speexdsp.git
3 years agoresample: avoid overflow in phase calculation
Wim Taymans [Thu, 4 Feb 2016 09:49:49 +0000 (10:49 +0100)]
resample: avoid overflow in phase calculation

Make sure we don't overflow when calculating the phase for the new
sample rate.

Signed-off-by: Tristan Matthews <tmatth@videolan.org>
3 years agoresample: Improve GCD calculation
Wim Taymans [Wed, 3 Feb 2016 14:01:35 +0000 (15:01 +0100)]
resample: Improve GCD calculation

Use Euclids algorithm to calculate the greatest common divisor to
simplify the resample ratio fraction instead of the slow iterative
method.

Signed-off-by: Tristan Matthews <tmatth@videolan.org>
4 years agoRemove spurious ACLOCAL_FLAGS variable from autogen.sh.
Ron [Sat, 26 Sep 2015 01:47:54 +0000 (11:17 +0930)]
Remove spurious ACLOCAL_FLAGS variable from autogen.sh.

This isn't needed (and isn't doing anything here anyway).

4 years agoDon't rely on HAVE_STDINT_H et al. being defined
Tanu Kaskinen [Sun, 5 Jul 2015 15:10:19 +0000 (18:10 +0300)]
Don't rely on HAVE_STDINT_H et al. being defined

Not everyone who includes speexdsp_config_types.h will have a test
which defines those, and if we've chosen to use the stdint types at
configure time then we know exactly which header(s) are available, so
just choose the best one then and generate the header to use it.

This patch, including the above text, is copied from a commit in the
speex repository[1]. The original commit for speex was made by Ron
<ron@debian.org>.

[1] https://git.xiph.org/?p=speex.git;a=commitdiff;h=774c87d6cb7dd8dabdd17677fc6da753ecf4aa87

Signed-off-by: Tanu Kaskinen <tanu.kaskinen@linux.intel.com>
4 years agopreprocess: prevent rare overflow on overlap-add
Tristan Matthews [Fri, 24 Apr 2015 19:55:32 +0000 (15:55 -0400)]
preprocess: prevent rare overflow on overlap-add

Reported-by: Fabian Henze <flyser42@gmx.de>
4 years agoremove trailing whitespace
Tristan Matthews [Wed, 1 Apr 2015 16:34:18 +0000 (12:34 -0400)]
remove trailing whitespace

4 years agoconfigure: bump to 1.2rc3 SpeexDSP-1.2rc3
Tristan Matthews [Mon, 15 Dec 2014 06:27:40 +0000 (01:27 -0500)]
configure: bump to 1.2rc3

4 years agocorrect INSTALL file
Tristan Matthews [Sun, 7 Dec 2014 18:18:57 +0000 (13:18 -0500)]
correct INSTALL file

4 years agoensure resample_neon.h is distributed
Tristan Matthews [Sun, 7 Dec 2014 16:34:33 +0000 (11:34 -0500)]
ensure resample_neon.h is distributed

4 years agoremove codec-only references from TODO
Tristan Matthews [Sun, 7 Dec 2014 00:44:43 +0000 (19:44 -0500)]
remove codec-only references from TODO

4 years agoupdate README.win32 for speexdsp
Tristan Matthews [Sun, 7 Dec 2014 00:42:47 +0000 (19:42 -0500)]
update README.win32 for speexdsp

4 years agoremove codec-only info from README.Trimedia
Tristan Matthews [Sun, 7 Dec 2014 00:42:15 +0000 (19:42 -0500)]
remove codec-only info from README.Trimedia

4 years agoremoved codec-only README.TI-DSP
Tristan Matthews [Sun, 7 Dec 2014 00:24:23 +0000 (19:24 -0500)]
removed codec-only README.TI-DSP

4 years agodropped codec-only README.symbian
Tristan Matthews [Sun, 7 Dec 2014 00:23:27 +0000 (19:23 -0500)]
dropped codec-only README.symbian

