special case because of packing, the DC term is real, so we
|
||||
are going to throw the N/2 term (also real) in with it. */
|
||||
ev.re = data[0];
|
||||
data[0] = ev.re+data[1];
|
||||
data[1] = ev.re-data[1];
|
||||
|
||||
#define RDFT_UNMANGLE(sign0, sign1) \
|
||||
for (i = 1; i < (n>>2); i++) { \
|
||||
i1 = 2*i; \
|
||||
i2 = n-i1; \
|
||||
/* Separate even and odd FFTs */ \
|
||||
ev.re = k1*(data[i1 ]+data[i2 ]); \
|
||||
od.im = k2*(data[i2 ]-data[i1 ]); \
|
||||
ev.im = k1*(data[i1+1]-data[i2+1]); \
|
||||
od.re = k2*(data[i1+1]+data[i2+1]); \
|
||||
/* Apply twiddle factors to the odd FFT and add to the even FFT */ \
|
||||
odsum.re = od.re*tcos[i] sign0 od.im*tsin[i]; \
|
||||
odsum.im = od.im*tcos[i] sign1 od.re*tsin[i]; \
|
||||
data[i1 ] = ev.re + odsum.re; \
|
||||
data[i1+1] = ev.im + odsum.im; \
|
||||
data[i2 ] = ev.re - odsum.re; \
|
||||
data[i2+1] = odsum.im - ev.im; \
|
||||
}
|
||||
|
||||
if (s->negative_sin) {
|
||||
RDFT_UNMANGLE(+,-)
|
||||
} else {
|
||||
RDFT_UNMANGLE(-,+)
|
||||
}
|
||||
|
||||
data[2*i+1]=s->sign_convention*data[2*i+1];
|
||||
if (s->inverse) {
|
||||
data[0] *= k1;
|
||||
data[1] *= k1;
|
||||
s->fft.fft_permute(&s->fft, (FFTComplex*)data);
|
||||
s->fft.fft_calc(&s->fft, (FFTComplex*)data);
|
||||
}
|
||||
}
|
||||
|
||||
av_cold int ff_rdft_init(RDFTContext *s, int nbits, enum RDFTransformType trans)
|
||||
{
|
||||
int n = 1 << nbits;
|
||||
int ret;
|
||||
|
||||
s->nbits = nbits;
|
||||
s->inverse = trans == IDFT_C2R || trans == DFT_C2R;
|
||||
s->sign_convention = trans == IDFT_R2C || trans == DFT_C2R ? 1 : -1;
|
||||
s->negative_sin = trans == DFT_C2R || trans == DFT_R2C;
|
||||
|
||||
if (nbits < 4 || nbits > 16)
|
||||
return AVERROR(EINVAL);
|
||||
|
||||
if ((ret = ff_fft_init(&s->fft, nbits-1, trans == IDFT_C2R || trans == IDFT_R2C)) < 0)
|
||||
return ret;
|
||||
|
||||
ff_init_ff_cos_tabs(nbits);
|
||||
s->tcos = ff_cos_tabs[nbits];
|
||||
s->tsin = ff_cos_tabs[nbits] + (n >> 2);
|
||||
s->rdft_calc = rdft_calc_c;
|
||||
|
||||
if (ARCH_ARM) ff_rdft_init_arm(s);
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
av_cold void ff_rdft_end(RDFTContext *s)
|
||||
{
|
||||
ff_fft_end(&s->fft);
|
||||
}
|
|
@ -0,0 +1,52 @@
|
|||
/*
|
||||
* (I)RDFT transforms
|
||||
* Copyright (c) 2009 Alex Converse <alex dot converse at gmail dot com>
|
||||
*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#if !defined(AVCODEC_RDFT_H) && (!defined(FFT_FLOAT) || FFT_FLOAT)
|
||||
#define AVCODEC_RDFT_H
|
||||
|
||||
#include "config.h"
|
||||
#include "fft.h"
|
||||
|
||||
struct RDFTContext {
|
||||
int nbits;
|
||||
int inverse;
|
||||
int sign_convention;
|
||||
|
||||
/* pre/post rotation tables */
|
||||
const FFTSample *tcos;
|
||||
const FFTSample *tsin;
|
||||
int negative_sin;
|
||||
FFTContext fft;
|
||||
void (*rdft_calc)(struct RDFTContext *s, FFTSample *z);
|
||||
};
|
||||
|
||||
/**
|
||||
* Set up a real FFT.
