Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
F
ffmpeg
Project
Project
Details
Activity
Releases
Cycle Analytics
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Charts
Issues
0
Issues
0
List
Board
Labels
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Charts
Wiki
Wiki
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Charts
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
Stefan Westerfeld
ffmpeg
Commits
9b8e2a87
Commit
9b8e2a87
authored
Dec 15, 2011
by
Nathan Caldwell
Committed by
Alex Converse
Jan 23, 2012
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
aacenc: Deinterleave input samples before processing.
Signed-off-by:
Alex Converse
<
alex.converse@gmail.com
>
parent
04af2efa
Changes
5
Hide whitespace changes
Inline
Side-by-side
Showing
5 changed files
with
80 additions
and
62 deletions
+80
-62
aacenc.c
libavcodec/aacenc.c
+61
-39
aacenc.h
libavcodec/aacenc.h
+5
-1
aacpsy.c
libavcodec/aacpsy.c
+5
-6
psymodel.c
libavcodec/psymodel.c
+6
-10
psymodel.h
libavcodec/psymodel.h
+3
-6
No files found.
libavcodec/aacenc.c
View file @
9b8e2a87
...
...
@@ -143,6 +143,18 @@ static const uint8_t aac_chan_configs[6][5] = {
{
4
,
TYPE_SCE
,
TYPE_CPE
,
TYPE_CPE
,
TYPE_LFE
},
// 6 channels - front center + stereo + back stereo + LFE
};
/**
* Table to remap channels from Libav's default order to AAC order.
*/
static
const
uint8_t
aac_chan_maps
[
AAC_MAX_CHANNELS
][
AAC_MAX_CHANNELS
]
=
{
{
0
},
{
0
,
1
},
{
2
,
0
,
1
},
{
2
,
0
,
1
,
3
},
{
2
,
0
,
1
,
3
,
4
},
{
2
,
0
,
1
,
4
,
5
,
3
},
};
/**
* Make AAC audio config object.
* @see 1.6.2.1 "Syntax - AudioSpecificConfig"
...
...
@@ -172,34 +184,29 @@ static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
float
*
audio
)
{
int
i
,
k
;
const
int
chans
=
s
->
channels
;
const
float
*
lwindow
=
sce
->
ics
.
use_kb_window
[
0
]
?
ff_aac_kbd_long_1024
:
ff_sine_1024
;
const
float
*
swindow
=
sce
->
ics
.
use_kb_window
[
0
]
?
ff_aac_kbd_short_128
:
ff_sine_128
;
const
float
*
pwindow
=
sce
->
ics
.
use_kb_window
[
1
]
?
ff_aac_kbd_short_128
:
ff_sine_128
;
float
*
output
=
sce
->
ret
;
if
(
sce
->
ics
.
window_sequence
[
0
]
!=
EIGHT_SHORT_SEQUENCE
)
{
memcpy
(
output
,
sce
->
saved
,
sizeof
(
float
)
*
1024
);
memcpy
(
output
,
sce
->
saved
,
sizeof
(
output
[
0
]
)
*
1024
);
if
(
sce
->
ics
.
window_sequence
[
0
]
==
LONG_STOP_SEQUENCE
)
{
memset
(
output
,
0
,
sizeof
(
output
[
0
])
*
448
);
for
(
i
=
448
;
i
<
576
;
i
++
)
output
[
i
]
=
sce
->
saved
[
i
]
*
pwindow
[
i
-
448
];
for
(
i
=
576
;
i
<
704
;
i
++
)
output
[
i
]
=
sce
->
saved
[
i
];
}
if
(
sce
->
ics
.
