This patch makes all symbols conditional on whether or not there's
enough space left in the buffer to code them, and eliminates much
of the redundancy in the side information.
A summary of the major changes:
* The isTransient flag is moved up to before the the coarse energy.
If there are not enough bits to code the coarse energy, the flag
would get forced to 0, meaning what energy values were coded
would get interpreted incorrectly.
This might not be the end of the world, and I'd be willing to
move it back given a compelling argument.
* Coarse energy switches coding schemes when there are less than 15
bits left in the packet:
- With at least 2 bits remaining, the change in energy is forced
to the range [-1...1] and coded with 1 bit (for 0) or 2 bits
(for +/-1).
- With only 1 bit remaining, the change in energy is forced to
the range [-1...0] and coded with one bit.
- If there is less than 1 bit remaining, the change in energy is
forced to -1.
This effectively low-passes bands whose energy is consistently
starved; this might be undesirable, but letting the default be
zero is unstable, which is worse.
* The tf_select flag gets moved back after the per-band tf_res
flags again, and is now skipped entirely when none of the
tf_res flags are set, and the default value is the same for
either alternative.
* dynalloc boosting is now limited so that it stops once it's given
a band all the remaining bits in the frame, or when it hits the
"stupid cap" of (64<<LM)*(C<<BITRES) used during allocation.
* If dynalloc boosing has allocated all the remaining bits in the
frame, the alloc trim parameter does not get encoded (it would
have no effect).
* The intensity stereo offset is now limited to the range
[start...codedBands], and thus doesn't get coded until after
all of the skip decisions.
Some space is reserved for it up front, and gradually given back
as each band is skipped.
* The dual stereo flag is coded only if intensity>start, since
otherwise it has no effect.
It is now coded after the intensity flag.
* The space reserved for the final skip flag, the intensity stereo
offset, and the dual stereo flag is now redistributed to all
bands equally if it is unused.
Before, the skip flag's bit was given to the band that stopped
skipping without it (usually a dynalloc boosted band).
In order to enable simple interaction between VBR and these
packet-size enforced limits, many of which are encountered before
VBR is run, the maximum packet size VBR will allow is computed at
the beginning of the encoding function, and the buffer reduced to
that size immediately.
Later, when it is time to make the VBR decision, the minimum packet
size is set high enough to ensure that no decision made thus far
will have been affected by the packet size.
As long as this is smaller than the up-front maximum, all of the
encoder's decisions will remain in-sync with the decoder.
If it is larger than the up-front maximum, the packet size is kept
at that maximum, also ensuring sync.
The minimum used now is slightly larger than it used to be, because
it also includes the bits added for dynalloc boosting.
Such boosting is shut off by the encoder at low rates, and so
should not cause any serious issues at the rates where we would
actually run out of room before compute_allocation().
cf874373 raised the limit from 7 to 8 for N>1 bands in
interp_bits2pulses(), but did not raise the corresponding limits
for N=1 bands, or for [un]quant_energy_finalise().
This commit raises all of the limits to the same value, 8.
This way if a band doesn't get the fine bits we want because it
wasn't allocated enough bits to start with, then we will still
give it priority for any spare bits after PVQ.
The modeline-bisection and interpolator have used different criteria
for the minimum coding threshold since the introduction of the
"backwards done" in 405e6a99. This meant that a lower modeline could be
selected which the interpolator was never able to get under the maximum
allocation. This patch makes the modeline selection search use the same
criteria as the interpolator.
All of our usage of ec_{enc|dec}_bit_prob had the probability of a
"one" being a power of two.
This adds a new ec_{enc|dec}_bit_logp() function that takes this
explicitly into account.
It introduces less rounding error than the bit_prob version, does not
require 17-bit integers to be emulated by ec_{encode|decode}_bin(),
and does not require any multiplies or divisions at all.
It is exactly equivalent to
ec_encode_bin(enc,_val?0:(1<<_logp)-1,(1<<_logp)-(_val?1:0),1<<_logp)
The old ec_{enc|dec}_bit_prob functions are left in place for now,
because I am not sure if SILK is still using them or not when
combined in Opus.
