Merge commit '3d5d46233cd81f78138a6d7418d480af04d3f6c8'

* commit '3d5d46233cd81f78138a6d7418d480af04d3f6c8':
  opus: Factor out imdct15 into a standalone component

Conflicts:
	configure
	libavcodec/opus_celt.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2015-02-02 20:43:13 +01:00
commit 7b32b35bf5
8 changed files with 46 additions and 41 deletions

2
configure vendored
View File

@ -1868,6 +1868,7 @@ CONFIG_EXTRA="
huffyuvencdsp
idctdsp
iirfilter
imdct15
intrax8
lgplv3
llauddsp
@ -2192,6 +2193,7 @@ nellymoser_encoder_select="audio_frame_queue mdct sinewin"
nuv_decoder_select="idctdsp lzo"
on2avc_decoder_select="mdct"
opus_decoder_deps="swresample"
opus_decoder_select="imdct15"
png_decoder_select="zlib"
png_encoder_select="huffyuvencdsp zlib"
prores_decoder_select="blockdsp idctdsp"

View File

@ -66,6 +66,7 @@ OBJS-$(CONFIG_HUFFYUVDSP) += huffyuvdsp.o
OBJS-$(CONFIG_HUFFYUVENCDSP) += huffyuvencdsp.o
OBJS-$(CONFIG_IDCTDSP) += idctdsp.o simple_idct.o jrevdct.o
OBJS-$(CONFIG_IIRFILTER) += iirfilter.o
OBJS-$(CONFIG_IMDCT15) += imdct15.o
OBJS-$(CONFIG_INTRAX8) += intrax8.o intrax8dsp.o
OBJS-$(CONFIG_LIBXVID) += libxvid_rc.o
OBJS-$(CONFIG_LLAUDDSP) += lossless_audiodsp.o
@ -351,8 +352,7 @@ OBJS-$(CONFIG_NUV_DECODER) += nuv.o rtjpeg.o
OBJS-$(CONFIG_NVENC_ENCODER) += nvenc.o
OBJS-$(CONFIG_ON2AVC_DECODER) += on2avc.o on2avcdata.o
OBJS-$(CONFIG_OPUS_DECODER) += opusdec.o opus.o opus_celt.o \
opus_imdct.o opus_silk.o \
vorbis_data.o
opus_silk.o vorbis_data.o
OBJS-$(CONFIG_PAF_AUDIO_DECODER) += pafaudio.o
OBJS-$(CONFIG_PAF_VIDEO_DECODER) += pafvideo.o
OBJS-$(CONFIG_PAM_DECODER) += pnmdec.o pnm.o

View File

@ -3,11 +3,11 @@ OBJS-$(CONFIG_H264CHROMA) += aarch64/h264chroma_init_aarch64.o
OBJS-$(CONFIG_H264DSP) += aarch64/h264dsp_init_aarch64.o
OBJS-$(CONFIG_H264QPEL) += aarch64/h264qpel_init_aarch64.o
OBJS-$(CONFIG_HPELDSP) += aarch64/hpeldsp_init_aarch64.o
OBJS-$(CONFIG_IMDCT15) += aarch64/imdct15_init.o
OBJS-$(CONFIG_MPEGAUDIODSP) += aarch64/mpegaudiodsp_init.o
OBJS-$(CONFIG_NEON_CLOBBER_TEST) += aarch64/neontest.o
OBJS-$(CONFIG_VIDEODSP) += aarch64/videodsp_init.o
OBJS-$(CONFIG_OPUS_DECODER) += aarch64/opus_imdct_init.o
OBJS-$(CONFIG_RV40_DECODER) += aarch64/rv40dsp_init_aarch64.o
OBJS-$(CONFIG_VC1_DECODER) += aarch64/vc1dsp_init_aarch64.o
OBJS-$(CONFIG_VORBIS_DECODER) += aarch64/vorbisdsp_init.o
@ -21,8 +21,8 @@ NEON-OBJS-$(CONFIG_H264DSP) += aarch64/h264dsp_neon.o \
NEON-OBJS-$(CONFIG_H264QPEL) += aarch64/h264qpel_neon.o \
aarch64/hpeldsp_neon.o
NEON-OBJS-$(CONFIG_HPELDSP) += aarch64/hpeldsp_neon.o
NEON-OBJS-$(CONFIG_IMDCT15) += aarch64/imdct15_neon.o
NEON-OBJS-$(CONFIG_MPEGAUDIODSP) += aarch64/mpegaudiodsp_neon.o
NEON-OBJS-$(CONFIG_MDCT) += aarch64/mdct_neon.o
NEON-OBJS-$(CONFIG_OPUS_DECODER) += aarch64/opus_imdct_neon.o
NEON-OBJS-$(CONFIG_VORBIS_DECODER) += aarch64/vorbisdsp_neon.o

