Koen Vos authored 9 years ago and Jean-Marc Valin committed 9 years ago

The bug was caused by an improper feedback of the per-frame bitrate, causing
the bitrate to jump up and down from frame to frame, within a packet.
The patch avoids this, and also gives a slight improvement in general for
multi-frame packets, even without FEC.

Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca>

maks authored 11 years ago and Jean-Marc Valin committed 11 years ago

The code was wrong, but did not have any impact on user-visible behaviour
because all it did was change an internal-only error code.

Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca>

Timothy B. Terriberry authored 11 years ago and Jean-Marc Valin committed 11 years ago

Optimizing celt_pitch_xcorr()/xcorr_kernel() which also speeds up
FIRs, IIRs and auto-correlations

Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca>

Works by turning off pitch and energy prediction in CELT, while
setting first_frame_after_reset in SILK to disable pitch and LSF interpolation
and reduce LPC gain.

Since the last patch originally had them mangled (presumably by
 mailer, http server, or something else), let's just get rid of
 them.

This makes all remaining large stack allocations use the vararray
 macros.
This continues the work of 6f2d9f50 to allow compiling with
 NONTHREADSAFE_PSEUDOSTACK to move the memory for large buffers
 off the stack for devices where it is very limited.

It also does this for some additional large buffers used by the
 PLC in the decoder.

Diego Elio Pettenò authored 12 years ago and Jean-Marc Valin committed 12 years ago

This allows the compiler to perform more optimization on them as it
knows how the functions are being called.

Signed-off-by: Diego Elio Pettenò <flameeyes@flameeyes.eu>

Koen Vos authored 13 years ago and Jean-Marc Valin committed 13 years ago

Also clamps the gain to avoid forcing a float decoder to emulate the
state rescaling.

Koen Vos authored 13 years ago and Jean-Marc Valin committed 13 years ago

decoder:
- fixed incorrect scaling of filter states for the smallest quantization
  step sizes
- NLSF2A now limits the prediction gain of LPC filters

encoder:
- increased damping of LTP coefficients in LTP analysis
- increased white noise fraction in noise shaping LPC analysis
- introduced maximum total prediction gain.  Used by Burg's method to
  exit early if prediction gain is exceeded.  This improves packet
  loss robustness and numerical robustness in Burg's method
- Prefiltered signal is now in int32 Q10 domain, from int16 Q0
- Increased max number of iterations in CBR gain control loop from 5 to 6
- Removed useless code from LTP scaling control
- Optimization: smarter LPC loop unrolling
- Switched default win32 compile mode to be floating-point

resampler:
- made resampler have constant delay of 0.75 ms; removed delay
  compensation from silk code.
- removed obsolete table entries (~850 Bytes)
- increased downsampling filter order from 16 to 18/24/36 (depending on
  frequency ratio)
- reoptimized filter coefficients

Koen Vos authored 13 years ago and Jean-Marc Valin committed 13 years ago

Also some comment/warning fixes

Only encoder changes were necessary because this uses the same
"redundant frames" mechanism as SILK<->CELT switching.
This also fixes a regression introduced in 78291b27 that was
causing the encoder to go back and forth between bandwidths when
SILK wasn't ready to change.

This is achieved by running the encoding process in a loop and
padding when we don't reach the exact rate. It also implements
VBR-with-cap, which means we no longer need to artificially decrease
the SILK bandwidth when it's close to the cap.

Correct encoder/decoder state mismatch and spurious state resetting in the encoder when side is reset.

Patch from Tim which corrects a glitch during mode switching tests.

Timothy B. Terriberry authored 13 years ago and Gregory Maxwell committed 13 years ago

If there is activity in a regular, side SILK frame, then it must
 be coded, so we don't need to send a mid-only flag.

Timothy B. Terriberry authored 13 years ago and Jean-Marc Valin committed 13 years ago

Neither the encoder nor decoder were incrementing the side-channel
 index for a mid-only frame.
Unfortunately, they used that index to index the VAD flags and LBRR
 flags.
This made the VAD flags for the side channel particularly useless,
 as you couldn't tell which frame a flag belonged to without
 decoding most of the packet.
It also desynched the LBRR information, as look at the wrong LBRR
 flags to decide when it had to code a mid-only flag.
If some frames were skipped in the side channel, then the last few
 VAD flags and LBRR flags would be garbage, but still get encoded.

This patch fixes this by continuing to increment nFramesDecoded or
 nFramesEncoded, even when skipping a frame in the side channel.
This makes the side-channel VAD and LBRR flags correspond to the
 correct time periods for frames greater than 20 ms.
It also fixes a bug where if DTX was not used on the packet where
 the side channel got turned off, it would never get used again
 until the encoder attempted to encode something in the side
 channel again.

Timothy B. Terriberry authored 13 years ago and Jean-Marc Valin committed 13 years ago

b24e5746 introduced changes to LastGainIndex which broke
 conditional coding for side frames after a mid-only frame (i.e.,
 in a 60 ms frame where the side is coded, not coded, then coded
 again).
These rules were a mess in general, however, because the side
 channel state kept a different nFramesDecoded count from the mid
 channel state, and had no way to tell if the prior side frame was
 coded.

This patch attempts to rationalize them by moving the conditional
 coding decision up to the top level, where all this information is
 available.
The first coded side frame after an uncoded side frame now always
 uses independent coding.
If such a frame is also not the first side frame in an Opus frame,
 then it doesn't include an LTP scaling parameter (because the LTP
 state is well-defined).

- There was a bug where the decoder resampler was not properly initialized
when fs_kHz == API_fs_kHz. In that case the resampler would continue to
upsample, and the output was corrupt.

- The delay value in the decoder was taken from the state before it was
potentially updated. This caused the decoder to apply the new dalay value one
frame late

- The encoder and decoder states are now updated more consistently, when
the sampling rate changes (pesq liked these changes)

- Properly resetting the side channel encoder and decoder for the first
frame with side coding active again

- Faster updating the "ratio" value in the LR_to_MS() code for large
prediction values means that for certain extreme/artificial input
signals the output looks better

Koen Vos authored 13 years ago and Jean-Marc Valin committed 13 years ago

- increases the max pitch lag by 1 (the thing Tim pointed out).  this brings the decoder in sync with the old one
- avoids that the first stereo frame is collapsed to mono

SILK now reports an "effective width" that takes into account side prediction.

1) averaging the output of the left and right resampling states when
   switching to mono
2) averaging the the delay buffers from left and right when switching

Adds SILK delay compensation that depends on encode and decode sampling
rate, as well as SILK internal coding rate. This ensures that the SILK
part of Opus is always in sync with the CELT part no matter what the
sampling rates are. It also increases the resampling delay to 1.15 ms
(was previously 0.48 ms).

Koen Vos authored 13 years ago and Jean-Marc Valin committed 13 years ago

Simplifies mono/stereo switching in SILK
Fixes a quantization mismatch between encoder and decoder
Constrains the pitch lags in the same way in the encoder and decoder

Gets rid of some more glitches caused by the decoder resampler

Delaying stereo->mono switching decisions so that SILK can do a smooth
downmix. Also, wrote proper float/fixed code for the hybrid variable
stereo collapse, including a smooth downmix for stereo<->mono switching

There's still a range coder mismatch on the first frame when using FEC.

For these transitions, we now start the left and right resamplers
from the same state.

The API permits the caller to freely copy the codec state on their
own, but this can't work if there are any any position dependant pointers
in the codec state.