RSS feed
e949b720
– Jean-Marc Valin <
[email protected]>
authored
on 2010/05/07 03:48
Fixed pitchEnd wrt to variable frame size. It should (sort of) all work now
ce4dd367
– Jean-Marc Valin <
[email protected]>
authored
on 2010/05/07 03:45
More work on variable frame size (getting rid of FRAMESIZE() )
3b918bac
– Jean-Marc Valin <
[email protected]>
authored
on 2010/05/04 20:02
replaced instances of mode->nbShortMdcts by the dynamic number of MDCTs (M)
f7547a4e
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/28 18:10
Shrinking the MDCT's table by separately rotating by the 1/(8N) factor
65ee67ac
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/26 03:08
Making the band definition the same at all frame sizes.
4fb95687
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/20 19:30
Bands are now (again) a multiple of the number of short MDCTs
d6b79eeb
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/20 13:31
Skipping transient shape renormalisation in the encoder because
281a63ea
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/20 06:43
Bump bitstream version now that we broke everything
8271dc8a
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/19 20:47
Tuning the amount of spreading based on Koen's feedback
9c30de57
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/19 09:32
Improved transient_analysis() by adding one frame of memory.
b8ba70c9
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/18 18:10
Disabling resynthesis when not needed (need to remove folding for this to work)
b8a06ee0
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/18 05:57
Re-introducing the successive spreading rotations, but in a two-step
628c0253
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/16 16:57
Converted a few double-precision constants to single precision
3ff5e4c2
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/14 14:02
Making sure compute_allocation_table() doesn't overrun at lower sampling rates
137f3366
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/14 13:42
Changed compute_allocation_table() so it handles ebands that start and end in the same allocation band. Also fixed a minor C89 issue.
31bec963
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/07 14:30
Fix small mode leak (forgot to free the new logN array).
f400a3c5
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/05 19:58
Caching log2_frac(N, BITRES) in the more data to save some CPU
64209a3e
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/05 05:26
A few minor optimisations (compute_allocation, denormalise_bands,
9a92d61e
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/03 12:30
A bit more tuning on the pseudo-frac-Hadamard. Also Trying to improve
354bf60b
– Jean-Marc Valin <
[email protected]>
authored
on 2010/04/03 05:23
Doing the spreading with a "pseudo-fractional-Hadamard" transform
8974f00d
– Jean-Marc Valin <
[email protected]>
authored
on 2010/03/19 20:41
fixed-point: improving accuracy of the energy prediction
c7b01bad
– Jean-Marc Valin <
[email protected]>
authored
on 2010/03/19 19:17
fixed-point: slightly increasing the decoder accuracy again
dc8a6eee
– Timothy B. Terriberry <
[email protected]>
authored
and Jean-Marc Valin <
[email protected]>
committed
on 2010/03/15 19:49
Update the log2 approximation to accomodate the increased bit precision from bd0610d21b28b76095a49e601df44fb792b96369. This requires another term in the power series to reduce the error to the level of truncation (peak absolute error 0.621785). Also refactor the rounding bias term so that further changes to DB_SHIFT will leave the result unbiased.
70bdf9bf
– Jean-Marc Valin <
[email protected]>
authored
on 2010/03/15 16:50
fixed-point: reducing the mismatch in the folded part
210ccafa
– Jean-Marc Valin <
[email protected]>
authored
on 2010/03/11 17:20
Improves fixed-point precision for quant_coarse_energy
6c79a932
– Jean-Marc Valin <
[email protected]>
authored
on 2010/01/16 18:35
Getting "make dist" to work again with the files that got removed
74128bef
– Jean-Marc Valin <
[email protected]>
authored
on 2010/01/01 04:33
fixed-point: converted PLC filters and LPC parameters
07fed1bf
– Jean-Marc Valin <
[email protected]>
authored
on 2009/12/28 02:59
Adding a safeguard against unstable LPC, so now there's no way (that I can
d69c1cb3
– Jean-Marc Valin <
[email protected]>
authored
on 2009/12/27 19:34
Making new PLC code work in fixed-point even though it's still using float
d5f99308
– Jean-Marc Valin <
[email protected]>
authored
on 2009/12/16 17:42
This fixes a bug in stereo PLC (offset wasn't changed between channels)
5a0fae53
– Jean-Marc Valin <
[email protected]>
authored
on 2009/12/14 16:19
PLC: Added lag windowing and constraint to synthesis energy
aebb6fd6
– Jean-Marc Valin <
[email protected]>
authored
on 2009/12/02 02:04
Some work on the pitch search weighting filter (still disabled)
e465c149
– Jean-Marc Valin <
[email protected]>
authored
on 2009/11/25 19:39
Reorganised the pitch code to simplify the PLC case and fixed a buffer overflow
6d584acd
– Jean-Marc Valin <
[email protected]>
authored
on 2009/11/21 16:57
Renamed MDCT functions to avoid symbol clashes with Vorbis
8c7bb4c9
– Timothy Terriberry <
[email protected]>
authored
and Jean-Marc Valin <
[email protected]>
committed
on 2009/10/31 06:19
Expose the normalized range for reciprocal square roots in fixed-point mode. This allows subsequnt calculations to use the full precision of the result.
630ee44a
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/24 10:29
SMALL_FOOTPRINT preserves the special n=1 cwrs case
45f11105
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/21 20:12
Making sure the VBR controller never busts the number of bytes allowed
25767d1c
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/21 19:24
Better computation of the VBR rate upper bound and reducing the coarse energy
ab4dcc5c
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/21 03:08
Allow coarse energy to take almost all the bits. Also, fixed a some issues with
73bf6332
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/20 03:22
Making the multiplication in normalise_residual() produce a 32-bit result just
a3bba38b
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/20 03:13
This should prevent a rare divide-by-zero in the pitch gain code
bd5d54ad
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/19 20:25
New VBR rate controller that doesn't allow more than one frame worth of excess
328b8bd7
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/19 15:07
fixed a few trivial bugs: exporting celt_strerror(), changed DB_SCALING to a
234969c9
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/17 18:12
Removed the _t from all the celt*_t types to avoid clashing with POSIX
30f7f813
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/17 10:35
Changed all the celt*int*_t types to remove the _t suffix, which is reserved
d35d6775
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/16 19:16
Constrain the pitch gain to prevent potential instabilities.
ece94a04
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/16 03:30
Improved error handling, and implemented celt_strerror()
80ed1476
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/15 17:45
The number of channels is now set when creating the states rather than when
bf2398b0
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/15 03:28
first step for removing the number of channels from the mode
6a978022
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/13 16:35
Adjusting the allocation for stereo directly in the alloc function rather than
a9947c48
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/08 19:10
merged the code for quant_bands() and unquant_bands()
f7a1e165
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/07 02:56
Renamed mix_pitch_and_residual() to normalise_residual(), after minor
f6dc1eba
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/06 16:08
Further simplifications to mix_pitch_and_residual() (which no longer does
8861955d
– Jean-Marc Valin <
[email protected]>
authored
on 2009/10/04 17:35
Doing the mdct shaping in the normalised domain, so that energy is preserved
Next