window_sequence
[
0
]
!=
LONG_START_SEQUENCE
)
{
for
(
i
=
0
;
i
<
1024
;
i
++
)
{
output
[
i
+
1024
]
=
audio
[
i
*
chans
]
*
lwindow
[
1024
-
i
-
1
];
sce
->
saved
[
i
]
=
audio
[
i
*
chans
]
*
lwindow
[
i
];
output
[
i
+
1024
]
=
audio
[
i
]
*
lwindow
[
1024
-
i
-
1
];
sce
->
saved
[
i
]
=
audio
[
i
]
*
lwindow
[
i
];
}
}
else
{
for
(
i
=
0
;
i
<
448
;
i
++
)
output
[
i
+
1024
]
=
audio
[
i
*
chans
];
memcpy
(
output
+
1024
,
audio
,
sizeof
(
output
[
0
])
*
448
);
for
(;
i
<
576
;
i
++
)
output
[
i
+
1024
]
=
audio
[
i
*
chans
]
*
swindow
[
576
-
i
-
1
];
output
[
i
+
1024
]
=
audio
[
i
]
*
swindow
[
576
-
i
-
1
];
memset
(
output
+
1024
+
576
,
0
,
sizeof
(
output
[
0
])
*
448
);
for
(
i
=
0
;
i
<
1024
;
i
++
)
sce
->
saved
[
i
]
=
audio
[
i
*
chans
];
memcpy
(
sce
->
saved
,
audio
,
sizeof
(
sce
->
saved
[
0
])
*
1024
);
}
s
->
mdct1024
.
mdct_calc
(
&
s
->
mdct1024
,
sce
->
coeffs
,
output
);
}
else
{
...
...
@@ -207,13 +214,12 @@ static void apply_window_and_mdct(AACEncContext *s, SingleChannelElement *sce,
for
(
i
=
448
+
k
;
i
<
448
+
k
+
256
;
i
++
)
output
[
i
-
448
-
k
]
=
(
i
<
1024
)
?
sce
->
saved
[
i
]
:
audio
[
(
i
-
1024
)
*
chans
];
:
audio
[
i
-
1024
];
s
->
dsp
.
vector_fmul
(
output
,
output
,
k
?
swindow
:
pwindow
,
128
);
s
->
dsp
.
vector_fmul_reverse
(
output
+
128
,
output
+
128
,
swindow
,
128
);
s
->
mdct128
.
mdct_calc
(
&
s
->
mdct128
,
sce
->
coeffs
+
k
,
output
);
}
for
(
i
=
0
;
i
<
1024
;
i
++
)
sce
->
saved
[
i
]
=
audio
[
i
*
chans
];
memcpy
(
sce
->
saved
,
audio
,
sizeof
(
sce
->
saved
[
0
])
*
1024
);
}
}
...
...
@@ -432,11 +438,37 @@ static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s,
put_bits
(
&
s
->
pb
,
12
-
padbits
,
0
);
}
/*
* Deinterleave input samples.
* Channels are reordered from Libav's default order to AAC order.
*/
static
void
deinterleave_input_samples
(
AACEncContext
*
s
,
const
float
*
samples
)
{
int
ch
,
i
;
const
int
sinc
=
s
->
channels
;
const
uint8_t
*
channel_map
=
aac_chan_maps
[
sinc
-
1
];
/* deinterleave and remap input samples */
for
(
ch
=
0
;
ch
<
sinc
;
ch
++
)
{
const
float
*
sptr
=
samples
+
channel_map
[
ch
];
/* copy last 1024 samples of previous frame to the start of the current frame */
memcpy
(
&
s
->
planar_samples
[
ch
][
0
],
&
s
->
planar_samples
[
ch
][
1024
],
1024
*
sizeof
(
s
->
planar_samples
[
0
][
0
]));
/* deinterleave */
for
(
i
=
1024
;
i
<
1024
*
2
;
i
++
)
{
s
->
planar_samples
[
ch
][
i
]
=
*
sptr
;
sptr
+=
sinc
;
}
}
}
static
int
aac_encode_frame
(
AVCodecContext
*
avctx
,
uint8_t
*
frame
,
int
buf_size
,
void
*
data
)
{
AACEncContext
*
s
=
avctx
->
priv_data
;
float
*
samples
=
s
->
samples
,
*
samples2
,
*
la
;
float
*
*
samples
=
s
->
planar_
samples
,
*
samples2
,
*
la
;
ChannelElement
*
cpe
;
int
i
,
ch
,
w
,
g
,
chans
,
tag
,
start_ch
;
int
chan_el_counter
[
4
];
...