The bisection search in compute_allocation() was not using the same
method to count psum as interp_bits2pulses, i.e., it did not
include the 64*C<<BITRES<<LM allocation ceiling (this adds at most
84 max operations/frame, and so should have a trivial CPU cost).
Again, I wouldn't want to try to explain why these are different in
a spec, so let's make them the same.
In addition, the procedure used to fill in bits1 and bits2 after the
bisection search was not the same as the one used during the
bisection search.
I.e., the
if (bits1[j] > 0)
bits1[j] += trim_offset[j];
step was not also done for bits2, so bits1[j] + bits2[j] would not
be equal to what was computed earlier for the hi line, and would
not be guaranteed to be larger than total.
We now compute both allocation lines in the same manner, and then
obtain bits2 by subtracting them, instead of trying to compute the
offset from bits1 up front.
Finally, there was nothing to stop a bitstream from boosting a band
beyond the number of bits remaining, which means that bits1 would
not produce an allocation less than or equal to total, which means
that some bands would receive a negative allocation in the decoder
when the "left over" negative bits were redistributed to other
bands.
This patch only adds the dynalloc offset to allocation lines greater
than 0, so that an all-zeros floor still exists; the effect is that
a dynalloc boost gets linearly scaled between allocation lines 0 and
1, and is constant (like it was before) after that.
We don't have to add the extra condition to the bisection search,
because it never examines allocation line 0.
This re-writes the indexing in the search to make that explicit;
it was tested and gives exactly the same results in exactly the
same number of iterations as the old search.
Commit 8e447678 increased the number of cases where we end skipping
without explicit signaling.
Before, this would cause the bit we reserved for this purpose to
either a) get grabbed by some N=1 band to code its sign bits or
b) wind up as part of the fine energy at the end.
This patch gives it back to the band where we stopped skipping,
which is either the first band, or a band that was boosted by
dynalloc.
This allows the bit to be used for shape coding in that band, and
allows the better computation of the fine offset, since the band
knows it will get that bit in advance.
With this change, we now guarantee that the number of bits allocated
by compute_allocation() is exactly equal to the input total, less
the bits consumed by skip flags during allocation itself (assuming
total was non-negative; for negative total, no bits are emitted,
and no bits are allocated).
Excess fractions of a bit can't be re-used in N=1 bands during
quant_all_bands() because there's no shape, only a sign bit.
This meant that all the fractional bits in these bands accumulated,
often up to 5 or 6 bits for stereo, until the first band with N>1,
where they were dumped all at once.
This patch moves the rebalancing for N=1 bands to
interp_bits2pulses() instead, where excess bits still have a
chance to be moved into fine energy.
In commit ffe10574 JM added a "done" flag to the allocation
interpolation loop: whenver a band did not have enough bits to
pass its threshold for receiving PVQ pulses, all of the rest of
band were given just enough bits for fine energy only.
This patch implements JM's "backwards done" idea: instead work
backwards, dropping bands until the first band that is over the
threshold is encountered, and don't artificially reduce the
allocation any more after that.
This is much more stable: we can continue to signal manual skips if
we want to, but we aren't forced to skip a large number of bands
because of an isolated hole in he allocation.
This makes low-bitrate 120-sample frames much less rough.
It also reduces the force skip threshold from
alloc_floor+(1<<BITRES)+1 to just alloc_floor+(1<<BITRES), because
the former can now cascade to cause many bands to be skipped.
The difference here is subtle, and increases signaling overhead by
0.11% of the total bitrate, but Monty confirmed that removing the
+1 reduces noise in the bass (i.e., in N=1 bands where such a skip
could cascade).
Finally the 64*C<<BITRES<<LM ceiling is moved into the bisection
search, instead of just being imposed afterwards, again because I
wouldn't want to try to explain in a spec why they're different.
1) Continue to update left and percoeff if we skip all the way to the
first band.
This doesn't actually matter for correctness, but I don't want to
try to explain in a spec why we aren't doing this.
2) Force all the bits in skipped bands to go to fine energy.
Before some of them could continue to be given to pulses, even though no
pulses would actually be allocate for them.
The margin of safety was supposed to be 1/8th bit, not 1 bit, and the
bit we reserved to terminate skip signalling before was actually 8
bits.
