shithub: opus

RSS feed

3fed34ae – Jean-Marc Valin <[email protected]> authored on 2010/12/17 09:17
Giving less bits to single-bin bands.

f33a7fb8 – Jean-Marc Valin <[email protected]> authored on 2010/12/17 08:38
Fixed the spreading probabilities (1-x)

320cf2e2 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/17 00:52
Re-organize spreading/folding constants.

cd84e3d0 – Jean-Marc Valin <[email protected]> authored on 2010/12/16 17:29
Re-enabling post-filter on 2.5 ms frames

428a77d6 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/16 11:50
More cleanups to compute_allocation().

76ea41e1 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/16 09:39
Give the bit we reserved to end skipping back when we don't use it.

8e31ab3d – Jean-Marc Valin <[email protected]> authored on 2010/12/16 11:45
Making lower sampling rates use the same allocation table

8e447678 – Gregory Maxwell <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/16 09:22
Terminate the coding of skip bits at the last dynalloc boosted band. Otherwise the bitstream allows non-sensible behavior by the encoder (dynallocing bits into a band and then skipping it). This reduces skip bit overhead by about 2-3% at moderate bitrates with the current encoder.

034172ba – Jean-Marc Valin <[email protected]> authored on 2010/12/16 09:40
Making the encoder fail less often at stupidly low bit-rates

5c80391b – Jean-Marc Valin <[email protected]> authored on 2010/12/16 09:11
Comments, low bit-rate busting avoidance

df6620eb – Jean-Marc Valin <[email protected]> authored on 2010/12/16 08:07
Setting fine_priority for skipped bands

66c5ab4f – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/16 03:39
Rebalance N=1 allocations during interp_bits2pulses().

405e6a99 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/15 15:46
Change strategies for allocation hole prevention.

7627b9f6 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/15 03:22
More compute_allocation() fixes.

4777f069 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/15 01:56
Store the total budget of compute_allocation in BITRES units.

283a9b60 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/15 00:35
Don't stop force-skipping on the first manually skipped band.

b2f59009 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/15 00:12
Move skip coding into interp_bits2pulses().

7cbf168c – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/12/14 16:55
More allocation function updates.

ffe10574 – Jean-Marc Valin <[email protected]> authored on 2010/12/14 19:36
Tim just rewrote half of the bit allocator -- hope it works now

9651ffdb – Jean-Marc Valin <[email protected]> authored on 2010/12/14 11:41
This fixes a bunch of bit allocation bugs

fa50d3ac – Jean-Marc Valin <[email protected]> authored on 2010/12/14 10:02
More negative bits prevention

dcacb73c – Jean-Marc Valin <[email protected]> authored on 2010/12/14 08:39
Preventing negative bit allocation

6cbfbc38 – Jean-Marc Valin <[email protected]> authored on 2010/12/14 06:53
Tuning for 2.5 ms frames

425cfd3e – Jean-Marc Valin <[email protected]> authored on 2010/12/13 11:56
Fix 2-phase stereo in the encoder

70d30ffc – Jean-Marc Valin <[email protected]> authored on 2010/12/13 08:50
Using overlap=shortMdctSize even for 2.5 ms frames.

546dfa19 – Jean-Marc Valin <[email protected]> authored on 2010/12/10 12:18
Adapting the allocation trim based on the spectral tilt

dfd6e714 – Jean-Marc Valin <[email protected]> authored on 2010/12/09 18:23
Adding some hysteresis on the folding threshold frequency

85bbab0b – Jean-Marc Valin <[email protected]> authored on 2010/12/09 10:23
Fixes a fixed-point overflow in stereo angle calculation

fddc521a – Jean-Marc Valin <[email protected]> authored on 2010/12/09 09:28
Completely new transient analysis algorithm

2b747c98 – Jean-Marc Valin <[email protected]> authored on 2010/12/08 06:06
Simplifying the allocation code and making sure both channels get a bit

52dc66b8 – Jean-Marc Valin <[email protected]> authored on 2010/12/06 16:31
Take into accound the 32-bit limit in the VQ

79b34eb8 – Jean-Marc Valin <[email protected]> authored on 2010/12/05 12:22
Add API support for unconstrained VBR

9faf7408 – Jean-Marc Valin <[email protected]> authored on 2010/12/04 05:27
Keeping the allocation of the intensity-codec bands

a4badac9 – Jean-Marc Valin <[email protected]> authored on 2010/12/03 10:20
Making VBR slightly exceed the budget rather than fail

30165bba – Jean-Marc Valin <[email protected]> authored on 2010/12/03 09:35
Fixing the most obvious problems with the VBR code

e5e9aa79 – Jean-Marc Valin <[email protected]> authored on 2010/12/02 11:09
Fixes some side-information rate control issues in VBR mode

e65978fe – Jean-Marc Valin <[email protected]> authored on 2010/12/02 08:46
Adding a dual stereo option.