...
@@ -444,27 +476,15 @@ static int aac_encode_frame(AVCodecContext *avctx,
if
(
s
->
last_frame
)
return
0
;
if
(
data
)
{
if
(
!
s
->
psypp
)
{
memcpy
(
s
->
samples
+
1024
*
s
->
channels
,
data
,
1024
*
s
->
channels
*
sizeof
(
s
->
samples
[
0
]));
}
else
{
start_ch
=
0
;
samples2
=
s
->
samples
+
1024
*
s
->
channels
;
for
(
i
=
0
;
i
<
s
->
chan_map
[
0
];
i
++
)
{
tag
=
s
->
chan_map
[
i
+
1
];
chans
=
tag
==
TYPE_CPE
?
2
:
1
;
ff_psy_preprocess
(
s
->
psypp
,
(
float
*
)
data
+
start_ch
,
samples2
+
start_ch
,
start_ch
,
chans
);
start_ch
+=
chans
;
}
}
deinterleave_input_samples
(
s
,
data
);
if
(
s
->
psypp
)
ff_psy_preprocess
(
s
->
psypp
,
s
->
planar_samples
,
s
->
channels
);
}
if
(
!
avctx
->
frame_number
)
{
memcpy
(
s
->
samples
,
s
->
samples
+
1024
*
s
->
channels
,
1024
*
s
->
channels
*
sizeof
(
s
->
samples
[
0
]));
if
(
!
avctx
->
frame_number
)
return
0
;
}
start_ch
=
0
;
for
(
i
=
0
;
i
<
s
->
chan_map
[
0
];
i
++
)
{
...
...
@@ -475,8 +495,8 @@ static int aac_encode_frame(AVCodecContext *avctx,
for
(
ch
=
0
;
ch
<
chans
;
ch
++
)
{
IndividualChannelStream
*
ics
=
&
cpe
->
ch
[
ch
].
ics
;
int
cur_channel
=
start_ch
+
ch
;
samples2
=
samples
+
cur_channel
;
la
=
samples2
+
(
448
+
64
)
*
s
->
channels
;
samples2
=
&
samples
[
cur_channel
][
0
]
;
la
=
samples2
+
(
448
+
64
);
if
(
!
data
)
la
=
NULL
;
if
(
tag
==
TYPE_LFE
)
{
...
...
@@ -592,8 +612,7 @@ static int aac_encode_frame(AVCodecContext *avctx,
if
(
!
data
)
s
->
last_frame
=
1
;
memcpy
(
s
->
samples
,
s
->
samples
+
1024
*
s
->
channels
,
1024
*
s
->
channels
*
sizeof
(
s
->
samples
[
0
]));
return
put_bits_count
(
&
s
->
pb
)
>>
3
;
}
...
...
@@ -606,7 +625,7 @@ static av_cold int aac_encode_end(AVCodecContext *avctx)
ff_psy_end
(
&
s
->
psy
);
if
(
s
->
psypp
)
ff_psy_preprocess_end
(
s
->
psypp
);
av_freep
(
&
s
->
samples
);
av_freep
(
&
s
->
buffer
.
samples
);
av_freep
(
&
s
->
cpe
);
return
0
;
}
...
...
@@ -633,10 +652,13 @@ static av_cold int dsp_init(AVCodecContext *avctx, AACEncContext *s)
static
av_cold
int
alloc_buffers
(
AVCodecContext
*
avctx
,
AACEncContext
*
s
)
{
FF_ALLOC
_OR_GOTO
(
avctx
,
s
->
samples
,
2
*
1024
*
s
->
channels
*
sizeof
(
s
->
samples
[
0
]),
alloc_fail
);
FF_ALLOC
Z_OR_GOTO
(
avctx
,
s
->
buffer
.
samples
,
2
*
1024
*
s
->
channels
*
sizeof
(
s
->
buffer
.
