Kári Tristan Helgason authored Nov 09, 2017 and Jean-Marc Valin committed Nov 16, 2017

This fixes issues when using neon intrinsics on 64-bit
systems.

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

Andrew Allen authored Nov 07, 2017 and Jean-Marc Valin committed Nov 10, 2017

Ensure matrix is properly exported from
projection encoder.

Change-Id: I5ede77e4f4d0dc82074e2230780777af542a416e
Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca>

Larger network, using 8-bit weights

Andrew Allen authored Nov 07, 2017 and Jean-Marc Valin committed Nov 07, 2017

OpusProjection* classes
MixingMatrix class
Projection tests

Change-Id: I98644466abf4ffd36e48bdecad1204d69e1539b9
Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca>

Roman Kalashnikov authored Oct 30, 2017 and Ralph Giles committed Oct 30, 2017

Fixes a warning from PVS-Studio. This was a cut-and-paste error
from the stanza above and had no effect.

Signed-off-by: Ralph Giles <giles@thaumas.net>

Put the colon ':' before the continuation dots like all the
other entries.

Addresses issue https://github.com/xiph/opus/issues/54

Addresses issue https://github.com/xiph/opus/issues/58

Reported in https://github.com/xiph/opus/issues/55

The regression was introduced in 783ad767
by changes to celt_fir() that make it no longer in-place.

The masking term was previously completely broken (even in 1.1). The
bandwidth detection relies mostly on the noise floor and can only use
masking to cut one extra band. The 12-24 kHz energy is now normalized properly
but uses a higher noise floor due to the leakage in the resampler. Bandwidth
detection is still mostly useless at identifying SWB speech (it usually says
FB).

Based on two dense layers with a GRU layer in the middle

SILK was being allocated too few bits for stereo hybrid, often resulting
in forced narrowing of the width. We now allocate more bits to SILK
and reduce the threshold for narrowing. This improves quality enough that
the bitrate threshold for switching to SILK can be increased to 44 kb/s.

Alexander Kochetkov authored Sep 11, 2017 and Ralph Giles committed Sep 11, 2017

Prefix passed using '--with-NE10=PFX' not used as NE10 include path.

Signed-off-by: Alexander Kochetkov <al.kochet@gmail.com>
Signed-off-by: Ralph Giles <giles@thaumas.net>

Make these consistent with the other optimization file
sets which use a suffix to indicate the extension they
use.

Signed-off-by: Jonathan Lennox <jonathan@vidyo.com>

Distinguish source files for the SSE 4.1 instruction
set extension consistently by their filename. This makes
it easier to check the correct flags are being set at
build time.

Signed-off-by: Jonathan Lennox <jonathan@vidyo.com>

Matt Brubeck authored Aug 24, 2017 and Ralph Giles committed Aug 29, 2017

Signed-off-by: Ralph Giles <giles@thaumas.net>

Ray Essick authored Aug 24, 2017 and Felicia Lim committed Aug 24, 2017

some SSE optimizations were using an instruction sequence that required
128-bit alignment, even though this is not always guaranteed. The error
was in handling a 64-bit entity; made the same macro changes that had
been done previously for 32-bit entities.

Signed-off-by: Felicia Lim <flim@google.com>

Pavel P authored Aug 10, 2017 and Ralph Giles committed Aug 11, 2017

Signed-off-by: Ricardo Constantino <wiiaboo@gmail.com>

Linfeng Zhang authored Jul 20, 2017 and Jean-Marc Valin committed Jul 25, 2017

This bug is introduced in commit 85fb8a1d

.

Change-Id: I3822cdfc1308ca62ce03da008260de8f4b04596c

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

|frame_size_enum| in tests/test_opus_encode.cl:117 is flagged as potentially
uninitialized but get_frame_size_enum() will fail anyway if a valid value is
not found.

Migrating opus-codec.org to a static site generator inserted
a redirect into the download path, which the actual testvectors
at https://opus-codec.org/static/testvectors/opus_testvectors.tar.gz.

Add the -L option to the suggested curl invocation so the redirect
is followed to download the actual resource. Without this one
gets the html redirection fallback page.

Andrew Larkin authored Jul 06, 2017 and Felicia Lim committed Jul 06, 2017

Signed-off-by: Felicia Lim <flim@google.com>

The new detector for SWB vs FB is unreliable due to the hack
that was required to get 24 kHz analysis working. We're now err
on the side of FB just to make sure.