This patch updates the margin of safety to the correct value and
accounts for the one bit (not 8) needed for skip signalling.
It also fixes the remainder calculation in the skip loop to work
correctly when start>0.
Now that manual skipping is in the same loop as forced skipping, there
is no reason to do all of one, then all of the other.
This ensures we won't propagate bits to bands that have almost nothing
later in quant_all_bands() because we didn't have enough bits to
signal them skipped.
This allows us to a) not pay a coding cost to avoid skipping bands that are
stupid to skip (e.g., the first band, or bands that have so few bits that we
wouldn't redistribute anything) and b) not reserve bits to pay that cost.
This moves more of the decisions about when to stop skipping bands into the
encoder-specific branch, so they are not forced in the decoder (because there
is currently no bit-savings from forcing them).
It also no longer requires an extra bit to code the fine energy in a skipped
band: that was meant to account for the skip flag, but we already subtracted
that.
Mimimum resolution is now 1/4 bit/sample instead of 3/8. The allocation
tilt now gives more LF bits for small frame sizes. There's also a new
"alternate" allocation that should soon replace the current one.
commit a2cc77cb2744a2cb0551b9bfdf06b97457b6d449
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Thu Nov 4 13:11:21 2010 -0400
Adding a switch to enable the post-filter (off by default)
commit 8e860dc0dfbe57e59fcbd5352588c5edff020e27
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Thu Nov 4 11:57:12 2010 -0400
Allowing pitches up to 3000 Hz
commit 837412d37bbca32bb34bfb5941e132ff4b0a568c
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Wed Nov 3 20:47:11 2010 -0400
Pitch estimation tuning to prevent some cases of pitch halving
commit 34e20f24c85b40fffd1a15c5b632f2f78b26f081
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Nov 3 16:31:51 2010 -0400
Resynthesis now purely a compile-time option with RESYNTH
commit d83fb5a9cc2ec4b6cce938662997643da1c5ed0d
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Nov 3 16:28:25 2010 -0400
Fixes a divide by zero in remove_doubling()
commit bb91e05b7f8f91fd15a8a0daae3d8cb6bd8d81db
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Nov 3 15:55:48 2010 -0400
Bring back resynthesis with RESYNTH macro
commit 31fe6f6b4997af0a46b8c62f523fe2dfdb7f56ae
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Tue Nov 2 17:55:04 2010 -0400
Tuning the allocation tilt to give more bits to higher frequencies.
Especially useful now that the post-filter can reduce low freq noise.
commit 919ba48f0369a87885334756cdfac2a448ce52d0
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Mon Nov 1 17:27:19 2010 -0400
C89 fix
commit ee0dbb1855a82ee8c132ddaffcab4d072bb3455e
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Mon Nov 1 11:45:10 2010 -0400
Complete fixed-point port of the pitch code (I think).
commit 4c7b3fd12a8f7469607b5ac57c85301a5de9fa81
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Mon Nov 1 10:55:43 2010 -0400
More fixed-point pitch gain work
commit 26f1412188900199b63e187fcb0bd04db53c898a
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Mon Nov 1 10:39:25 2010 -0400
Fixed-point version of the pitch gain calculation code
commit 27c73d008e9f50d282c3ad08e2f05f7006013ae1
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Sun Oct 31 16:50:26 2010 -0400
Some more fixed-point work in remove_doubling()
commit 59354672cb3af794a0e46c0b2097d6441c75cdd1
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Sun Oct 31 09:57:36 2010 -0400
Fixed a stupid fixed-point pf bug in the gain handling
commit be9e7dabf6c8b32bc049da260b58ff6085dc1ac3
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Sat Oct 30 01:52:09 2010 -0400
Fixed-point: fixed frac_div32() that was broken a few commits ago.
commit 5b06270afc41a88915252cea14411be43650e704
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Fri Oct 29 17:45:44 2010 -0400
This fixes VBR when encoding the pitch period with raw bits
commit 10e0488458ae558aa80d0b30cce70841ad081f73
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Fri Oct 29 16:50:31 2010 -0400
Pitch period is now encoder with equal probability for each octave (rather than each lag).