samples
[
0
]),
alloc_fail
);
FF_ALLOCZ_OR_GOTO
(
avctx
,
s
->
cpe
,
sizeof
(
ChannelElement
)
*
s
->
chan_map
[
0
],
alloc_fail
);
FF_ALLOCZ_OR_GOTO
(
avctx
,
avctx
->
extradata
,
5
+
FF_INPUT_BUFFER_PADDING_SIZE
,
alloc_fail
);
for
(
int
ch
=
0
;
ch
<
s
->
channels
;
ch
++
)
s
->
planar_samples
[
ch
]
=
s
->
buffer
.
samples
+
2
*
1024
*
ch
;
return
0
;
alloc_fail:
return
AVERROR
(
ENOMEM
);
...
...
libavcodec/aacenc.h
View file @
9b8e2a87
...
...
@@ -58,7 +58,7 @@ typedef struct AACEncContext {
FFTContext
mdct1024
;
///< long (1024 samples) frame transform context
FFTContext
mdct128
;
///< short (128 samples) frame transform context
DSPContext
dsp
;
float
*
samples
;
///< saved preprocessed input
float
*
planar_samples
[
6
];
///< saved preprocessed input
int
samplerate_index
;
///< MPEG-4 samplerate index
int
channels
;
///< channel count
...
...
@@ -73,6 +73,10 @@ typedef struct AACEncContext {
float
lambda
;
DECLARE_ALIGNED
(
16
,
int
,
qcoefs
)[
96
];
///< quantized coefficients
DECLARE_ALIGNED
(
32
,
float
,
scoefs
)[
1024
];
///< scaled coefficients
struct
{
float
*
samples
;
}
buffer
;
}
AACEncContext
;
extern
float
ff_aac_pow34sf_tab
[
428
];
...
...
libavcodec/aacpsy.c
View file @
9b8e2a87
...
...
@@ -400,7 +400,7 @@ static av_unused FFPsyWindowInfo psy_3gpp_window(FFPsyContext *ctx,
int
stay_short
=
0
;
for
(
i
=
0
;
i
<
8
;
i
++
)
{
for
(
j
=
0
;
j
<
128
;
j
++
)
{
v
=
iir_filter
(
la
[
(
i
*
128
+
j
)
*
ctx
->
avctx
->
channels
],
pch
->
iir_state
);
v
=
iir_filter
(
la
[
i
*
128
+
j
],
pch
->
iir_state
);
sum
+=
v
*
v
;
}
s
[
i
]
=
sum
;
...
...
@@ -794,18 +794,17 @@ static FFPsyWindowInfo psy_lame_window(FFPsyContext *ctx, const float *audio,
float
attack_intensity
[(
AAC_NUM_BLOCKS_SHORT
+
1
)
*
PSY_LAME_NUM_SUBBLOCKS
];
float
energy_subshort
[(
AAC_NUM_BLOCKS_SHORT
+
1
)
*
PSY_LAME_NUM_SUBBLOCKS
];
float
energy_short
[
AAC_NUM_BLOCKS_SHORT
+
1
]
=
{
0
};
int
chans
=
ctx
->
avctx
->
channels
;
const
float
*
firbuf
=
la
+
(
AAC_BLOCK_SIZE_SHORT
/
4
-
PSY_LAME_FIR_LEN
)
*
chans
;
const
float
*
firbuf
=
la
+
(
AAC_BLOCK_SIZE_SHORT
/
4
-
PSY_LAME_FIR_LEN
);
int
j
,
att_sum
=
0
;
/* LAME comment: apply high pass filter of fs/4 */
for
(
i
=
0
;
i
<
AAC_BLOCK_SIZE_LONG
;
i
++
)
{
float
sum1
,
sum2
;
sum1
=
firbuf
[
(
i
+
((
PSY_LAME_FIR_LEN
-
1
)
/
2
))
*
chans
];
sum1
=
firbuf
[
i
+
(
PSY_LAME_FIR_LEN
-
1
)
/
2
];
sum2
=
0
.