100ae8ce – Jean-Marc Valin <[email protected]> authored on 2010/12/01 13:45
nothing to see here

1bfa18cb – Jean-Marc Valin <[email protected]> authored on 2010/12/01 11:11
Fix totally broken bit allocation for non-mainstream modes (e.g. powers of two).

bad42a7d – Jean-Marc Valin <[email protected]> authored on 2010/12/01 09:54
Better tuning for 5 ms frames

4b087df5 – Jean-Marc Valin <[email protected]> authored on 2010/11/30 16:08
Increasing resolution of the alloc trim

dd4bf63a – Jean-Marc Valin <[email protected]> authored on 2010/11/30 10:31
Tuning higher bit-rate to behave a bit more like Vorbis.

4f177e85 – Jean-Marc Valin <[email protected]> authored on 2010/11/26 05:32
Intensity stereo now in the bit-stream

fb031119 – Jean-Marc Valin <[email protected]> authored on 2010/11/25 11:32
Allocate remaining bits with constant SNR offset

9b4ed5e2 – Jean-Marc Valin <[email protected]> authored on 2010/11/24 13:01
Implementing intensity stereo.

128d0c1c – Jean-Marc Valin <[email protected]> authored on 2010/11/24 12:43
Tuning the allocation trim_offset for shorter frames

e6d7c084 – Jean-Marc Valin <[email protected]> authored on 2010/11/19 09:43
Minor tuning to the last band

c992a0d7 – Jean-Marc Valin <[email protected]> authored on 2010/11/19 09:06
New bit allocation tuning now enabled by default

54d84c01 – Jean-Marc Valin <[email protected]> authored on 2010/11/19 06:45
Bit allocation wotk.

0722c16d – Jean-Marc Valin <[email protected]> authored on 2010/11/16 05:55
Good catch by John Ridges

ef2e6505 – Timothy B. Terriberry <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/11/08 20:43
Add coarse energy entropy model tuning.

99a19474 – Jean-Marc Valin <[email protected]> authored on 2010/11/08 19:13
0.9.1 release

1ad93cf4 – Jean-Marc Valin <[email protected]> authored on 2010/11/06 18:02
Fixes several fixed-point overflows in the PLC code

e53c4bc5 – Jean-Marc Valin <[email protected]> authored on 2010/11/06 17:41
Fixes a silly fixed-point scaling PLC bug

d7231dd1 – Jean-Marc Valin <[email protected]> authored on 2010/11/06 16:30
Giving up on reusing the saved overlap in the PLC

bc4a0023 – Jean-Marc Valin <[email protected]> authored on 2010/11/06 14:11
PLC fixes

8f6d12be – Jean-Marc Valin <[email protected]> authored on 2010/11/06 08:11
Adding float|fixed suffixes to static mode files

6c12497c – Jean-Marc Valin <[email protected]> authored on 2010/11/05 10:55
Increases the probability of alloc_trim==2 to reflect the latest changes

44a96007 – Jean-Marc Valin <[email protected]> authored on 2010/11/05 07:39
Minor tuning

9743bf38 – Gregory Maxwell <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/11/04 16:45
Switch iteration over channels to the do{}while(); construct in order to inform the compiler that the these loops execute at least once. (This results in more intelligent output from the clang static analysis tool and should also produce faster code on at least some architectures.)

60c316b4 – Gregory Maxwell <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/11/04 16:14
Eliminate some promotions to double. A fair number of implicit promotions remain but they all involve math functions which exist only as double precision form in C89.

fa1ca199 – Jean-Marc Valin <[email protected]> authored on 2010/11/04 11:59
Make sure that we always code the sign of N=1 bands

a3a066cb – Jean-Marc Valin <[email protected]> authored on 2010/11/04 11:15
Fixes some stereo issues where the right channel wasn't taken into account

35095c69 – Jean-Marc Valin <[email protected]> authored on 2010/11/04 09:24
Squashed commit of the following:

bc2c4548 – Jean-Marc Valin <[email protected]> authored on 2010/10/28 10:27
Fixed the PLC, which had been broken for a while

fac6c98c – Gregory Maxwell <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/10/28 11:00
Fix crashes with VBR for short duration frames and very low bitrates. This may, however, cause the encoder to violate the rate target at insanely low rates.

a9411472 – Gregory Maxwell <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/10/27 23:52
Switch example tools to use VBR and 960 sample frames by default on the basis that if the user doesn't have any particular requirements that they probably want this.

eedb4228 – Jean-Marc Valin <[email protected]> authored on 2010/10/23 20:22
Further simplications to compute_mdcts()

933dd833 – Jean-Marc Valin <[email protected]> authored on 2010/10/23 20:08
De-interleaves the MDCT input and overlap memory.

9037757c – Jean-Marc Valin <[email protected]> authored on 2010/10/22 11:12
Tuning the allocation trim

c40addcb – Jean-Marc Valin <[email protected]> authored on 2010/10/22 10:57
Reworked the allocation trim to be absolute (in bits/sample) rather relative

fd54a99e – Jean-Marc Valin <[email protected]> authored on 2010/10/21 13:50
Prevents a folding artefact due to "folding the folding"

94491653 – Jean-Marc Valin <[email protected]> authored on 2010/10/21 13:10
Alternate tuning found using exp_tuning_knobs (disabled by default)

6bf15c55 – Jean-Marc Valin <[email protected]> authored on 2010/10/21 11:08
Tuning the folding bands.

0110301d – Jean-Marc Valin <[email protected]> authored on 2010/10/19 12:40
allocation trim doesn't make sense for stereo after all

c5792dee – Jean-Marc Valin <[email protected]> authored on 2010/10/19 10:24
First shot at automatically adjusting the "allocation trim" for stereo.

5790fba7 – Jean-Marc Valin <[email protected]> authored on 2010/10/18 13:27
Simplifying transient_analysis() now that we don't care about the time window

7fff572a – Jean-Marc Valin <[email protected]> authored on 2010/10/18 12:20
Making sure we can use up to 128 pulses.

d50b6287 – Jean-Marc Valin <[email protected]> authored on 2010/10/18 11:26
Removing renormalise_bands()

27fc0c42 – Jean-Marc Valin <[email protected]> authored on 2010/10/18 11:10
s/const/static const/

7a08ddd1 – Jean-Marc Valin <[email protected]> authored on 2010/10/18 10:55
Removing both the transient window and the mdct_weight_shift.

35fceef3 – Jean-Marc Valin <[email protected]> authored on 2010/10/17 20:34
Turning off time-domain window pending decision on what to do with it.

e0aa9d18 – Jean-Marc Valin <[email protected]> authored on 2010/10/17 12:25
Removing dead code found by LLVM's static analysis

e0ae9801 – John Ridges <[email protected]> authored and Jean-Marc Valin <[email protected]> committed on 2010/10/16 14:00
Optimised float->int cast for Win64

bb918ef7 – Jean-Marc Valin <[email protected]> authored on 2010/10/15 20:01
Fixed fixed-point overflow for 2.5 ms frames.

3295b5d5 – Jean-Marc Valin <[email protected]> authored on 2010/10/15 19:43
Folding code moved to quant_band() to prevent duplication.

c4711e4e – Jean-Marc Valin <[email protected]> authored on 2010/10/15 19:31
More VQ search simplifications

bbc010c4 – Jean-Marc Valin <[email protected]> authored on 2010/10/15 18:55
Removing yshift from the fixed-point VQ search

3d25e71b – Jean-Marc Valin <[email protected]> authored on 2010/10/15 18:45
More simplifications to the VQ search

aca7a630 – Jean-Marc Valin <[email protected]> authored on 2010/10/15 18:06
VQ search no longer needs to put more than one pulse at a time

dfb3e686 – Jean-Marc Valin <[email protected]> authored on 2010/10/15 13:45
tweaks to the split delta for transients

4d2d9fc9 – Jean-Marc Valin <[email protected]> authored on 2010/10/15 10:17
Transient detection fix for the case where a transient occurs during the overlap

54fb7e56 – Jean-Marc Valin <[email protected]> authored on 2010/10/15 07:26
Tuning the bias for the L1 tf metric

8852563a – Jean-Marc Valin <[email protected]> authored on 2010/10/15 06:24
Unb0rked the VBR code: several units issues that made VBR completely useless

8200b2d1 – Jean-Marc Valin <[email protected]> authored on 2010/10/14 22:18
Modifying the L1 metric for tf analysis by merging together all frequencies.

4a8c1f11 – Jean-Marc Valin <[email protected]> authored on 2010/10/13 14:03
Making it possible to have two transients in a row for longer frames

2794b630 – Jean-Marc Valin <[email protected]> authored on 2010/10/13 13:32
Adding a high-pass filter in front of the transient analysis code

ccd51fe9 – Jean-Marc Valin <[email protected]> authored on 2010/10/13 07:48
No longer filling the stereo side with noise.

ccd5a61a – Jean-Marc Valin <[email protected]> authored on 2010/10/12 21:11
Taking into account TF decisions for VBR

0f8fc0be – Jean-Marc Valin <[email protected]> authored on 2010/10/12 19:25
New TF decision code based on L1-norm. Needs more work.

Next