|
||||
* @param nbits log2 of the length of the input array
|
||||
* @param trans the type of transform
|
||||
*/
|
||||
int ff_rdft_init(RDFTContext *s, int nbits, enum RDFTransformType trans);
|
||||
void ff_rdft_end(RDFTContext *s);
|
||||
|
||||
void ff_rdft_init_arm(RDFTContext *s);
|
||||
|
||||
|
||||
#endif /* AVCODEC_RDFT_H */
|
Разница между файлами не показана из-за своего большого размера
Загрузить разницу
|
@ -0,0 +1,38 @@
|
|||
/*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#ifndef AVCODEC_X86_FFT_H
|
||||
#define AVCODEC_X86_FFT_H
|
||||
|
||||
#include "libavcodec/fft.h"
|
||||
|
||||
void ff_fft_permute_sse(FFTContext *s, FFTComplex *z);
|
||||
void ff_fft_calc_avx(FFTContext *s, FFTComplex *z);
|
||||
void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
|
||||
void ff_fft_calc_3dnow(FFTContext *s, FFTComplex *z);
|
||||
void ff_fft_calc_3dnowext(FFTContext *s, FFTComplex *z);
|
||||
|
||||
void ff_imdct_calc_3dnow(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void ff_imdct_half_3dnow(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void ff_imdct_calc_3dnowext(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void ff_imdct_half_3dnowext(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void ff_imdct_calc_sse(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void ff_imdct_half_sse(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
void ff_imdct_half_avx(FFTContext *s, FFTSample *output, const FFTSample *input);
|
||||
|
||||
#endif /* AVCODEC_X86_FFT_H */
|
|
@ -0,0 +1,61 @@
|
|||
/*
|
||||
* This file is part of FFmpeg.
|
||||
*
|
||||
* FFmpeg is free software; you can redistribute it and/or
|
||||
* 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.
|
||||
*
|
||||
* FFmpeg is distributed in the hope that it will be useful,
|
||||
* 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
|
||||
* License along with FFmpeg; if not, write to the Free Software
|
||||
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
||||
*/
|
||||
|
||||
#include "config.h"
|
||||
|
||||
#include "libavutil/attributes.h"
|
||||
#include "libavutil/cpu.h"
|
||||
#include "libavutil/x86/cpu.h"
|
||||
|
||||
#include "fft.h"
|
||||
|
||||
av_cold void ff_fft_init_x86(FFTContext *s)
|
||||
{
|
||||
int cpu_flags = av_get_cpu_flags();
|
||||
|
||||
if (s->nbits > 16)
|
||||
return;
|
||||
|
||||
#if ARCH_X86_32
|
||||
if (EXTERNAL_AMD3DNOW(cpu_flags)) {
|
||||
s->imdct_calc = ff_imdct_calc_3dnow;
|
||||
s->imdct_half = ff_imdct_half_3dnow;
|
||||
s->fft_calc = ff_fft_calc_3dnow;
|
||||
}
|
||||
|
||||
if (EXTERNAL_AMD3DNOWEXT(cpu_flags)) {
|
||||
s->imdct_calc = ff_imdct_calc_3dnowext;
|
||||
s->imdct_half = ff_imdct_half_3dnowext;
|
||||
s->fft_calc = ff_fft_calc_3dnowext;
|
||||
}
|
||||
#endif /* ARCH_X86_32 */
|
||||
|
||||
if (EXTERNAL_SSE(cpu_flags)) {
|
||||
s->imdct_calc = ff_imdct_calc_sse;
|
||||
s->imdct_half = ff_imdct_half_sse;
|
||||
s->fft_permute = ff_fft_permute_sse;
|
||||
s->fft_calc = ff_fft_calc_sse;
|
||||
s->fft_permutation = FF_FFT_PERM_SWAP_LSBS;
|
||||
}
|
||||
|
||||
if (EXTERNAL_AVX_FAST(cpu_flags) && s->nbits >= 5) {
|
||||
s->imdct_half = ff_imdct_half_avx;
|
||||
s->fft_calc = ff_fft_calc_avx;
|
||||
s->fft_permutation = FF_FFT_PERM_AVX;
|
||||
}
|
||||
}
|
|
@ -30,6 +30,12 @@ SOURCES += [
|
|||
'vp9mc_16bpp.asm',
|
||||
]
|
||||
|
||||
if CONFIG['MOZ_LIBAV_FFT']:
|
||||
SOURCES += [
|
||||
'fft.asm',
|
||||
'fft_init.c',
|
||||
]
|
||||
|
||||
FINAL_LIBRARY = 'mozavcodec'
|
||||
|
||||
include('/media/ffvpx/ffvpxcommon.mozbuild')
|
||||
|
|
Загрузка…
Ссылка в новой задаче