View File

@ -21,21 +21,22 @@
#include "libavutil/cpu.h"
#include "libavutil/aarch64/cpu.h"
#include "libavutil/internal.h"
#include "libavcodec/opus_imdct.h"
#include "libavcodec/imdct15.h"
#include "asm-offsets.h"
AV_CHECK_OFFSET(CeltIMDCTContext, exptab, CELT_EXPTAB);
AV_CHECK_OFFSET(CeltIMDCTContext, fft_n, CELT_FFT_N);
AV_CHECK_OFFSET(CeltIMDCTContext, len2, CELT_LEN2);
AV_CHECK_OFFSET(CeltIMDCTContext, len4, CELT_LEN4);
AV_CHECK_OFFSET(CeltIMDCTContext, tmp, CELT_TMP);
AV_CHECK_OFFSET(CeltIMDCTContext, twiddle_exptab, CELT_TWIDDLE);
AV_CHECK_OFFSET(IMDCT15Context, exptab, CELT_EXPTAB);
AV_CHECK_OFFSET(IMDCT15Context, fft_n, CELT_FFT_N);
AV_CHECK_OFFSET(IMDCT15Context, len2, CELT_LEN2);
AV_CHECK_OFFSET(IMDCT15Context, len4, CELT_LEN4);
AV_CHECK_OFFSET(IMDCT15Context, tmp, CELT_TMP);
AV_CHECK_OFFSET(IMDCT15Context, twiddle_exptab, CELT_TWIDDLE);
void ff_celt_imdct_half_neon(CeltIMDCTContext *s, float *dst, const float *src,
void ff_celt_imdct_half_neon(IMDCT15Context *s, float *dst, const float *src,
ptrdiff_t stride, float scale);
void ff_celt_imdct_init_aarch64(CeltIMDCTContext *s)
void ff_imdct15_init_aarch64(IMDCT15Context *s)
{
int cpu_flags = av_get_cpu_flags();

View File

@ -33,8 +33,8 @@
#include "libavutil/common.h"
#include "avfft.h"
#include "imdct15.h"
#include "opus.h"
#include "opus_imdct.h"
// minimal iMDCT size to make SIMD opts easier
#define CELT_MIN_IMDCT_SIZE 120
@ -66,9 +66,9 @@ do { \
(d).im = -ri + ir; \
} while (0)
av_cold void ff_celt_imdct_uninit(CeltIMDCTContext **ps)
av_cold void ff_imdct15_uninit(IMDCT15Context **ps)
{
CeltIMDCTContext *s = *ps;
IMDCT15Context *s = *ps;
int i;
if (!s)
@ -84,12 +84,12 @@ av_cold void ff_celt_imdct_uninit(CeltIMDCTContext **ps)
av_freep(ps);
}
static void celt_imdct_half(CeltIMDCTContext *s, float *dst, const float *src,
ptrdiff_t stride, float scale);
static void imdct15_half(IMDCT15Context *s, float *dst, const float *src,
ptrdiff_t stride, float scale);
av_cold int ff_celt_imdct_init(CeltIMDCTContext **ps, int N)
av_cold int ff_imdct15_init(IMDCT15Context **ps, int N)
{
CeltIMDCTContext *s;
IMDCT15Context *s;
int len2 = 15 * (1 << N);
int len = 2 * len2;
int i, j;
@ -134,16 +134,17 @@ av_cold int ff_celt_imdct_init(CeltIMDCTContext **ps, int N)
for (j = 15; j < 19; j++)
s->exptab[0][j] = s->exptab[0][j - 15];
s->imdct_half = celt_imdct_half;
s->imdct_half = imdct15_half;
if (ARCH_AARCH64)
ff_celt_imdct_init_aarch64(s);
ff_imdct15_init_aarch64(s);
*ps = s;
return 0;
fail:
ff_celt_imdct_uninit(&s);
ff_imdct15_uninit(&s);
return AVERROR(ENOMEM);
}
@ -180,7 +181,8 @@ static void fft5(FFTComplex *out, const FFTComplex *in, ptrdiff_t stride)
out[4].im = in[0].im + z[0][3].im + z[1][2].im + z[2][1].im + z[3][0].im;
}
static void fft15(CeltIMDCTContext *s, FFTComplex *out, const FFTComplex *in, ptrdiff_t stride)
static void fft15(IMDCT15Context *s, FFTComplex *out, const FFTComplex *in,
ptrdiff_t stride)
{
const FFTComplex *exptab = s->exptab[0];
FFTComplex tmp[5];
@ -215,7 +217,7 @@ static void fft15(CeltIMDCTContext *s, FFTComplex *out, const FFTComplex *in, pt
/*
* FFT of the length 15 * (2^N)
*/
static void fft_calc(CeltIMDCTContext *s, FFTComplex *out, const FFTComplex *in,
static void fft_calc(IMDCT15Context *s, FFTComplex *out, const FFTComplex *in,
int N, ptrdiff_t stride)
{
if (N) {
@ -241,8 +243,8 @@ static void fft_calc(CeltIMDCTContext *s, FFTComplex *out, const FFTComplex *in,
fft15(s, out, in, stride);
}
static void celt_imdct_half(CeltIMDCTContext *s, float *dst, const float *src,
ptrdiff_t stride, float scale)
static void imdct15_half(IMDCT15Context *s, float *dst, const float *src,
ptrdiff_t stride, float scale)
{
FFTComplex *z = (FFTComplex *)dst;
const int len8 = s->len4 / 2;