Max pitch gain allowed is now 0.625.
commit ca19396c1c1511c0e208b400efb51384fc7c200d
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Fri Oct 29 16:00:01 2010 -0400
More fixed-point post-filter work
commit f3e42fde1b575bc587b2557b8b31a6085421a99c
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Fri Oct 29 14:39:23 2010 -0400
More fixed-point work for the prefilter/postfilter
commit db945132d12b25ff25acc0701b91a1d8a81417d5
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Fri Oct 29 14:14:02 2010 -0400
Making the pitch estimation work in fixed-point
Even if there's still lots of float operations left.
commit acb3f96e04802ac4601295f83bef1f32593e261a
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Fri Oct 29 10:57:39 2010 -0400
Making the PLC code consistent with the prefilter/postfilter
commit 8f64f5974ac846b8c35d0b692e0472f279206cf0
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Thu Oct 28 00:33:53 2010 -0400
More tuning for remove_doubling()
commit 0c08f2ee9dcc135dd222fef30f5ad93e95e0d364
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Oct 27 17:48:02 2010 -0400
Doing an interpolation step to improve the accuracy of the pitch estimate
Also increasing the gain slightly.
commit 23d303e992f1fdc3d2668652603ae6311d3b91c5
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Oct 27 16:56:42 2010 -0400
Implements a fixed 3-tap prefilter/postfilter to make the gain roll off with frequency
commit 881c5928adc1af9eb75c4b68e9eba94ab1d65adc
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Oct 27 14:47:30 2010 -0400
Partially whitening the down-sampled signal before the pitch search
commit 4a8687deea8587007f14051cb966f6fd748893a1
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Oct 27 14:27:47 2010 -0400
pitch_search() no longer computes the gain
commit a7f85bb6b10d9c509caec521ca444efb3f27df05
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Oct 27 14:00:53 2010 -0400
remove_doubling() now works on the down-sampled signal
commit 06cb70e876873f79fed214ebbca35cb4c5057ec8
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Oct 27 11:28:53 2010 -0400
Simplification to the pitch continuity code
commit 5201927c284a424eb8f21f63d358844b3de8c285
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Wed Oct 27 11:04:02 2010 -0400
Some more pitch doubling prevention code
commit 7ef63fbe1f78f79e1923bc42e06fbdf1ec28ffd3
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Wed Oct 27 06:49:28 2010 -0400
Minor fix
commit eb37eaab32e7df074a7ddf0ae4781e57f827c4ad
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Tue Oct 26 18:32:25 2010 -0400
Enforcing some pitch continuity
commit 751ef6edf2ee7721252cedb264bdf9b3f6244a9d
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Tue Oct 26 17:29:47 2010 -0400
Code for preventing pitch doubling/halving
commit c12647ecb55b645005efbeede91880db72936f8d
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Tue Oct 26 00:04:34 2010 -0400
Finally getting perfect reconstruction when pitch changes
Post-filter now delays the filter coefs by the overlap so that the pre-filter
and post-filter are synchronised.
commit f854311d945bb375039a4a4a4fea782b648581f8
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Mon Oct 25 14:59:13 2010 -0400
Very simple/inefficient signalling of the prefilter period/gain
commit b4e1215432e3d89a29c998639a6d8b07e28c5a2a
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Mon Oct 25 14:09:17 2010 -0400
using the actual pitch gain
commit e7cd4f07bb073b6955a001e56c0bbf16156f4195
Author: Jean-Marc Valin <jean-marc.valin@octasic.com>
Date: Mon Oct 25 12:16:11 2010 -0400
Adding some pitch prediction though side information still isn't coded
commit 77a03aa27c9b6ed2fe80c27a1196b460ccb5079e
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Mon Oct 25 00:12:48 2010 -0400
prefilter implemented as well
commit a3fd81b6ca213d4a9f8ddfa2883fd0e238d64d04
Author: Jean-Marc Valin <jean-marc.valin@usherbrooke.ca>
Date: Sun Oct 24 01:14:10 2010 -0400
Implementing Raymond Chen's comb filter idea
So far, only the post-filter is there.
Also making use of alloc_trim_analysis() again because the effect of
inter-channel correlation on the bitstream is really in terms of absolute
number of bits/samples.
