1
0
mirror of https://github.com/musix-org/musix-oss synced 2024-11-10 08:10:18 +00:00
This commit is contained in:
MatteZ02 2020-03-04 13:43:21 +02:00
parent 79f8a18164
commit da84fcfed1
1292 changed files with 93623 additions and 35760 deletions

15
node_modules/.bin/detect-libc generated vendored Normal file
View File

@ -0,0 +1,15 @@
#!/bin/sh
basedir=$(dirname "$(echo "$0" | sed -e 's,\\,/,g')")
case `uname` in
*CYGWIN*|*MINGW*|*MSYS*) basedir=`cygpath -w "$basedir"`;;
esac
if [ -x "$basedir/node" ]; then
"$basedir/node" "$basedir/../detect-libc/bin/detect-libc.js" "$@"
ret=$?
else
node "$basedir/../detect-libc/bin/detect-libc.js" "$@"
ret=$?
fi
exit $ret

17
node_modules/.bin/detect-libc.cmd generated vendored Normal file
View File

@ -0,0 +1,17 @@
@ECHO off
SETLOCAL
CALL :find_dp0
IF EXIST "%dp0%\node.exe" (
SET "_prog=%dp0%\node.exe"
) ELSE (
SET "_prog=node"
SET PATHEXT=%PATHEXT:;.JS;=;%
)
"%_prog%" "%dp0%\..\detect-libc\bin\detect-libc.js" %*
ENDLOCAL
EXIT /b %errorlevel%
:find_dp0
SET dp0=%~dp0
EXIT /b

18
node_modules/.bin/detect-libc.ps1 generated vendored Normal file
View File

@ -0,0 +1,18 @@
#!/usr/bin/env pwsh
$basedir=Split-Path $MyInvocation.MyCommand.Definition -Parent
$exe=""
if ($PSVersionTable.PSVersion -lt "6.0" -or $IsWindows) {
# Fix case when both the Windows and Linux builds of Node
# are installed in the same directory
$exe=".exe"
}
$ret=0
if (Test-Path "$basedir/node$exe") {
& "$basedir/node$exe" "$basedir/../detect-libc/bin/detect-libc.js" $args
$ret=$LASTEXITCODE
} else {
& "node$exe" "$basedir/../detect-libc/bin/detect-libc.js" $args
$ret=$LASTEXITCODE
}
exit $ret

15
node_modules/.bin/mkdirp generated vendored Normal file
View File

@ -0,0 +1,15 @@
#!/bin/sh
basedir=$(dirname "$(echo "$0" | sed -e 's,\\,/,g')")
case `uname` in
*CYGWIN*|*MINGW*|*MSYS*) basedir=`cygpath -w "$basedir"`;;
esac
if [ -x "$basedir/node" ]; then
"$basedir/node" "$basedir/../mkdirp/bin/cmd.js" "$@"
ret=$?
else
node "$basedir/../mkdirp/bin/cmd.js" "$@"
ret=$?
fi
exit $ret

17
node_modules/.bin/mkdirp.cmd generated vendored Normal file
View File

@ -0,0 +1,17 @@
@ECHO off
SETLOCAL
CALL :find_dp0
IF EXIST "%dp0%\node.exe" (
SET "_prog=%dp0%\node.exe"
) ELSE (
SET "_prog=node"
SET PATHEXT=%PATHEXT:;.JS;=;%
)
"%_prog%" "%dp0%\..\mkdirp\bin\cmd.js" %*
ENDLOCAL
EXIT /b %errorlevel%
:find_dp0
SET dp0=%~dp0
EXIT /b

18
node_modules/.bin/mkdirp.ps1 generated vendored Normal file
View File

@ -0,0 +1,18 @@
#!/usr/bin/env pwsh
$basedir=Split-Path $MyInvocation.MyCommand.Definition -Parent
$exe=""
if ($PSVersionTable.PSVersion -lt "6.0" -or $IsWindows) {
# Fix case when both the Windows and Linux builds of Node
# are installed in the same directory
$exe=".exe"
}
$ret=0
if (Test-Path "$basedir/node$exe") {
& "$basedir/node$exe" "$basedir/../mkdirp/bin/cmd.js" $args
$ret=$LASTEXITCODE
} else {
& "node$exe" "$basedir/../mkdirp/bin/cmd.js" $args
$ret=$LASTEXITCODE
}
exit $ret

15
node_modules/.bin/needle generated vendored Normal file
View File

@ -0,0 +1,15 @@
#!/bin/sh
basedir=$(dirname "$(echo "$0" | sed -e 's,\\,/,g')")
case `uname` in
*CYGWIN*|*MINGW*|*MSYS*) basedir=`cygpath -w "$basedir"`;;
esac
if [ -x "$basedir/node" ]; then
"$basedir/node" "$basedir/../needle/bin/needle" "$@"
ret=$?
else
node "$basedir/../needle/bin/needle" "$@"
ret=$?
fi
exit $ret

17
node_modules/.bin/needle.cmd generated vendored Normal file
View File

@ -0,0 +1,17 @@
@ECHO off
SETLOCAL
CALL :find_dp0
IF EXIST "%dp0%\node.exe" (
SET "_prog=%dp0%\node.exe"
) ELSE (
SET "_prog=node"
SET PATHEXT=%PATHEXT:;.JS;=;%
)
"%_prog%" "%dp0%\..\needle\bin\needle" %*
ENDLOCAL
EXIT /b %errorlevel%
:find_dp0
SET dp0=%~dp0
EXIT /b

18
node_modules/.bin/needle.ps1 generated vendored Normal file
View File

@ -0,0 +1,18 @@
#!/usr/bin/env pwsh
$basedir=Split-Path $MyInvocation.MyCommand.Definition -Parent
$exe=""
if ($PSVersionTable.PSVersion -lt "6.0" -or $IsWindows) {
# Fix case when both the Windows and Linux builds of Node
# are installed in the same directory
$exe=".exe"
}
$ret=0
if (Test-Path "$basedir/node$exe") {
& "$basedir/node$exe" "$basedir/../needle/bin/needle" $args
$ret=$LASTEXITCODE
} else {
& "node$exe" "$basedir/../needle/bin/needle" $args
$ret=$LASTEXITCODE
}
exit $ret

15
node_modules/.bin/node-pre-gyp generated vendored Normal file
View File

@ -0,0 +1,15 @@
#!/bin/sh
basedir=$(dirname "$(echo "$0" | sed -e 's,\\,/,g')")
case `uname` in
*CYGWIN*|*MINGW*|*MSYS*) basedir=`cygpath -w "$basedir"`;;
esac
if [ -x "$basedir/node" ]; then
"$basedir/node" "$basedir/../node-pre-gyp/bin/node-pre-gyp" "$@"
ret=$?
else
node "$basedir/../node-pre-gyp/bin/node-pre-gyp" "$@"
ret=$?
fi
exit $ret

17
node_modules/.bin/node-pre-gyp.cmd generated vendored Normal file
View File

@ -0,0 +1,17 @@
@ECHO off
SETLOCAL
CALL :find_dp0
IF EXIST "%dp0%\node.exe" (
SET "_prog=%dp0%\node.exe"
) ELSE (
SET "_prog=node"
SET PATHEXT=%PATHEXT:;.JS;=;%
)
"%_prog%" "%dp0%\..\node-pre-gyp\bin\node-pre-gyp" %*
ENDLOCAL
EXIT /b %errorlevel%
:find_dp0
SET dp0=%~dp0
EXIT /b

18
node_modules/.bin/node-pre-gyp.ps1 generated vendored Normal file
View File

@ -0,0 +1,18 @@
#!/usr/bin/env pwsh
$basedir=Split-Path $MyInvocation.MyCommand.Definition -Parent
$exe=""
if ($PSVersionTable.PSVersion -lt "6.0" -or $IsWindows) {
# Fix case when both the Windows and Linux builds of Node
# are installed in the same directory
$exe=".exe"
}
$ret=0
if (Test-Path "$basedir/node$exe") {
& "$basedir/node$exe" "$basedir/../node-pre-gyp/bin/node-pre-gyp" $args
$ret=$LASTEXITCODE
} else {
& "node$exe" "$basedir/../node-pre-gyp/bin/node-pre-gyp" $args
$ret=$LASTEXITCODE
}
exit $ret

15
node_modules/.bin/nopt generated vendored Normal file
View File

@ -0,0 +1,15 @@
#!/bin/sh
basedir=$(dirname "$(echo "$0" | sed -e 's,\\,/,g')")
case `uname` in
*CYGWIN*|*MINGW*|*MSYS*) basedir=`cygpath -w "$basedir"`;;
esac
if [ -x "$basedir/node" ]; then
"$basedir/node" "$basedir/../nopt/bin/nopt.js" "$@"
ret=$?
else
node "$basedir/../nopt/bin/nopt.js" "$@"
ret=$?
fi
exit $ret