4 years agoREADME: drop references to codec
Tristan Matthews [Sun, 7 Dec 2014 00:22:55 +0000 (19:22 -0500)]
README: drop references to codec

4 years agoUpdated README.blackfin for speexdsp
Tristan Matthews [Sun, 7 Dec 2014 00:15:40 +0000 (19:15 -0500)]
Updated README.blackfin for speexdsp

4 years agoensure speex_buffer.h is included in dist but not installed
Tristan Matthews [Sat, 6 Dec 2014 23:49:24 +0000 (18:49 -0500)]
ensure speex_buffer.h is included in dist but not installed

4 years agofix configure typo
Tristan Matthews [Sat, 6 Dec 2014 23:39:55 +0000 (18:39 -0500)]
fix configure typo

4 years agoconfigure: bump to 1.2rc2 Speex-1.2rc2
Tristan Matthews [Sat, 6 Dec 2014 20:40:23 +0000 (15:40 -0500)]
configure: bump to 1.2rc2

5 years agoconfigure: add --disable-examples option
Tristan Matthews [Sun, 12 Oct 2014 01:38:08 +0000 (21:38 -0400)]
configure: add --disable-examples option

5 years agodoc: remove codec-specific documentation
Tristan Matthews [Fri, 26 Sep 2014 19:06:48 +0000 (15:06 -0400)]
doc: remove codec-specific documentation

5 years agospeexdsp: fix SSE2 support
Tristan Matthews [Sun, 17 Aug 2014 17:24:50 +0000 (10:24 -0700)]
speexdsp: fix SSE2 support

_USE_SSE2 was only being defined for win32 builds on x86-64

5 years agoos_support: fix misleading comments
Tristan Matthews [Sun, 10 Aug 2014 17:42:15 +0000 (13:42 -0400)]
os_support: fix misleading comments

5 years agoos_support: fix comment
Tristan Matthews [Sun, 10 Aug 2014 16:43:00 +0000 (12:43 -0400)]
os_support: fix comment

5 years agoresample: recover from integer overflow in update_filter()
Karl Tomlinson [Thu, 7 Aug 2014 04:44:46 +0000 (00:44 -0400)]
resample: recover from integer overflow in update_filter()

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoresample: recover from larger memory allocation failures
Karl Tomlinson [Thu, 7 Aug 2014 04:04:02 +0000 (00:04 -0400)]
resample: recover from larger memory allocation failures

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoresample: use a single common block to allocate sinc_table
Karl Tomlinson [Thu, 7 Aug 2014 03:42:15 +0000 (23:42 -0400)]
resample: use a single common block to allocate sinc_table

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agotestresample: fix underlinking of libm
Tristan Matthews [Thu, 7 Aug 2014 04:24:45 +0000 (00:24 -0400)]
testresample: fix underlinking of libm

5 years agoconfigure: fix PKG_CHECK_MODULES invocation
Tristan Matthews [Wed, 23 Jul 2014 21:22:19 +0000 (17:22 -0400)]
configure: fix PKG_CHECK_MODULES invocation

Patch by chutz.
https://trac.xiph.org/ticket/1825

5 years agoresample: make some tables constant
Diego Elio Pettenò [Wed, 23 Jul 2014 20:41:40 +0000 (16:41 -0400)]
resample: make some tables constant

Patch adapted from https://trac.xiph.org/ticket/1298

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoresample: use common statement to calculate st->mem required size per channel
Karl Tomlinson [Thu, 3 Jul 2014 14:53:22 +0000 (10:53 -0400)]
resample: use common statement to calculate st->mem required size per channel

5 years agoresample: unify remaining allocations of st->mem in update_filter()
Karl Tomlinson [Thu, 3 Jul 2014 05:38:26 +0000 (01:38 -0400)]
resample: unify remaining allocations of st->mem in update_filter()

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoresample: use spx_uint32_t for index and length
Karl Tomlinson [Thu, 3 Jul 2014 05:35:54 +0000 (01:35 -0400)]
resample: use spx_uint32_t for index and length

This is a little more consistent than spx_int32_t with other lengths.

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoresample: unify alloc/realloc paths in update_filter()
Karl Tomlinson [Thu, 3 Jul 2014 05:29:55 +0000 (01:29 -0400)]
resample: unify alloc/realloc paths in update_filter()

The only difference between speex_alloc(size) and speex_realloc(NULL, size)
is that speex_alloc zero-initializes the memory, but every byte of memory
allocated is subsequently set anyway.

This also means that st->sinc_table_length is initialized even on the first
allocation of sinc_table.

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agobuild testresample
Tristan Matthews [Wed, 2 Jul 2014 15:39:04 +0000 (11:39 -0400)]
build testresample

5 years agoresample: Add NEON optimized inner_product_single for floating point
Jyri Sarha [Thu, 1 Sep 2011 12:30:31 +0000 (15:30 +0300)]
resample: Add NEON optimized inner_product_single for floating point

Also adds inline asm implementations of WORD2INT(x) macro for fixed
and floating point.

5 years agoresample: Add NEON optimized inner_product_single for fixed point
Jyri Sarha [Thu, 1 Sep 2011 12:30:29 +0000 (15:30 +0300)]
resample: Add NEON optimized inner_product_single for fixed point

Semantics of inner_product_single have also been changed to contain
the final right shift and saturation so it can also be implemented in
the optimal way for the used platform. This change affects fixed point
calculations only.

I also added a new fixed point macro SATURATE32PSHR(x, shift, a). It
does pretty much the same thing as SATURATE32(PSHR32(x, shift), a),
but it avoids over flowing in rounding up phase in the rare occasion
where x has already been saturated. It should also be slightly faster.

5 years agoconfigure.ac: Add ARM NEON support
Jyri Sarha [Thu, 1 Sep 2011 12:30:30 +0000 (15:30 +0300)]
configure.ac: Add ARM NEON support

Use --enable-neon to force NEON optimization on. The auto detection
should also work if your CFLAGS supports NEON.

5 years agoresample: Calculate full sinc table (e.g. no sinc interpolation)
Jyri Sarha [Thu, 1 Sep 2011 12:30:27 +0000 (15:30 +0300)]
resample: Calculate full sinc table (e.g. no sinc interpolation)

On many systems taking couple of kilo-bytes for full sinc table is not
a problem. With a straight filter table the number of MAC operations
needed drops to one quarter. Also the inner loop of the filter becomes
more simple and easier to optimize. So the full sinc table version
runs much faster in most cases. The trade off is some kilo-bytes
bigger memory usage and more CPU needed when initializing the
resampler.

configure.ac: Add --enable-resample-full-sinc-table conf flag

Use this flag to enable resampling with full sinc table.

5 years agoMake M_PI the same as what math.h defines it as
Ron [Wed, 2 Jul 2014 02:46:46 +0000 (12:16 +0930)]
Make M_PI the same as what math.h defines it as

Since the math.h one is what most systems will use anyway.

5 years agoContinue to install the headers in $prefix/include/speex
Ron [Fri, 27 Jun 2014 07:07:18 +0000 (16:37 +0930)]
Continue to install the headers in $prefix/include/speex

The $pkgincludedir for this package is now $includedir/speexdsp, which
isn't where established users are looking for these headers. Thanks to
the vlc folks for noticing this.

5 years agoRemove codec specific file
Tristan Matthews [Fri, 27 Jun 2014 06:54:55 +0000 (02:54 -0400)]
Remove codec specific file

5 years agoresample: complete speex_resampler_reset_mem
Karl Tomlinson [Wed, 25 Jun 2014 18:53:50 +0000 (14:53 -0400)]
resample: complete speex_resampler_reset_mem

sample_frac_num is the array most likely to be non-zero after resampling a
previous stream, but the other arrays can also be left at non-zero values in
other situations.

Using a previous sample_frac_num can mean that the first input sample is
ignored.  This may usually not be significant, but may just be detectable when
only one channel of a stereo signal is affected.

Reviewed by Jean-Marc Valin <jmvalin@mozilla.com> at
https://bugzilla.mozilla.org/show_bug.cgi?id=973374

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoresample: kill some warnings when using sse
Karl Tomlinson [Wed, 25 Jun 2014 18:52:21 +0000 (14:52 -0400)]
resample: kill some warnings when using sse

Cherry-picked from opus-tools d1354febb0fad9e5e9016530666e5b5974c2aba7

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoresample: rename sinc to sinct to fix warnings
Karl Tomlinson [Wed, 25 Jun 2014 18:49:40 +0000 (14:49 -0400)]
resample: rename sinc to sinct to fix warnings

Cherry-picked from opus-tools 25c9d378d527604f86819d1cc33e77b4f36f3ae5

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoscal: remove dead code
Tristan Matthews [Tue, 22 Apr 2014 05:40:52 +0000 (01:40 -0400)]
scal: remove dead code

beta is set but immediately overrwritten before being read, therefore this
line had no effect

5 years agojitter: remove unused variable
Tristan Matthews [Tue, 22 Apr 2014 05:37:23 +0000 (01:37 -0400)]
jitter: remove unused variable

5 years agoFix file permission of libspeexdsp/scal.c
Peter Meerwald [Tue, 15 Apr 2014 16:33:50 +0000 (18:33 +0200)]
Fix file permission of libspeexdsp/scal.c

Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>
Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoResampler buffer_size is always 160, drop FIXED_POINT #ifdef
Peter Meerwald [Tue, 15 Apr 2014 16:33:48 +0000 (18:33 +0200)]
Resampler buffer_size is always 160, drop FIXED_POINT #ifdef

Signed-off-by: Peter Meerwald <pmeerw@pmeerw.net>
Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoRemove speex codec specific file
Tristan Matthews [Fri, 31 Jan 2014 22:37:31 +0000 (17:37 -0500)]
Remove speex codec specific file

5 years agoRemove dependency on speex
Rafaël Carré [Fri, 31 Jan 2014 21:39:05 +0000 (16:39 -0500)]
Remove dependency on speex

Cherry-picked from VLC commit 0f025b2eea9e2ae08b48209cbe5456c82caa3f47

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoDon't overlink with libm
Ron [Sat, 7 Dec 2013 16:38:35 +0000 (03:08 +1030)]
Don't overlink with libm

This will also work on systems where libm is not required.

5 years agoPrefer the stdint types if available
Ron [Sat, 7 Dec 2013 16:38:07 +0000 (03:08 +1030)]
Prefer the stdint types if available

This is similar to the change we did in libogg some time back,
it means the generated header is less likely to vary unnecessarily
between system architectures.

5 years agoDrop the --disable-vbr option from autoconf
Ron [Sat, 7 Dec 2013 14:42:58 +0000 (01:12 +1030)]
Drop the --disable-vbr option from autoconf

That's not relevant here anymore, it's a speex option.

5 years agoMake autotools a bit less whiny on modern systems
Ron [Sat, 7 Dec 2013 00:47:22 +0000 (11:17 +1030)]
Make autotools a bit less whiny on modern systems

and clean up some more things that belonged to speex but aren't used
in this source tree.

Drop the SPEEX*_VERSION variables completely, we only use PACKAGE_VERSION
in the .pc file in this code.

5 years agoDrop the hardcoded SPEEX_VERSION from arch.h
Ron [Fri, 6 Dec 2013 22:40:35 +0000 (09:10 +1030)]
Drop the hardcoded SPEEX_VERSION from arch.h

It's out of date, but isn't used anywhere in this source tree.
The only place the one from configure.ac is used is in speexdsp.pc.in,
and that should probably be changed to SPEEXDSP_VERSION now.

5 years agoDrop doc/nb_celp.c
Ron [Fri, 6 Dec 2013 22:39:24 +0000 (09:09 +1030)]
Drop doc/nb_celp.c

It's no longer referenced by doc/manual.lyx in this tree.

5 years agoTidy the speex/*.h includes some more
Ron [Fri, 6 Dec 2013 22:33:14 +0000 (09:03 +1030)]
Tidy the speex/*.h includes some more

This is a partial revert of 785e61a593c61ebc0afee5d87eca9cc73e16d6a6.

The _BUILD_SPEEX special case isn't needed in the public headers, and
we don't need '../include' relative paths in the internal files.

Don't include speex.h in jitter.c, it's not needed there, and no longer
present in this source tree.

Switch from speex_types.h to speexdsp_types.h in arch.h, it's not clear
yet what the best thing to do with the latter will be (since they are
duplicates of each other and must always stay synchronised), but the
former is no longer present in this source tree anymore either.

5 years agoFix typos in manual for speex_preprocess API
Tristan Matthews [Fri, 6 Dec 2013 19:09:46 +0000 (14:09 -0500)]
Fix typos in manual for speex_preprocess API

5 years agoSwitch autogen.sh to use autoreconf
Ron [Fri, 6 Dec 2013 06:20:40 +0000 (01:20 -0500)]
Switch autogen.sh to use autoreconf

The previous version didn't actually update everything needed
(like config.{guess,sub} at the very least).

5 years agoEnsure the filter length is a multiple of 8
Ron [Fri, 6 Dec 2013 06:18:13 +0000 (01:18 -0500)]
Ensure the filter length is a multiple of 8

Fixes a bug where the SSE implementation would read past the end of the
buffer if it wasn't.

Also taken from f426ab5260fe9f1618a187fc2a8264dbba4aef08 in opus-tools.

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoBe precise computing the resampler with the smallest memory footprint
Ron [Fri, 6 Dec 2013 06:16:20 +0000 (01:16 -0500)]
Be precise computing the resampler with the smallest memory footprint

The previous version was only wrong in some fairly obscure corner cases,
when oversample < den_rate <= oversample + 8 / filt_len, but since the
direct filter seems to mostly be the fastest, and that would err on the
side of choosing the interpolated one, let's do it right now anyway.

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoProperly save in_len for multiple channels in the resampler.
Jean-Marc Valin [Fri, 6 Dec 2013 06:14:28 +0000 (01:14 -0500)]
Properly save in_len for multiple channels in the resampler.

This fixes issues with clicking in one channel and/or truncation
with some unusual sample rates.

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
5 years agoFixed potential over-allocation / (fatal) under-allocation on platforms where sizeof...
Sherief N. Farouk [Fri, 6 Dec 2013 06:07:16 +0000 (01:07 -0500)]
Fixed potential over-allocation / (fatal) under-allocation on platforms where sizeof(int) != 4.

Signed-off-by: Tristan Matthews <le.businessman@gmail.com>
9 years agoAdd missing \ so headers are included in tarball
Thorvald Natvig [Mon, 1 Feb 2010 12:29:03 +0000 (13:29 +0100)]
Add missing \ so headers are included in tarball

9 years agoFix a few make dist issues with win32 directory
Thorvald Natvig [Sun, 17 Jan 2010 04:07:50 +0000 (05:07 +0100)]
Fix a few make dist issues with win32 directory

9 years agoClean up other platforms; remove codec files
Thorvald Natvig [Thu, 7 Jan 2010 16:56:39 +0000 (17:56 +0100)]
Clean up other platforms; remove codec files

9 years agoClean up embedded targets
slicer [Thu, 7 Jan 2010 16:18:55 +0000 (17:18 +0100)]
Clean up embedded targets

9 years agoRemove unused tests from configure.ac
slicer [Thu, 7 Jan 2010 16:05:18 +0000 (17:05 +0100)]
Remove unused tests from configure.ac

9 years agoRemove codec-only files
slicer [Thu, 7 Jan 2010 15:59:45 +0000 (16:59 +0100)]
Remove codec-only files

9 years agoSplit libspeex and libspeexdsp
slicer [Tue, 22 Dec 2009 19:02:03 +0000 (20:02 +0100)]
Split libspeex and libspeexdsp

10 years agoBetter Doxygen comments on latency
Alexander Chemeris [Wed, 8 Jul 2009 04:29:03 +0000 (00:29 -0400)]
Better Doxygen comments on latency

10 years agoAdding info on the algorithmic delay
Alexander Chemeris [Wed, 8 Jul 2009 04:26:06 +0000 (00:26 -0400)]
Adding info on the algorithmic delay

10 years agoFixing typos in the manual
Alexander Chemeris [Wed, 8 Jul 2009 04:25:11 +0000 (00:25 -0400)]
Fixing typos in the manual

10 years agoSupport for Intel IPP FFT
Jeff Wallace [Wed, 8 Jul 2009 04:24:21 +0000 (00:24 -0400)]
Support for Intel IPP FFT

10 years agoTI project fix (for EXPORT)
Jim Crichton [Wed, 8 Jul 2009 04:22:51 +0000 (00:22 -0400)]
TI project fix (for EXPORT)

10 years agoRemoving h1_mem from the encoder state since it wasn't used at all.
John Ridges [Fri, 26 Jun 2009 22:43:43 +0000 (18:43 -0400)]
Removing h1_mem from the encoder state since it wasn't used at all.

10 years agoThis patch simply flips on the _USE_SSE/_USE_SSE2 automatically if the visual
Jeff Wallace [Tue, 23 Jun 2009 11:40:35 +0000 (07:40 -0400)]
This patch simply flips on the _USE_SSE/_USE_SSE2 automatically if the visual
studio compiler options for SSE/SSE2 are specified or if compiling for x86-64.

10 years agoFixes a regression in the fixed-point code and adds saturation for fixed-point.
Stephane Lesage [Thu, 18 Jun 2009 11:19:22 +0000 (07:19 -0400)]
Fixes a regression in the fixed-point code and adds saturation for fixed-point.

10 years agofix -elf2flt usage for bfin-uclinux
Mike Frysinger [Wed, 3 Jun 2009 21:22:44 +0000 (17:22 -0400)]
fix -elf2flt usage for bfin-uclinux

There are multiple problems with the elf2flt usage in the configure script:
 - clobbers LDFLAGS
 - is used for all targets so breaks non-FLAT targets
 - is only used for Blackfin FLAT targets

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
10 years agomake sure new bfin.h is included in dist
Mike Frysinger [Wed, 3 Jun 2009 21:16:09 +0000 (17:16 -0400)]
make sure new bfin.h is included in dist

Looks like I didn't test the `make dist` target and so broke it by not
listing the bfin.h in one of the headers vars.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
10 years agoBlackfin: cleanup astat/cc/hardware loop asm clobbers
Mike Frysinger [Fri, 24 Apr 2009 23:15:42 +0000 (19:15 -0400)]
Blackfin: cleanup astat/cc/hardware loop asm clobbers

Most asm statements clobber ASTAT bits (shifts, maxes, etc...) but do
declare the register as clobbered.  Same thing with CC in a few places.
Some places make an attempt at clobbering some hardware loop registers,
but it's very incomplete compared with how many asm statements actually
use hardware loops.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
10 years agoAdded the --no-highpass and --print-rate options to speexenc
Jean-Marc Valin [Wed, 15 Apr 2009 19:02:23 +0000 (15:02 -0400)]
Added the --no-highpass and --print-rate options to speexenc

10 years agoResurected the --enable-vorbis-psy option by linking with libspeexdsp for the
Jean-Marc Valin [Wed, 15 Apr 2009 18:19:04 +0000 (14:19 -0400)]
Resurected the --enable-vorbis-psy option by linking with libspeexdsp for the
fft.

10 years agoMore VBR tuning
Jean-Marc Valin [Wed, 15 Apr 2009 18:00:43 +0000 (14:00 -0400)]
More VBR tuning

10 years agoFixed a "conditional jump depends on uninitialised value in speex_decode_int()
Jean-Marc Valin [Wed, 8 Apr 2009 18:42:28 +0000 (14:42 -0400)]
Fixed a "conditional jump depends on uninitialised value in speex_decode_int()
even though it wasn't an actual bug.

10 years agoVBR tuning (along with open-loop gamma2 tweak)
Jean-Marc Valin [Fri, 3 Apr 2009 15:52:04 +0000 (11:52 -0400)]
VBR tuning (along with open-loop gamma2 tweak)

10 years agoTuning the enhancer and gamm1
Jean-Marc Valin [Fri, 3 Apr 2009 14:25:53 +0000 (10:25 -0400)]
Tuning the enhancer and gamm1

10 years agoBit-stream definition in the manual with simplified source code
Alfred E. Heggestad [Sun, 1 Mar 2009 14:18:40 +0000 (09:18 -0500)]
Bit-stream definition in the manual with simplified source code

10 years agoConst correctness patch
Alfred E. Heggestad [Tue, 10 Feb 2009 12:20:03 +0000 (07:20 -0500)]
Const correctness patch

10 years agoCosmetic fixes
Alfred E. Heggestad [Fri, 30 Jan 2009 16:30:16 +0000 (11:30 -0500)]
Cosmetic fixes

10 years agoUpdated draft
Jean-Marc Valin [Thu, 11 Dec 2008 03:53:58 +0000 (22:53 -0500)]
Updated draft

10 years agopatch: speex doc typos
Alfred E. Heggestad [Wed, 10 Dec 2008 23:11:02 +0000 (00:11 +0100)]
patch: speex doc typos

hi

here is just a small patch for speex doc, fixing some typos..
I was reading through the whole manual a couple of times.

could we also update the Internet-Draft in the Manual?

latest version is here:

  http://tools.ietf.org/html/draft-ietf-avt-rtp-speex-05

/alfred

10 years agoAt least VS2005 (what I'm using) won't compile resample_sse.h with
John Ridges [Mon, 1 Dec 2008 00:45:28 +0000 (19:45 -0500)]
At least VS2005 (what I'm using) won't compile resample_sse.h with
_USE_SSE2 defined because it refuses to cast __m128 to __m128d and vice
versa. While there are intrinsics to do the casts, I thought it would be
simpler to just use an intrinsic that accomplishes the same thing
without all the casting.

10 years agoDocumenting the pitch decoding
Jean-Marc Valin [Mon, 1 Dec 2008 00:16:59 +0000 (19:16 -0500)]
Documenting the pitch decoding

10 years agoPatch : Make speex_bits_read_from and speex_bits_read_whole_bytes const correct
Erik de Castro Lopo [Fri, 7 Nov 2008 22:49:11 +0000 (09:49 +1100)]
Patch : Make speex_bits_read_from and speex_bits_read_whole_bytes const correct

Jean-Marc,

The following patch makes the above two read functions const correct.

Cheers,
Erik

10 years agoDocumenting the decoder
Jean-Marc Valin [Sun, 9 Nov 2008 15:59:38 +0000 (10:59 -0500)]
Documenting the decoder

11 years agoFix for VS files to include multi-channel changes. Also patches scal.c for
Alexander Chemeris [Fri, 3 Oct 2008 00:38:47 +0000 (20:38 -0400)]
Fix for VS files to include multi-channel changes. Also patches scal.c for
compilers that don't have M_PI defined.

11 years agooops, setting the max packet length correctly in speex_jitter_get()
Jean-Marc Valin [Tue, 16 Sep 2008 11:22:43 +0000 (07:22 -0400)]
oops, setting the max packet length correctly in speex_jitter_get()

11 years agoThe LPC analysis window and lag window are now in Q15.
Jean-Marc Valin [Sat, 6 Sep 2008 02:31:56 +0000 (22:31 -0400)]
The LPC analysis window and lag window are now in Q15.