View File

@ -16,14 +16,14 @@
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#ifndef AVCODEC_OPUS_IMDCT_H
#define AVCODEC_OPUS_IMDCT_H
#ifndef AVCODEC_IMDCT15_H
#define AVCODEC_IMDCT15_H
#include <stddef.h>
#include "avfft.h"
typedef struct CeltIMDCTContext {
typedef struct IMDCT15Context {
int fft_n;
int len2;
int len4;
@ -37,21 +37,21 @@ typedef struct CeltIMDCTContext {
/**
* Calculate the middle half of the iMDCT
*/
void (*imdct_half)(struct CeltIMDCTContext *s, float *dst, const float *src,
void (*imdct_half)(struct IMDCT15Context *s, float *dst, const float *src,
ptrdiff_t src_stride, float scale);
} CeltIMDCTContext;
} IMDCT15Context;
/**
* Init an iMDCT of the length 2 * 15 * (2^N)
*/
int ff_celt_imdct_init(CeltIMDCTContext **s, int N);
int ff_imdct15_init(IMDCT15Context **s, int N);
/**
* Free an iMDCT.
*/
void ff_celt_imdct_uninit(CeltIMDCTContext **s);
void ff_imdct15_uninit(IMDCT15Context **s);
void ff_celt_imdct_init_aarch64(CeltIMDCTContext *s);
void ff_imdct15_init_aarch64(IMDCT15Context *s);
#endif /* AVCODEC_OPUS_IMDCT_H */
#endif /* AVCODEC_IMDCT15_H */

View File

@ -28,8 +28,8 @@
#include "libavutil/float_dsp.h"
#include "imdct15.h"
#include "opus.h"
#include "opus_imdct.h"
enum CeltSpread {
CELT_SPREAD_NONE,
@ -61,7 +61,7 @@ typedef struct CeltFrame {
struct CeltContext {
// constant values that do not change during context lifetime
AVCodecContext *avctx;
CeltIMDCTContext *imdct[4];
IMDCT15Context *imdct[4];
AVFloatDSPContext *dsp;
int output_channels;
@ -1983,7 +1983,7 @@ int ff_celt_decode_frame(CeltContext *s, OpusRangeCoder *rc,
int silence = 0;
int transient = 0;
int anticollapse = 0;
CeltIMDCTContext *imdct;
IMDCT15Context *imdct;
float imdct_scale = 1.0;
if (coded_channels != 1 && coded_channels != 2) {
@ -2179,7 +2179,7 @@ void ff_celt_free(CeltContext **ps)
return;
for (i = 0; i < FF_ARRAY_ELEMS(s->imdct); i++)
ff_celt_imdct_uninit(&s->imdct[i]);
ff_imdct15_uninit(&s->imdct[i]);
av_freep(&s->dsp);
av_freep(ps);
@ -2204,7 +2204,7 @@ int ff_celt_init(AVCodecContext *avctx, CeltContext **ps, int output_channels)
s->output_channels = output_channels;
for (i = 0; i < FF_ARRAY_ELEMS(s->imdct); i++) {
ret = ff_celt_imdct_init(&s->imdct[i], i + 3);
ret = ff_imdct15_init(&s->imdct[i], i + 3);
if (ret < 0)
goto fail;
}