17
node_modules/.bin/nopt.cmd generated vendored Normal file
View File

@ -0,0 +1,17 @@
@ECHO off
SETLOCAL
CALL :find_dp0
IF EXIST "%dp0%\node.exe" (
SET "_prog=%dp0%\node.exe"
) ELSE (
SET "_prog=node"
SET PATHEXT=%PATHEXT:;.JS;=;%
)
"%_prog%" "%dp0%\..\nopt\bin\nopt.js" %*
ENDLOCAL
EXIT /b %errorlevel%
:find_dp0
SET dp0=%~dp0
EXIT /b

18
node_modules/.bin/nopt.ps1 generated vendored Normal file
View File

@ -0,0 +1,18 @@
#!/usr/bin/env pwsh
$basedir=Split-Path $MyInvocation.MyCommand.Definition -Parent
$exe=""
if ($PSVersionTable.PSVersion -lt "6.0" -or $IsWindows) {
# Fix case when both the Windows and Linux builds of Node
# are installed in the same directory
$exe=".exe"
}
$ret=0
if (Test-Path "$basedir/node$exe") {
& "$basedir/node$exe" "$basedir/../nopt/bin/nopt.js" $args
$ret=$LASTEXITCODE
} else {
& "node$exe" "$basedir/../nopt/bin/nopt.js" $args
$ret=$LASTEXITCODE
}
exit $ret

15
node_modules/.bin/rc generated vendored Normal file
View File

@ -0,0 +1,15 @@
#!/bin/sh
basedir=$(dirname "$(echo "$0" | sed -e 's,\\,/,g')")
case `uname` in
*CYGWIN*|*MINGW*|*MSYS*) basedir=`cygpath -w "$basedir"`;;
esac
if [ -x "$basedir/node" ]; then
"$basedir/node" "$basedir/../rc/cli.js" "$@"
ret=$?
else
node "$basedir/../rc/cli.js" "$@"
ret=$?
fi
exit $ret

17
node_modules/.bin/rc.cmd generated vendored Normal file
View File

@ -0,0 +1,17 @@
@ECHO off
SETLOCAL
CALL :find_dp0
IF EXIST "%dp0%\node.exe" (
SET "_prog=%dp0%\node.exe"
) ELSE (
SET "_prog=node"
SET PATHEXT=%PATHEXT:;.JS;=;%
)
"%_prog%" "%dp0%\..\rc\cli.js" %*
ENDLOCAL
EXIT /b %errorlevel%
:find_dp0
SET dp0=%~dp0
EXIT /b

18
node_modules/.bin/rc.ps1 generated vendored Normal file
View File

@ -0,0 +1,18 @@
#!/usr/bin/env pwsh
$basedir=Split-Path $MyInvocation.MyCommand.Definition -Parent
$exe=""
if ($PSVersionTable.PSVersion -lt "6.0" -or $IsWindows) {
# Fix case when both the Windows and Linux builds of Node
# are installed in the same directory
$exe=".exe"
}
$ret=0
if (Test-Path "$basedir/node$exe") {
& "$basedir/node$exe" "$basedir/../rc/cli.js" $args
$ret=$LASTEXITCODE
} else {
& "node$exe" "$basedir/../rc/cli.js" $args
$ret=$LASTEXITCODE
}
exit $ret

15
node_modules/.bin/rimraf generated vendored Normal file
View File

@ -0,0 +1,15 @@
#!/bin/sh
basedir=$(dirname "$(echo "$0" | sed -e 's,\\,/,g')")
case `uname` in
*CYGWIN*|*MINGW*|*MSYS*) basedir=`cygpath -w "$basedir"`;;
esac
if [ -x "$basedir/node" ]; then
"$basedir/node" "$basedir/../rimraf/bin.js" "$@"
ret=$?
else
node "$basedir/../rimraf/bin.js" "$@"
ret=$?
fi
exit $ret

17
node_modules/.bin/rimraf.cmd generated vendored Normal file
View File

@ -0,0 +1,17 @@
@ECHO off
SETLOCAL
CALL :find_dp0
IF EXIST "%dp0%\node.exe" (
SET "_prog=%dp0%\node.exe"
) ELSE (
SET "_prog=node"
SET PATHEXT=%PATHEXT:;.JS;=;%
)
"%_prog%" "%dp0%\..\rimraf\bin.js" %*
ENDLOCAL
EXIT /b %errorlevel%
:find_dp0
SET dp0=%~dp0
EXIT /b

18
node_modules/.bin/rimraf.ps1 generated vendored Normal file
View File

@ -0,0 +1,18 @@
#!/usr/bin/env pwsh
$basedir=Split-Path $MyInvocation.MyCommand.Definition -Parent
$exe=""
if ($PSVersionTable.PSVersion -lt "6.0" -or $IsWindows) {
# Fix case when both the Windows and Linux builds of Node
# are installed in the same directory
$exe=".exe"
}
$ret=0
if (Test-Path "$basedir/node$exe") {
& "$basedir/node$exe" "$basedir/../rimraf/bin.js" $args
$ret=$LASTEXITCODE
} else {
& "node$exe" "$basedir/../rimraf/bin.js" $args
$ret=$LASTEXITCODE
}
exit $ret

1
node_modules/@discordjs/opus/.gitattributes generated vendored Normal file
View File

@ -0,0 +1 @@
* text=auto eol=lf

View File

@ -0,0 +1,22 @@
name: Build
on: [push, pull_request]
jobs:
build:
name: Build
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [macos-latest, ubuntu-latest, windows-latest]
node: [12, 13]
fail-fast: true
steps:
- name: Checkout repository
uses: actions/checkout@v2
- name: Install Node v${{ matrix.node }}
uses: actions/setup-node@v1
with:
node-version: ${{ matrix.node }}
- name: Install dependencies
run: yarn install

View File

@ -0,0 +1,34 @@
name: Prebuild
on:
release:
types: [published]
jobs:
build:
name: Prebuild
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [macos-latest, ubuntu-latest, windows-latest]
node: [12, 13]
fail-fast: true
steps:
- name: Checkout repository
uses: actions/checkout@v2
- name: Install Node v${{ matrix.node }}
uses: actions/setup-node@v1
with:
node-version: ${{ matrix.node }}
- name: Install dependencies
run: yarn install
- name: Package prebuild
run: yarn build
- name: Upload prebuild asset
uses: icrawl/action-artifact@v1
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
with:
path: 'build/stage/**/*.tar.gz'

View File

@ -1,6 +1,7 @@
The MIT License (MIT) MIT License
Copyright (c) 2016 Zeit, Inc. Copyright (c) 2020 iCrawl
Copyright (c) 2013-2019 Mikko Rantanen
Permission is hereby granted, free of charge, to any person obtaining a copy Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal of this software and associated documentation files (the "Software"), to deal

51
node_modules/@discordjs/opus/binding.gyp generated vendored Normal file
View File

@ -0,0 +1,51 @@
{
"targets": [
{
"target_name": "<(module_name)",
"product_dir": "<(module_path)",
"dependencies": [
"deps/binding.gyp:libopus",
],
"cflags!": [
"-fno-exceptions",
],
"cflags_cc!": [
"-fno-exceptions",
],
"cflags": [
"-pthread",
"-fno-strict-aliasing",
"-Wall",
"-Wno-unused-parameter",
"-Wno-missing-field-initializers",
"-Wextra",
"-pipe",
"-fno-ident",
"-fdata-sections",
"-ffunction-sections",
"-fPIC",
],
"defines": [
"LARGEFILE_SOURCE",
"_FILE_OFFSET_BITS=64",
"WEBRTC_TARGET_PC",
"WEBRTC_LINUX",
"WEBRTC_THREAD_RR",
"EXPAT_RELATIVE_PATH",
"GTEST_RELATIVE_PATH",
"JSONCPP_RELATIVE_PATH",
"WEBRTC_RELATIVE_PATH",
"POSIX,"
"__STDC_FORMAT_MACROS",
"DYNAMIC_ANNOTATIONS_ENABLED=0",
"NAPI_DISABLE_CPP_EXCEPTIONS",
],
"include_dirs": [
"<!@(node -p \"require('node-addon-api').include\")"
],
"sources": [
"src/node-opus.cc",
],
},
],
}

224
node_modules/@discordjs/opus/deps/binding.gyp generated vendored Normal file
View File

@ -0,0 +1,224 @@
{
"variables": {
"target_arch%": "x64"
},
"target_defaults": {
"default_configuration": "Debug",
"configuration": {
"Debug": {
"defines": [ "DEBUG", "_DEBUG" ],
"msvs_settings": {
"VSSLCompilerTool": {
"RuntimeLibrary": 1,
},
},
},
"Release": {
"defines": [ "NODEBUG" ],
"msvs_settings": {
"VSSLCompilerTool": {
"RuntimeLibrary": 0,
},
},
},
},
"msvs_settings": {
"VCLinkerTool": {
"GenerateDebugInformation": "true",
},
},
},
"targets": [
{
"target_name": "libopus",
"type": "static_library",
"sources": [
"opus/src/opus_multistream.c",
"opus/src/opus_projection_encoder.c",
"opus/src/analysis.c",
"opus/src/mlp_data.c",
"opus/src/opus_multistream_encoder.c",
"opus/src/opus_projection_decoder.c",
"opus/src/mapping_matrix.c",
"opus/src/opus_compare.c",
"opus/src/mlp.c",
"opus/src/opus.c",
"opus/src/opus_multistream_decoder.c",
"opus/src/opus_decoder.c",
"opus/src/repacketizer.c",
"opus/src/opus_encoder.c",
"opus/silk/decode_frame.c",
"opus/silk/float/inner_product_FLP.c",
"opus/silk/float/scale_vector_FLP.c",
"opus/silk/float/find_pred_coefs_FLP.c",
"opus/silk/float/schur_FLP.c",
"opus/silk/float/warped_autocorrelation_FLP.c",
"opus/silk/float/burg_modified_FLP.c",
"opus/silk/float/find_LPC_FLP.c",
"opus/silk/float/LPC_inv_pred_gain_FLP.c",
"opus/silk/float/scale_copy_vector_FLP.c",
"opus/silk/float/noise_shape_analysis_FLP.c",
"opus/silk/float/pitch_analysis_core_FLP.c",
"opus/silk/float/bwexpander_FLP.c",
"opus/silk/float/LTP_analysis_filter_FLP.c",
"opus/silk/float/LTP_scale_ctrl_FLP.c",
"opus/silk/float/corrMatrix_FLP.c",
"opus/silk/float/encode_frame_FLP.c",
"opus/silk/float/sort_FLP.c",
"opus/silk/float/find_pitch_lags_FLP.c",
"opus/silk/float/residual_energy_FLP.c",
"opus/silk/float/LPC_analysis_filter_FLP.c",
"opus/silk/float/autocorrelation_FLP.c",
"opus/silk/float/k2a_FLP.c",
"opus/silk/float/regularize_correlations_FLP.c",
"opus/silk/float/find_LTP_FLP.c",
"opus/silk/float/energy_FLP.c",
"opus/silk/float/apply_sine_window_FLP.c",
"opus/silk/float/wrappers_FLP.c",
"opus/silk/float/process_gains_FLP.c",
"opus/silk/stereo_quant_pred.c",
"opus/silk/LPC_inv_pred_gain.c",
"opus/silk/process_NLSFs.c",
"opus/silk/NSQ.c",
"opus/silk/check_control_input.c",
"opus/silk/NLSF_del_dec_quant.c",
"opus/silk/LPC_analysis_filter.c",
"opus/silk/dec_API.c",
"opus/silk/resampler.c",
"opus/silk/sort.c",
"opus/silk/VAD.c",
"opus/silk/resampler_private_AR2.c",
"opus/silk/LPC_fit.c",
"opus/silk/control_SNR.c",
"opus/silk/decode_parameters.c",
"opus/silk/pitch_est_tables.c",
"opus/silk/fixed/warped_autocorrelation_FIX.c",
"opus/silk/fixed/apply_sine_window_FIX.c",
"opus/silk/fixed/residual_energy16_FIX.c",
"opus/silk/fixed/schur64_FIX.c",
"opus/silk/fixed/residual_energy_FIX.c",
"opus/silk/fixed/noise_shape_analysis_FIX.c",
"opus/silk/fixed/encode_frame_FIX.c",
"opus/silk/fixed/schur_FIX.c",
"opus/silk/fixed/autocorr_FIX.c",
"opus/silk/fixed/burg_modified_FIX.c",
"opus/silk/fixed/pitch_analysis_core_FIX.c",
"opus/silk/fixed/find_LTP_FIX.c",
"opus/silk/fixed/find_LPC_FIX.c",
"opus/silk/fixed/corrMatrix_FIX.c",
"opus/silk/fixed/k2a_FIX.c",
"opus/silk/fixed/LTP_scale_ctrl_FIX.c",
"opus/silk/fixed/process_gains_FIX.c",
"opus/silk/fixed/k2a_Q16_FIX.c",
"opus/silk/fixed/regularize_correlations_FIX.c",
"opus/silk/fixed/LTP_analysis_filter_FIX.c",
"opus/silk/fixed/vector_ops_FIX.c",
"opus/silk/fixed/find_pitch_lags_FIX.c",
"opus/silk/fixed/find_pred_coefs_FIX.c",
"opus/silk/control_audio_bandwidth.c",
"opus/silk/decoder_set_fs.c",
"opus/silk/NLSF_unpack.c",
"opus/silk/bwexpander.c",
"opus/silk/resampler_rom.c",
"opus/silk/shell_coder.c",
"opus/silk/decode_pulses.c",
"opus/silk/bwexpander_32.c",
"opus/silk/decode_core.c",
"opus/silk/PLC.c",
"opus/silk/tables_NLSF_CB_WB.c",
"opus/silk/table_LSF_cos.c",
"opus/silk/tables_pulses_per_block.c",
"opus/silk/tables_gain.c",
"opus/silk/inner_prod_aligned.c",
"opus/silk/resampler_down2_3.c",
"opus/silk/NSQ_del_dec.c",
"opus/silk/decode_pitch.c",
"opus/silk/NLSF_VQ_weights_laroia.c",
"opus/silk/interpolate.c",
"opus/silk/debug.c",
"opus/silk/tables_other.c",
"opus/silk/LP_variable_cutoff.c",
"opus/silk/NLSF_decode.c",
"opus/silk/encode_pulses.c",
"opus/silk/control_codec.c",
"opus/silk/stereo_LR_to_MS.c",
"opus/silk/HP_variable_cutoff.c",
"opus/silk/encode_indices.c",
"opus/silk/init_decoder.c",
"opus/silk/stereo_encode_pred.c",
"opus/silk/NLSF_VQ.c",
"opus/silk/init_encoder.c",
"opus/silk/resampler_private_IIR_FIR.c",
"opus/silk/resampler_private_up2_HQ.c",
"opus/silk/sigm_Q15.c",
"opus/silk/sum_sqr_shift.c",
"opus/silk/tables_LTP.c",
"opus/silk/resampler_down2.c",
"opus/silk/code_signs.c",
"opus/silk/tables_NLSF_CB_NB_MB.c",
"opus/silk/gain_quant.c",
"opus/silk/tables_pitch_lag.c",
"opus/silk/NLSF_stabilize.c",
"opus/silk/stereo_find_predictor.c",
"opus/silk/A2NLSF.c",
"opus/silk/NLSF2A.c",
"opus/silk/VQ_WMat_EC.c",
"opus/silk/NLSF_encode.c",
"opus/silk/log2lin.c",
"opus/silk/stereo_decode_pred.c",
"opus/silk/lin2log.c",
"opus/silk/CNG.c",
"opus/silk/enc_API.c",
"opus/silk/biquad_alt.c",
"opus/silk/quant_LTP_gains.c",
"opus/silk/resampler_private_down_FIR.c",
"opus/silk/ana_filt_bank_1.c",
"opus/silk/stereo_MS_to_LR.c",
"opus/silk/decode_indices.c",
"opus/celt/rate.c",
"opus/celt/entdec.c",
"opus/celt/modes.c",
"opus/celt/celt_lpc.c",
"opus/celt/laplace.c",
"opus/celt/cwrs.c",
"opus/celt/celt.c",
"opus/celt/entcode.c",
"opus/celt/celt_decoder.c",
"opus/celt/celt_encoder.c",
"opus/celt/mdct.c",
"opus/celt/quant_bands.c",
"opus/celt/vq.c",
"opus/celt/bands.c",
"opus/celt/kiss_fft.c",
"opus/celt/entenc.c",
"opus/celt/mathops.c",
"opus/celt/pitch.c",
],
"cflags": [
"-fvisibility=hidden",
"-W",
"-Wstrict-prototypes",
"-Wall",
"-Wextra",
"-Wcast-align",
"-Wnested-externs",
"-Wshadow",
"-Wno-parentheses",
"-Wno-unused-parameter",
"-Wno-sign-compare",
"-Wno-maybe-uninitialized",
],
"include_dirs": [
"config/opus/<(OS)/<(target_arch)",
"opus/include",
"opus/celt",
"opus/silk",
"opus/silk/float",
],
"defines": [
"PIC",
"HAVE_CONFIG_H",
],
},
],
}

View File

@ -152,7 +152,7 @@
#define PACKAGE_NAME "opus" #define PACKAGE_NAME "opus"
/* Define to the full name and version of this package. */ /* Define to the full name and version of this package. */
#define PACKAGE_STRING "opus 1.1.2" #define PACKAGE_STRING "opus 1.3.1"
/* Define to the one symbol short name of this package. */ /* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "opus" #define PACKAGE_TARNAME "opus"
@ -161,7 +161,7 @@
#define PACKAGE_URL "" #define PACKAGE_URL ""
/* Define to the version of this package. */ /* Define to the version of this package. */
#define PACKAGE_VERSION "1.1.2" #define PACKAGE_VERSION "1.3.1"
/* Define to 1 if you have the ANSI C header files. */ /* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1 #define STDC_HEADERS 1

View File

@ -153,7 +153,7 @@
#define PACKAGE_NAME "opus" #define PACKAGE_NAME "opus"
/* Define to the full name and version of this package. */ /* Define to the full name and version of this package. */
#define PACKAGE_STRING "opus 1.1.2" #define PACKAGE_STRING "opus 1.3.1"
/* Define to the one symbol short name of this package. */ /* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "opus" #define PACKAGE_TARNAME "opus"
@ -162,7 +162,7 @@
#define PACKAGE_URL "" #define PACKAGE_URL ""
/* Define to the version of this package. */ /* Define to the version of this package. */
#define PACKAGE_VERSION "1.1.2" #define PACKAGE_VERSION "1.3.1"
/* Define to 1 if you have the ANSI C header files. */ /* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1 #define STDC_HEADERS 1

View File

@ -132,7 +132,7 @@
#define PACKAGE_NAME "opus" #define PACKAGE_NAME "opus"
/* Define to the full name and version of this package. */ /* Define to the full name and version of this package. */
#define PACKAGE_STRING "opus 1.1.1-beta-34-g5b20cb0" #define PACKAGE_STRING "opus 1.3.1"
/* Define to the one symbol short name of this package. */ /* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "opus" #define PACKAGE_TARNAME "opus"
@ -141,7 +141,7 @@
#define PACKAGE_URL "" #define PACKAGE_URL ""
/* Define to the version of this package. */ /* Define to the version of this package. */
#define PACKAGE_VERSION "1.1.1-beta-34-g5b20cb0" #define PACKAGE_VERSION "1.3.1"
/* Define to 1 if you have the ANSI C header files. */ /* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1 #define STDC_HEADERS 1

View File

@ -152,7 +152,7 @@
#define PACKAGE_NAME "opus" #define PACKAGE_NAME "opus"
/* Define to the full name and version of this package. */ /* Define to the full name and version of this package. */
#define PACKAGE_STRING "opus 1.1.2" #define PACKAGE_STRING "opus 1.3.1"
/* Define to the one symbol short name of this package. */ /* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "opus" #define PACKAGE_TARNAME "opus"
@ -161,7 +161,7 @@
#define PACKAGE_URL "" #define PACKAGE_URL ""
/* Define to the version of this package. */ /* Define to the version of this package. */
#define PACKAGE_VERSION "1.1.2" #define PACKAGE_VERSION "1.3.1"
/* Define to 1 if you have the ANSI C header files. */ /* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1 #define STDC_HEADERS 1

View File

@ -128,7 +128,7 @@
#define PACKAGE_NAME "opus" #define PACKAGE_NAME "opus"
/* Define to the full name and version of this package. */ /* Define to the full name and version of this package. */
#define PACKAGE_STRING "opus unknown" #define PACKAGE_STRING "opus 1.3.1"
/* Define to the one symbol short name of this package. */ /* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "opus" #define PACKAGE_TARNAME "opus"
@ -137,7 +137,7 @@
#define PACKAGE_URL "" #define PACKAGE_URL ""
/* Define to the version of this package. */ /* Define to the version of this package. */
#define PACKAGE_VERSION "unknown" #define PACKAGE_VERSION "1.3.1"
/* Define to 1 if you have the ANSI C header files. */ /* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1 #define STDC_HEADERS 1

View File

@ -129,7 +129,7 @@
#define PACKAGE_NAME "opus" #define PACKAGE_NAME "opus"
/* Define to the full name and version of this package. */ /* Define to the full name and version of this package. */
#define PACKAGE_STRING "opus 1.1.1-beta-34-g5b20cb0" #define PACKAGE_STRING "opus 1.3.1"
/* Define to the one symbol short name of this package. */ /* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "opus" #define PACKAGE_TARNAME "opus"
@ -138,7 +138,7 @@
#define PACKAGE_URL "" #define PACKAGE_URL ""
/* Define to the version of this package. */ /* Define to the version of this package. */
#define PACKAGE_VERSION "1.1.1-beta-34-g5b20cb0" #define PACKAGE_VERSION "1.3.1"
/* Define to 1 if you have the ANSI C header files. */ /* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1 #define STDC_HEADERS 1

View File

@ -128,7 +128,7 @@
#define PACKAGE_NAME "opus" #define PACKAGE_NAME "opus"
/* Define to the full name and version of this package. */ /* Define to the full name and version of this package. */
#define PACKAGE_STRING "opus 1.1.1-beta-34-g5b20cb0" #define PACKAGE_STRING "opus 1.3.1"
/* Define to the one symbol short name of this package. */ /* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "opus" #define PACKAGE_TARNAME "opus"
@ -137,7 +137,7 @@
#define PACKAGE_URL "" #define PACKAGE_URL ""
/* Define to the version of this package. */ /* Define to the version of this package. */
#define PACKAGE_VERSION "1.1.1-beta-34-g5b20cb0" #define PACKAGE_VERSION "1.3.1"
/* Define to 1 if you have the ANSI C header files. */ /* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1 #define STDC_HEADERS 1

View File

@ -28,8 +28,6 @@ POSSIBILITY OF SUCH DAMAGE.
#ifndef CONFIG_H #ifndef CONFIG_H
#define CONFIG_H #define CONFIG_H
#define inline __inline
#define USE_ALLOCA 1 #define USE_ALLOCA 1
/* Comment out the next line for floating-point code */ /* Comment out the next line for floating-point code */
@ -37,16 +35,30 @@ POSSIBILITY OF SUCH DAMAGE.
#define OPUS_BUILD 1 #define OPUS_BUILD 1
/* Get rid of the CELT VS compile warnings */ #if defined(_M_IX86) || defined(_M_X64)
#if 1 /* Can always compile SSE intrinsics (no special compiler flags necessary) */
#pragma warning(disable : 4996)/* This function or variable may be unsafe. Consider using fopen_s instead. To disable deprecation, use _CRT_SECURE_NO_WARNINGS. See online help for details. */ /* #define OPUS_X86_MAY_HAVE_SSE
#define OPUS_X86_MAY_HAVE_SSE2
#define OPUS_X86_MAY_HAVE_SSE4_1 */
/* Presume SSE functions, if compiled to use SSE/SSE2/AVX (note that AMD64 implies SSE2, and AVX
implies SSE4.1) */
#if defined(_M_X64) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 1)) || defined(__AVX__)
#define OPUS_X86_PRESUME_SSE 1
#endif
#if defined(_M_X64) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 2)) || defined(__AVX__)
#define OPUS_X86_PRESUME_SSE2 1
#endif
#if defined(__AVX__)
#define OPUS_X86_PRESUME_SSE4_1 1
#endif #endif
/* Enable SSE functions, if compiled with SSE/SSE2 (note that AMD64 implies SSE2) */ #if !defined(OPUS_X86_PRESUME_SSE4_1) || !defined(OPUS_X86_PRESUME_SSE2) || !defined(OPUS_X86_PRESUME_SSE)
#if defined(_M_X64) || (defined(_M_IX86_FP) && (_M_IX86_FP >= 1)) #define OPUS_HAVE_RTCD 1
#define __SSE__ 1
#endif #endif
#include "version.h" #endif
#define PACKAGE_VERSION "1.3.1"
#endif /* CONFIG_H */ #endif /* CONFIG_H */

411
node_modules/@discordjs/opus/deps/opus/CMakeLists.txt generated vendored Normal file
View File