0
;
for
(
j
=
0
;
j
<
((
PSY_LAME_FIR_LEN
-
1
)
/
2
)
-
1
;
j
+=
2
)
{
sum1
+=
psy_fir_coeffs
[
j
]
*
(
firbuf
[
(
i
+
j
)
*
chans
]
+
firbuf
[(
i
+
PSY_LAME_FIR_LEN
-
j
)
*
chans
]);
sum2
+=
psy_fir_coeffs
[
j
+
1
]
*
(
firbuf
[
(
i
+
j
+
1
)
*
chans
]
+
firbuf
[(
i
+
PSY_LAME_FIR_LEN
-
j
-
1
)
*
chans
]);
sum1
+=
psy_fir_coeffs
[
j
]
*
(
firbuf
[
i
+
j
]
+
firbuf
[
i
+
PSY_LAME_FIR_LEN
-
j
]);
sum2
+=
psy_fir_coeffs
[
j
+
1
]
*
(
firbuf
[
i
+
j
+
1
]
+
firbuf
[
i
+
PSY_LAME_FIR_LEN
-
j
-
1
]);
}
/* NOTE: The LAME psymodel expects it's input in the range -32768 to 32768. Tuning this for normalized floats would be difficult. */
hpfsmpl
[
i
]
=
(
sum1
+
sum2
)
*
32768
.
0
f
;
...
...
libavcodec/psymodel.c
View file @
9b8e2a87
...
...
@@ -112,19 +112,15 @@ av_cold struct FFPsyPreprocessContext* ff_psy_preprocess_init(AVCodecContext *av
return
ctx
;
}
void
ff_psy_preprocess
(
struct
FFPsyPreprocessContext
*
ctx
,
const
float
*
audio
,
float
*
dest
,
int
tag
,
int
channels
)
void
ff_psy_preprocess
(
struct
FFPsyPreprocessContext
*
ctx
,
float
**
audio
,
int
channels
)
{
int
ch
,
i
;
int
ch
;
int
frame_size
=
ctx
->
avctx
->
frame_size
;
if
(
ctx
->
fstate
)
{
for
(
ch
=
0
;
ch
<
channels
;
ch
++
)
ff_iir_filter_flt
(
ctx
->
fcoeffs
,
ctx
->
fstate
[
tag
+
ch
],
ctx
->
avctx
->
frame_size
,
audio
+
ch
,
ctx
->
avctx
->
channels
,
dest
+
ch
,
ctx
->
avctx
->
channels
);
}
else
{
for
(
ch
=
0
;
ch
<
channels
;
ch
++
)
for
(
i
=
0
;
i
<
ctx
->
avctx
->
frame_size
;
i
++
)
dest
[
i
*
ctx
->
avctx
->
channels
+
ch
]
=
audio
[
i
*
ctx
->
avctx
->
channels
+
ch
];
ff_iir_filter_flt
(
ctx
->
fcoeffs
,
ctx
->
fstate
[
ch
],
frame_size
,
&
audio
[
ch
][
frame_size
],
1
,
&
audio
[
ch
][
frame_size
],
1
);
}
}
...
...
libavcodec/psymodel.h
View file @
9b8e2a87
...
...
@@ -174,13 +174,10 @@ av_cold struct FFPsyPreprocessContext* ff_psy_preprocess_init(AVCodecContext *av
* Preprocess several channel in audio frame in order to compress it better.
*
* @param ctx preprocessing context
* @param audio samples to preprocess
* @param dest place to put filtered samples
* @param tag channel number
* @param channels number of channel to preprocess (some additional work may be done on stereo pair)
* @param audio samples to be filtered (in place)
* @param channels number of channel to preprocess
*/
void
ff_psy_preprocess
(
struct
FFPsyPreprocessContext
*
ctx
,
const
float
*
audio
,
float
*
dest
,
int
tag
,
int
channels
);
void
ff_psy_preprocess
(
struct
FFPsyPreprocessContext
*
ctx
,
float
**
audio
,
int
channels
);
/**
* Cleanup audio preprocessing module.
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment