With 120 ms frames and 6 or more channels, the total number of
 samples in the buffer could overflow.

This causes warnings with the ALLOC() macro.

The original code was causing parts of a sine sweep to be completely dropped
due to the 300 Hz margin and the hysteresis. Also, fixes scaling for the
analysis downmix when the input is 16-bit PCM.

We weren't doing the prefilling for SILK->SILK transitions.

We keep accidentally adding these which break the -pedantic build,
so complain about it in the normal build.

Also de-duplicate the warning list.

Koen Vos authored 12 years ago and Jean-Marc Valin committed 12 years ago

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

Taihei Momma authored 12 years ago and Jean-Marc Valin committed 12 years ago

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

Robert Meakins authored 12 years ago and Jean-Marc Valin committed 12 years ago

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

Using this to change the encoder based on mono/stereo rather than looking
at the number of input channels.

Conflicts:
	celt/celt_encoder.c

This not only avoids sudden changes in audio quality, but also increases
quality when we don't have enough bits to code a wide stereo image.

Improved version of:
https://github.com/berkus/mettanode/commit/f44b28ab56ce8afa4b9d8c7460b184197519281c

James Zern authored 12 years ago and Gregory Maxwell committed 12 years ago

Quiets:
warning: this decimal constant is unsigned only in ISO C90

when building with e.g., gcc -m32 -std=gnu90

opus_decode() and opus_multistream_decode() now apply soft clipping
before converting to 16-bit int. This should produce better a higher
quality result than hard clipping like we were doing before. The _float()
API isn't affected, but the clipping function is exported so users can
manually apply the soft clipping.

This is done using an adaptive beta and an estimate of the speech
and music detection confidence

Avoids biasing the decision when it's all silence/noise.

Variable duration option renamed to OPUS_SET_EXPERT_FRAME_DURATION,
with new API.
Also moves up the analysis to avoid having to do int->float conversion
on large buffers.

1) In cases where the SILK desired bandwidth went down, then quickly up, we
count get stuck in a mode with the LP variation going the wrong way.
2) Bandwidth detection can no longer force SILK to go below wideband to
avoid switches that require redundancy.

Detects a transient when the average band energy jumps by more than 6 dB
between two consecutive frames. This fixes some undetected transients on
Gainless' Muse_Breaks_Rmx sample.

Now supports OPUS_AUTO and OPUS_BITRATE_MAX

Also fixes a bug with stereo streams where the initial memory was only
using the left channel.

The search is based on minimizing the bitrate increase due to transients
by considering we can reduce the "transient boost" by reducing the frame
size, at the cost of increasing the normal overhead.

Also, the encode+decode mode now produces an output of the same size
as the original.

This means 2.5 ms frames can now use a higher bitrate for transients.

By not reinitializing the stereo state during a switch, the old stereo
memory would create a false impulse (i.e. single sample) in the decoded audio.
This change affects the normative part of the decoder. Fortunately,
the modified decoder is still compliant with the specification because
it still easily passes the testvectors. For example, for the float decoder
at 48 kHz, the opus_compare (arbitrary) "quality score" changes from
from 99.9333% to 99.925%.

This improves quality of CELT->SILK switches by not having the SILK encoder
attempt to encode a discontinuity. Overall, it's a small PQ improvement,
confirmed by some listening.