@ -0,0 +1,411 @@
cmake_minimum_required(VERSION 3.1)
include(opus_functions.cmake)
get_library_version(OPUS_LIBRARY_VERSION OPUS_LIBRARY_VERSION_MAJOR)
message(STATUS "Opus library version: ${OPUS_LIBRARY_VERSION}")
get_package_version(PACKAGE_VERSION)
message(STATUS "Opus package version: ${PACKAGE_VERSION}")
string(REGEX
REPLACE "^([0-9]+.[0-9]+\\.?([0-9]+)?).*"
"\\1"
PROJECT_VERSION
${PACKAGE_VERSION})
message(STATUS "Opus project version: ${PROJECT_VERSION}")
project(Opus LANGUAGES C VERSION ${PROJECT_VERSION})
include(opus_buildtype.cmake)
option(OPUS_STACK_PROTECTOR "Use stack protection" ON)
option(OPUS_USE_ALLOCA "Use alloca for stack arrays (on non-C99 compilers)" OFF)
option(OPUS_CUSTOM_MODES "Enable non-Opus modes, e.g. 44.1 kHz & 2^n frames"
OFF)
option(OPUS_BUILD_PROGRAMS "Build programs" OFF)
option(OPUS_FIXED_POINT
"Compile as fixed-point (for machines without a fast enough FPU)" OFF)
option(OPUS_ENABLE_FLOAT_API
"Compile with the floating point API (for machines with float library"
ON)
option(OPUS_INSTALL_PKG_CONFIG_MODULE "Install PkgConfig module" ON)
option(OPUS_INSTALL_CMAKE_CONFIG_MODULE "Install CMake package config module"
ON)
include(opus_config.cmake)
include(opus_sources.cmake)
include(GNUInstallDirs)
include(CMakeDependentOption)
include(FeatureSummary)
if(OPUS_STACK_PROTECTOR)
if(NOT MSVC) # GC on by default on MSVC
check_and_set_flag(STACK_PROTECTION_STRONG -fstack-protector-strong)
endif()
else()
if(MSVC)
check_and_set_flag(BUFFER_SECURITY_CHECK /GS-)
endif()
endif()
if(OPUS_CPU_X86 OR OPUS_CPU_X64)
cmake_dependent_option(OPUS_X86_MAY_HAVE_SSE
"Does runtime check for SSE1 support"
ON
"SSE1_SUPPORTED"
OFF)
cmake_dependent_option(OPUS_X86_MAY_HAVE_SSE2
"Does runtime check for SSE2 support"
ON
"SSE2_SUPPORTED"
OFF)
cmake_dependent_option(OPUS_X86_MAY_HAVE_SSE4_1
"Does runtime check for SSE4.1 support"
ON
"SSE4_1_SUPPORTED"
OFF)
cmake_dependent_option(OPUS_X86_MAY_HAVE_AVX
"Does runtime check for AVX support"
ON
"AVX_SUPPORTED"
OFF)
if(OPUS_CPU_X64) # Assume 64 bit has SSE2 support
cmake_dependent_option(OPUS_X86_PRESUME_SSE
"Assume target CPU has SSE1 support"
ON
"OPUS_X86_MAY_HAVE_SSE"
OFF)
cmake_dependent_option(OPUS_X86_PRESUME_SSE2
"Assume target CPU has SSE2 support"
ON
"OPUS_X86_MAY_HAVE_SSE2"
OFF)
else()
cmake_dependent_option(OPUS_X86_PRESUME_SSE
"Assume target CPU has SSE1 support"
OFF
"OPUS_X86_MAY_HAVE_SSE"
OFF)
cmake_dependent_option(OPUS_X86_PRESUME_SSE2
"Assume target CPU has SSE2 support"
OFF
"OPUS_X86_MAY_HAVE_SSE2"
OFF)
endif()
cmake_dependent_option(OPUS_X86_PRESUME_SSE4_1
"Assume target CPU has SSE4.1 support"
OFF
"OPUS_X86_MAY_HAVE_SSE4_1"
OFF)
cmake_dependent_option(OPUS_X86_PRESUME_AVX
"Assume target CPU has AVX support"
OFF
"OPUS_X86_MAY_HAVE_AVX"
OFF)
endif()
set_package_properties(Git
PROPERTIES
TYPE
REQUIRED
DESCRIPTION
"fast, scalable, distributed revision control system"
URL
"https://git-scm.com/"
PURPOSE
"required to set up package version")
add_feature_info(STACK_PROTECTOR OPUS_STACK_PROTECTOR "Use stack protection")
add_feature_info(USE_ALLOCA OPUS_USE_ALLOCA
"Use alloca for stack arrays (on non-C99 compilers)")
add_feature_info(CUSTOM_MODES OPUS_CUSTOM_MODES
"Enable non-Opus modes, e.g. 44.1 kHz & 2^n frames")
add_feature_info(BUILD_PROGRAMS OPUS_BUILD_PROGRAMS "Build programs")
add_feature_info(
FIXED_POINT OPUS_FIXED_POINT
"compile as fixed-point (for machines without a fast enough FPU)")
add_feature_info(
FLOAT_API OPUS_ENABLE_FLOAT_API
"compile with the floating point API (for machines with float library)")
add_feature_info(INSTALL_PKG_CONFIG_MODULE OPUS_INSTALL_PKG_CONFIG_MODULE
"install PkgConfig module")
add_feature_info(INSTALL_CMAKE_CONFIG_MODULE OPUS_INSTALL_CMAKE_CONFIG_MODULE
"install CMake package config module")
if(OPUS_CPU_X86 OR OPUS_CPU_X64)
add_feature_info(X86_MAY_HAVE_SSE OPUS_X86_MAY_HAVE_SSE
"does runtime check for SSE1 support")
add_feature_info(X86_MAY_HAVE_SSE2 OPUS_X86_MAY_HAVE_SSE2
"does runtime check for SSE2 support")
add_feature_info(X86_MAY_HAVE_SSE4_1 OPUS_X86_MAY_HAVE_SSE4_1
"does runtime check for SSE4_1 support")
add_feature_info(X86_MAY_HAVE_AVX OPUS_X86_MAY_HAVE_AVX
"does runtime check for AVX support")
add_feature_info(X86_PRESUME_SSE OPUS_X86_PRESUME_SSE
"assume target CPU has SSE1 support")
add_feature_info(X86_PRESUME_SSE2 OPUS_X86_PRESUME_SSE2
"assume target CPU has SSE2 support")
add_feature_info(X86_PRESUME_SSE4_1 OPUS_X86_PRESUME_SSE4_1
"assume target CPU has SSE4_1 support")
add_feature_info(X86_PRESUME_AVX OPUS_X86_PRESUME_AVX
"assume target CPU has AVX support")
endif()
feature_summary(WHAT ALL)
add_library(opus ${opus_sources} ${opus_sources_float})
set(Opus_PUBLIC_HEADER
${CMAKE_CURRENT_SOURCE_DIR}/include/opus.h
${CMAKE_CURRENT_SOURCE_DIR}/include/opus_custom.h
${CMAKE_CURRENT_SOURCE_DIR}/include/opus_defines.h
${CMAKE_CURRENT_SOURCE_DIR}/include/opus_multistream.h
${CMAKE_CURRENT_SOURCE_DIR}/include/opus_projection.h
${CMAKE_CURRENT_SOURCE_DIR}/include/opus_types.h)
set_target_properties(opus
PROPERTIES SOVERSION
${OPUS_LIBRARY_VERSION_MAJOR}
VERSION
${OPUS_LIBRARY_VERSION}
PUBLIC_HEADER
"${Opus_PUBLIC_HEADER}")
target_include_directories(
opus
PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>
$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>
PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
${CMAKE_CURRENT_SOURCE_DIR}
celt
silk)
target_link_libraries(opus PRIVATE ${OPUS_REQUIRED_LIBRARIES})
target_compile_definitions(opus PRIVATE OPUS_BUILD ENABLE_HARDENING)
if(NOT MSVC)
target_compile_definitions(opus PRIVATE FORTIFY_SOURCE=2)
endif()
# It is strongly recommended to uncomment one of these VAR_ARRAYS: Use C99
# variable-length arrays for stack allocation USE_ALLOCA: Use alloca() for stack
# allocation If none is defined, then the fallback is a non-threadsafe global
# array
if(OPUS_USE_ALLOCA OR MSVC)
target_compile_definitions(opus PRIVATE USE_ALLOCA)
else()
target_compile_definitions(opus PRIVATE VAR_ARRAYS)
endif()
if(OPUS_CUSTOM_MODES)
target_compile_definitions(opus PRIVATE CUSTOM_MODES)
endif()
if(BUILD_SHARED_LIBS)
if(WIN32)
target_compile_definitions(opus PRIVATE DLL_EXPORT)
else()
include(CheckCCompilerFlag)
check_c_compiler_flag(-fvisibility=hidden COMPILER_HAS_HIDDEN_VISIBILITY)
if(COMPILER_HAS_HIDDEN_VISIBILITY)
set_target_properties(opus PROPERTIES C_VISIBILITY_PRESET hidden)
endif()
endif()
endif()
add_sources_group(opus silk ${silk_sources})
add_sources_group(opus celt ${celt_sources})
if(OPUS_FIXED_POINT)
add_sources_group(opus silk ${silk_sources_fixed})
target_include_directories(opus PRIVATE silk/fixed)
target_compile_definitions(opus PRIVATE FIXED_POINT=1)
else()
add_sources_group(opus silk ${silk_sources_float})
target_include_directories(opus PRIVATE silk/float)
endif()
if(NOT OPUS_ENABLE_FLOAT_API)
target_compile_definitions(opus PRIVATE DISABLE_FLOAT_API)
endif()
if(OPUS_X86_MAY_HAVE_SSE
OR OPUS_X86_MAY_HAVE_SSE2
OR OPUS_X86_MAY_HAVE_SSE4_1
OR OPUS_X86_MAY_HAVE_AVX)
target_compile_definitions(opus PRIVATE OPUS_HAVE_RTCD)
endif()
if(OPUS_X86_MAY_HAVE_SSE)
add_sources_group(opus celt ${celt_sources_sse})
target_compile_definitions(opus PRIVATE OPUS_X86_MAY_HAVE_SSE)
endif()
if(OPUS_X86_PRESUME_SSE)
target_compile_definitions(opus PRIVATE OPUS_X86_PRESUME_SSE)
endif()
if(OPUS_X86_MAY_HAVE_SSE2)
add_sources_group(opus celt ${celt_sources_sse2})
target_compile_definitions(opus PRIVATE OPUS_X86_MAY_HAVE_SSE2)
endif()
if(OPUS_X86_PRESUME_SSE2)
target_compile_definitions(opus PRIVATE OPUS_X86_PRESUME_SSE2)
endif()
if(OPUS_X86_MAY_HAVE_SSE)
add_sources_group(opus celt ${celt_sources_sse4_1})
add_sources_group(opus silk ${silk_sources_sse4_1})
if(OPUS_FIXED_POINT)
add_sources_group(opus silk ${silk_sources_fixed_sse4_1})
endif()
target_compile_definitions(opus PRIVATE OPUS_X86_MAY_HAVE_SSE4_1)
endif()
if(OPUS_X86_PRESUME_SSE4_1)
target_compile_definitions(opus PRIVATE OPUS_X86_PRESUME_SSE4_1)
endif()
if(CMAKE_SYSTEM_PROCESSOR MATCHES "(armv7-a)")
add_sources_group(opus celt ${celt_sources_arm})
endif()
if(COMPILER_SUPPORT_NEON AND OPUS_USE_NEON)
if(OPUS_MAY_HAVE_NEON)
if(RUNTIME_CPU_CAPABILITY_DETECTION)
message(STATUS "OPUS_MAY_HAVE_NEON enabling runtime detection")
target_compile_definitions(opus PRIVATE OPUS_HAVE_RTCD)
else()
message(ERROR "Runtime cpu capability detection needed for MAY_HAVE_NEON")
endif()
# Do runtime check for NEON
target_compile_definitions(opus
PRIVATE
OPUS_ARM_MAY_HAVE_NEON
OPUS_ARM_MAY_HAVE_NEON_INTR)
endif()
add_sources_group(opus celt ${celt_sources_arm_neon_intr})
add_sources_group(opus silk ${silk_sources_arm_neon_intr})
# silk arm neon depends on main_Fix.h
target_include_directories(opus PRIVATE silk/fixed)
if(OPUS_FIXED_POINT)
add_sources_group(opus silk ${silk_sources_fixed_arm_neon_intr})
endif()
if(OPUS_PRESUME_NEON)
target_compile_definitions(opus
PRIVATE
OPUS_ARM_PRESUME_NEON
OPUS_ARM_PRESUME_NEON_INTR)
endif()
endif()
install(TARGETS opus
EXPORT OpusTargets
ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR}
PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/opus)
if(OPUS_INSTALL_PKG_CONFIG_MODULE)
set(prefix ${CMAKE_INSTALL_PREFIX})
set(exec_prefix ${CMAKE_INSTALL_PREFIX})
set(libdir ${CMAKE_INSTALL_FULL_LIBDIR})
set(includedir ${CMAKE_INSTALL_FULL_INCLUDEDIR})
set(VERSION ${OPUS_LIBRARY_VERSION})
set(VERSION ${OPUS_LIBRARY_VERSION})
if(HAVE_LIBM)
set(LIBM "-lm")
endif()
configure_file(opus.pc.in opus.pc)
install(FILES ${CMAKE_CURRENT_BINARY_DIR}/opus.pc
DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig)
endif()
if(OPUS_INSTALL_CMAKE_CONFIG_MODULE)
set(CMAKE_INSTALL_PACKAGEDIR ${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME})
install(EXPORT OpusTargets
NAMESPACE Opus::
DESTINATION ${CMAKE_INSTALL_PACKAGEDIR})
include(CMakePackageConfigHelpers)
set(INCLUDE_INSTALL_DIR ${CMAKE_INSTALL_INCLUDEDIR})
configure_package_config_file(OpusConfig.cmake.in
OpusConfig.cmake
INSTALL_DESTINATION
${CMAKE_INSTALL_PACKAGEDIR}
PATH_VARS
INCLUDE_INSTALL_DIR
INSTALL_PREFIX
${CMAKE_INSTALL_PREFIX})
write_basic_package_version_file(OpusConfigVersion.cmake
VERSION ${PROJECT_VERSION}
COMPATIBILITY SameMajorVersion)
install(FILES ${CMAKE_CURRENT_BINARY_DIR}/OpusConfig.cmake
${CMAKE_CURRENT_BINARY_DIR}/OpusConfigVersion.cmake
DESTINATION ${CMAKE_INSTALL_PACKAGEDIR})
endif()
if(OPUS_BUILD_PROGRAMS)
# demo
if(OPUS_CUSTOM_MODES)
add_executable(opus_custom_demo ${opus_custom_demo_sources})
target_include_directories(opus_custom_demo
PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
target_link_libraries(opus_custom_demo PRIVATE opus)
endif()
add_executable(opus_demo ${opus_demo_sources})
target_include_directories(opus_demo PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
target_include_directories(opus_demo PRIVATE silk) # debug.h
target_include_directories(opus_demo PRIVATE celt) # arch.h
target_link_libraries(opus_demo PRIVATE opus)
# compare
add_executable(opus_compare ${opus_compare_sources})
target_include_directories(opus_compare PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
target_link_libraries(opus_compare PRIVATE opus)
endif()
if(BUILD_TESTING)
enable_testing()
# tests
add_executable(test_opus_decode ${test_opus_decode_sources})
target_include_directories(test_opus_decode
PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
target_link_libraries(test_opus_decode PRIVATE opus)
if(OPUS_FIXED_POINT)
target_compile_definitions(test_opus_decode PRIVATE DISABLE_FLOAT_API)
endif()
add_test(test_opus_decode test_opus_decode)
add_executable(test_opus_padding ${test_opus_padding_sources})
target_include_directories(test_opus_padding
PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
target_link_libraries(test_opus_padding PRIVATE opus)
add_test(test_opus_padding test_opus_padding)
if(NOT BUILD_SHARED_LIBS)
# disable tests that depends on private API when building shared lib
add_executable(test_opus_api ${test_opus_api_sources})
target_include_directories(test_opus_api
PRIVATE ${CMAKE_CURRENT_BINARY_DIR} celt)
target_link_libraries(test_opus_api PRIVATE opus)
if(OPUS_FIXED_POINT)
target_compile_definitions(test_opus_api PRIVATE DISABLE_FLOAT_API)
endif()
add_test(test_opus_api test_opus_api)
add_executable(test_opus_encode ${test_opus_encode_sources})
target_include_directories(test_opus_encode
PRIVATE ${CMAKE_CURRENT_BINARY_DIR} celt)
target_link_libraries(test_opus_encode PRIVATE opus)
add_test(test_opus_encode test_opus_encode)
endif()
endif()

368
node_modules/@discordjs/opus/deps/opus/INSTALL generated vendored Normal file
View File

@ -0,0 +1,368 @@
Installation Instructions
*************************
Copyright (C) 1994-1996, 1999-2002, 2004-2016 Free Software
Foundation, Inc.
Copying and distribution of this file, with or without modification,
are permitted in any medium without royalty provided the copyright
notice and this notice are preserved. This file is offered as-is,
without warranty of any kind.
Basic Installation
==================
Briefly, the shell command './configure && make && make install'
should configure, build, and install this package. The following
more-detailed instructions are generic; see the 'README' file for
instructions specific to this package. Some packages provide this
'INSTALL' file but do not implement all of the features documented
below. The lack of an optional feature in a given package is not
necessarily a bug. More recommendations for GNU packages can be found
in *note Makefile Conventions: (standards)Makefile Conventions.
The 'configure' shell script attempts to guess correct values for
various system-dependent variables used during compilation. It uses
those values to create a 'Makefile' in each directory of the package.
It may also create one or more '.h' files containing system-dependent
definitions. Finally, it creates a shell script 'config.status' that
you can run in the future to recreate the current configuration, and a
file 'config.log' containing compiler output (useful mainly for
debugging 'configure').
It can also use an optional file (typically called 'config.cache' and
enabled with '--cache-file=config.cache' or simply '-C') that saves the
results of its tests to speed up reconfiguring. Caching is disabled by
default to prevent problems with accidental use of stale cache files.
If you need to do unusual things to compile the package, please try
to figure out how 'configure' could check whether to do them, and mail
diffs or instructions to the address given in the 'README' so they can
be considered for the next release. If you are using the cache, and at
some point 'config.cache' contains results you don't want to keep, you
may remove or edit it.
The file 'configure.ac' (or 'configure.in') is used to create
'configure' by a program called 'autoconf'. You need 'configure.ac' if
you want to change it or regenerate 'configure' using a newer version of
'autoconf'.
The simplest way to compile this package is:
1. 'cd' to the directory containing the package's source code and type
'./configure' to configure the package for your system.
Running 'configure' might take a while. While running, it prints
some messages telling which features it is checking for.
2. Type 'make' to compile the package.
3. Optionally, type 'make check' to run any self-tests that come with
the package, generally using the just-built uninstalled binaries.
4. Type 'make install' to install the programs and any data files and
documentation. When installing into a prefix owned by root, it is
recommended that the package be configured and built as a regular
user, and only the 'make install' phase executed with root
privileges.
5. Optionally, type 'make installcheck' to repeat any self-tests, but
this time using the binaries in their final installed location.
This target does not install anything. Running this target as a
regular user, particularly if the prior 'make install' required
root privileges, verifies that the installation completed
correctly.
6. You can remove the program binaries and object files from the
source code directory by typing 'make clean'. To also remove the
files that 'configure' created (so you can compile the package for
a different kind of computer), type 'make distclean'. There is
also a 'make maintainer-clean' target, but that is intended mainly
for the package's developers. If you use it, you may have to get
all sorts of other programs in order to regenerate files that came
with the distribution.
7. Often, you can also type 'make uninstall' to remove the installed
files again. In practice, not all packages have tested that
uninstallation works correctly, even though it is required by the
GNU Coding Standards.
8. Some packages, particularly those that use Automake, provide 'make
distcheck', which can by used by developers to test that all other
targets like 'make install' and 'make uninstall' work correctly.
This target is generally not run by end users.
Compilers and Options
=====================
Some systems require unusual options for compilation or linking that
the 'configure' script does not know about. Run './configure --help'
for details on some of the pertinent environment variables.
You can give 'configure' initial values for configuration parameters
by setting variables in the command line or in the environment. Here is
an example:
./configure CC=c99 CFLAGS=-g LIBS=-lposix
*Note Defining Variables::, for more details.
Compiling For Multiple Architectures
====================================
You can compile the package for more than one kind of computer at the
same time, by placing the object files for each architecture in their
own directory. To do this, you can use GNU 'make'. 'cd' to the
directory where you want the object files and executables to go and run
the 'configure' script. 'configure' automatically checks for the source
code in the directory that 'configure' is in and in '..'. This is known
as a "VPATH" build.
With a non-GNU 'make', it is safer to compile the package for one
architecture at a time in the source code directory. After you have
installed the package for one architecture, use 'make distclean' before
reconfiguring for another architecture.
On MacOS X 10.5 and later systems, you can create libraries and
executables that work on multiple system types--known as "fat" or
"universal" binaries--by specifying multiple '-arch' options to the
compiler but only a single '-arch' option to the preprocessor. Like
this:
./configure CC="gcc -arch i386 -arch x86_64 -arch ppc -arch ppc64" \
CXX="g++ -arch i386 -arch x86_64 -arch ppc -arch ppc64" \
CPP="gcc -E" CXXCPP="g++ -E"
This is not guaranteed to produce working output in all cases, you
may have to build one architecture at a time and combine the results
using the 'lipo' tool if you have problems.
Installation Names
==================
By default, 'make install' installs the package's commands under
'/usr/local/bin', include files under '/usr/local/include', etc. You
can specify an installation prefix other than '/usr/local' by giving
'configure' the option '--prefix=PREFIX', where PREFIX must be an
absolute file name.
You can specify separate installation prefixes for
architecture-specific files and architecture-independent files. If you
pass the option '--exec-prefix=PREFIX' to 'configure', the package uses
PREFIX as the prefix for installing programs and libraries.
Documentation and other data files still use the regular prefix.
In addition, if you use an unusual directory layout you can give
options like '--bindir=DIR' to specify different values for particular
kinds of files. Run 'configure --help' for a list of the directories
you can set and what kinds of files go in them. In general, the default
for these options is expressed in terms of '${prefix}', so that
specifying just '--prefix' will affect all of the other directory
specifications that were not explicitly provided.
The most portable way to affect installation locations is to pass the
correct locations to 'configure'; however, many packages provide one or
both of the following shortcuts of passing variable assignments to the
'make install' command line to change installation locations without
having to reconfigure or recompile.
The first method involves providing an override variable for each
affected directory. For example, 'make install
prefix=/alternate/directory' will choose an alternate location for all
directory configuration variables that were expressed in terms of
'${prefix}'. Any directories that were specified during 'configure',
but not in terms of '${prefix}', must each be overridden at install time
for the entire installation to be relocated. The approach of makefile
variable overrides for each directory variable is required by the GNU
Coding Standards, and ideally causes no recompilation. However, some
platforms have known limitations with the semantics of shared libraries
that end up requiring recompilation when using this method, particularly
noticeable in packages that use GNU Libtool.
The second method involves providing the 'DESTDIR' variable. For
example, 'make install DESTDIR=/alternate/directory' will prepend
'/alternate/directory' before all installation names. The approach of
'DESTDIR' overrides is not required by the GNU Coding Standards, and
does not work on platforms that have drive letters. On the other hand,
it does better at avoiding recompilation issues, and works well even
when some directory options were not specified in terms of '${prefix}'
at 'configure' time.
Optional Features
=================
If the package supports it, you can cause programs to be installed
with an extra prefix or suffix on their names by giving 'configure' the
option '--program-prefix=PREFIX' or '--program-suffix=SUFFIX'.
Some packages pay attention to '--enable-FEATURE' options to
'configure', where FEATURE indicates an optional part of the package.
They may also pay attention to '--with-PACKAGE' options, where PACKAGE
is something like 'gnu-as' or 'x' (for the X Window System). The
'README' should mention any '--enable-' and '--with-' options that the
package recognizes.
For packages that use the X Window System, 'configure' can usually
find the X include and library files automatically, but if it doesn't,
you can use the 'configure' options '--x-includes=DIR' and
'--x-libraries=DIR' to specify their locations.
Some packages offer the ability to configure how verbose the
execution of 'make' will be. For these packages, running './configure
--enable-silent-rules' sets the default to minimal output, which can be
overridden with 'make V=1'; while running './configure
--disable-silent-rules' sets the default to verbose, which can be
overridden with 'make V=0'.
Particular systems
==================
On HP-UX, the default C compiler is not ANSI C compatible. If GNU CC
is not installed, it is recommended to use the following options in
order to use an ANSI C compiler:
./configure CC="cc -Ae -D_XOPEN_SOURCE=500"
and if that doesn't work, install pre-built binaries of GCC for HP-UX.
HP-UX 'make' updates targets which have the same time stamps as their
prerequisites, which makes it generally unusable when shipped generated
files such as 'configure' are involved. Use GNU 'make' instead.
On OSF/1 a.k.a. Tru64, some versions of the default C compiler cannot
parse its '<wchar.h>' header file. The option '-nodtk' can be used as a
workaround. If GNU CC is not installed, it is therefore recommended to
try
./configure CC="cc"
and if that doesn't work, try
./configure CC="cc -nodtk"
On Solaris, don't put '/usr/ucb' early in your 'PATH'. This
directory contains several dysfunctional programs; working variants of
these programs are available in '/usr/bin'. So, if you need '/usr/ucb'
in your 'PATH', put it _after_ '/usr/bin'.
On Haiku, software installed for all users goes in '/boot/common',
not '/usr/local'. It is recommended to use the following options:
./configure --prefix=/boot/common
Specifying the System Type
==========================
There may be some features 'configure' cannot figure out
automatically, but needs to determine by the type of machine the package
will run on. Usually, assuming the package is built to be run on the
_same_ architectures, 'configure' can figure that out, but if it prints
a message saying it cannot guess the machine type, give it the
'--build=TYPE' option. TYPE can either be a short name for the system
type, such as 'sun4', or a canonical name which has the form:
CPU-COMPANY-SYSTEM
where SYSTEM can have one of these forms:
OS
KERNEL-OS
See the file 'config.sub' for the possible values of each field. If
'config.sub' isn't included in this package, then this package doesn't
need to know the machine type.
If you are _building_ compiler tools for cross-compiling, you should
use the option '--target=TYPE' to select the type of system they will
produce code for.
If you want to _use_ a cross compiler, that generates code for a
platform different from the build platform, you should specify the
"host" platform (i.e., that on which the generated programs will
eventually be run) with '--host=TYPE'.
Sharing Defaults
================
If you want to set default values for 'configure' scripts to share,
you can create a site shell script called 'config.site' that gives
default values for variables like 'CC', 'cache_file', and 'prefix'.
'configure' looks for 'PREFIX/share/config.site' if it exists, then
'PREFIX/etc/config.site' if it exists. Or, you can set the
'CONFIG_SITE' environment variable to the location of the site script.
A warning: not all 'configure' scripts look for a site script.
Defining Variables
==================
Variables not defined in a site shell script can be set in the
environment passed to 'configure'. However, some packages may run
configure again during the build, and the customized values of these
variables may be lost. In order to avoid this problem, you should set
them in the 'configure' command line, using 'VAR=value'. For example:
./configure CC=/usr/local2/bin/gcc
causes the specified 'gcc' to be used as the C compiler (unless it is
overridden in the site shell script).
Unfortunately, this technique does not work for 'CONFIG_SHELL' due to an
Autoconf limitation. Until the limitation is lifted, you can use this
workaround:
CONFIG_SHELL=/bin/bash ./configure CONFIG_SHELL=/bin/bash
'configure' Invocation
======================
'configure' recognizes the following options to control how it
operates.
'--help'
'-h'
Print a summary of all of the options to 'configure', and exit.
'--help=short'
'--help=recursive'
Print a summary of the options unique to this package's
'configure', and exit. The 'short' variant lists options used only
in the top level, while the 'recursive' variant lists options also
present in any nested packages.
'--version'
'-V'
Print the version of Autoconf used to generate the 'configure'
script, and exit.
'--cache-file=FILE'
Enable the cache: use and save the results of the tests in FILE,
traditionally 'config.cache'. FILE defaults to '/dev/null' to
disable caching.
'--config-cache'
'-C'
Alias for '--cache-file=config.cache'.
'--quiet'
'--silent'
'-q'
Do not print messages saying which checks are being made. To
suppress all normal output, redirect it to '/dev/null' (any error
messages will still be shown).
'--srcdir=DIR'
Look for the package's source code in directory DIR. Usually
'configure' can determine that directory automatically.
'--prefix=DIR'
Use DIR as the installation prefix. *note Installation Names:: for
more details, including other options available for fine-tuning the
installation locations.
'--no-create'
'-n'
Run the configure checks, but stop before creating any output
files.
'configure' also accepts some other, not widely useful, options. Run
'configure --help' for more details.

View File

@ -21,6 +21,9 @@ SILK_SOURCES += $(SILK_SOURCES_FIXED)
if HAVE_SSE4_1 if HAVE_SSE4_1
SILK_SOURCES += $(SILK_SOURCES_SSE4_1) $(SILK_SOURCES_FIXED_SSE4_1) SILK_SOURCES += $(SILK_SOURCES_SSE4_1) $(SILK_SOURCES_FIXED_SSE4_1)
endif endif
if HAVE_ARM_NEON_INTR
SILK_SOURCES += $(SILK_SOURCES_FIXED_ARM_NEON_INTR)
endif
else else
SILK_SOURCES += $(SILK_SOURCES_FLOAT) SILK_SOURCES += $(SILK_SOURCES_FLOAT)
if HAVE_SSE4_1 if HAVE_SSE4_1
@ -47,8 +50,9 @@ if CPU_ARM
CELT_SOURCES += $(CELT_SOURCES_ARM) CELT_SOURCES += $(CELT_SOURCES_ARM)
SILK_SOURCES += $(SILK_SOURCES_ARM) SILK_SOURCES += $(SILK_SOURCES_ARM)
if OPUS_ARM_NEON_INTR if HAVE_ARM_NEON_INTR
CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR) CELT_SOURCES += $(CELT_SOURCES_ARM_NEON_INTR)
SILK_SOURCES += $(SILK_SOURCES_ARM_NEON_INTR)
endif endif
if HAVE_ARM_NE10 if HAVE_ARM_NE10
@ -78,14 +82,43 @@ if OPUS_ARM_EXTERNAL_ASM
libopus_la_LIBADD += libarmasm.la libopus_la_LIBADD += libarmasm.la
endif endif
pkginclude_HEADERS = include/opus.h include/opus_multistream.h include/opus_types.h include/opus_defines.h pkginclude_HEADERS = include/opus.h include/opus_multistream.h include/opus_types.h include/opus_defines.h include/opus_projection.h
noinst_HEADERS = $(OPUS_HEAD) $(SILK_HEAD) $(CELT_HEAD) noinst_HEADERS = $(OPUS_HEAD) $(SILK_HEAD) $(CELT_HEAD)
if EXTRA_PROGRAMS if EXTRA_PROGRAMS
noinst_PROGRAMS = opus_demo repacketizer_demo opus_compare tests/test_opus_api tests/test_opus_encode tests/test_opus_decode tests/test_opus_padding celt/tests/test_unit_cwrs32 celt/tests/test_unit_dft celt/tests/test_unit_entropy celt/tests/test_unit_laplace celt/tests/test_unit_mathops celt/tests/test_unit_mdct celt/tests/test_unit_rotation celt/tests/test_unit_types noinst_PROGRAMS = celt/tests/test_unit_cwrs32 \
celt/tests/test_unit_dft \
celt/tests/test_unit_entropy \
celt/tests/test_unit_laplace \
celt/tests/test_unit_mathops \
celt/tests/test_unit_mdct \
celt/tests/test_unit_rotation \
celt/tests/test_unit_types \
opus_compare \
opus_demo \
repacketizer_demo \
silk/tests/test_unit_LPC_inv_pred_gain \
tests/test_opus_api \
tests/test_opus_decode \
tests/test_opus_encode \
tests/test_opus_padding \
tests/test_opus_projection
TESTS = celt/tests/test_unit_types celt/tests/test_unit_mathops celt/tests/test_unit_entropy celt/tests/test_unit_laplace celt/tests/test_unit_dft celt/tests/test_unit_mdct celt/tests/test_unit_rotation celt/tests/test_unit_cwrs32 tests/test_opus_api tests/test_opus_decode tests/test_opus_encode tests/test_opus_padding TESTS = celt/tests/test_unit_cwrs32 \
celt/tests/test_unit_dft \
celt/tests/test_unit_entropy \
celt/tests/test_unit_laplace \
celt/tests/test_unit_mathops \
celt/tests/test_unit_mdct \
celt/tests/test_unit_rotation \
celt/tests/test_unit_types \
silk/tests/test_unit_LPC_inv_pred_gain \
tests/test_opus_api \
tests/test_opus_decode \
tests/test_opus_encode \
tests/test_opus_padding \
tests/test_opus_projection
opus_demo_SOURCES = src/opus_demo.c opus_demo_SOURCES = src/opus_demo.c
@ -101,7 +134,7 @@ opus_compare_LDADD = $(LIBM)
tests_test_opus_api_SOURCES = tests/test_opus_api.c tests/test_opus_common.h tests_test_opus_api_SOURCES = tests/test_opus_api.c tests/test_opus_common.h
tests_test_opus_api_LDADD = libopus.la $(NE10_LIBS) $(LIBM) tests_test_opus_api_LDADD = libopus.la $(NE10_LIBS) $(LIBM)
tests_test_opus_encode_SOURCES = tests/test_opus_encode.c tests/test_opus_common.h tests_test_opus_encode_SOURCES = tests/test_opus_encode.c tests/opus_encode_regressions.c tests/test_opus_common.h
tests_test_opus_encode_LDADD = libopus.la $(NE10_LIBS) $(LIBM) tests_test_opus_encode_LDADD = libopus.la $(NE10_LIBS) $(LIBM)
tests_test_opus_decode_SOURCES = tests/test_opus_decode.c tests/test_opus_common.h tests_test_opus_decode_SOURCES = tests/test_opus_decode.c tests/test_opus_common.h
@ -110,11 +143,27 @@ tests_test_opus_decode_LDADD = libopus.la $(NE10_LIBS) $(LIBM)
tests_test_opus_padding_SOURCES = tests/test_opus_padding.c tests/test_opus_common.h tests_test_opus_padding_SOURCES = tests/test_opus_padding.c tests/test_opus_common.h
tests_test_opus_padding_LDADD = libopus.la $(NE10_LIBS) $(LIBM) tests_test_opus_padding_LDADD = libopus.la $(NE10_LIBS) $(LIBM)
CELT_OBJ = $(CELT_SOURCES:.c=.lo)
SILK_OBJ = $(SILK_SOURCES:.c=.lo)
OPUS_OBJ = $(OPUS_SOURCES:.c=.lo)
tests_test_opus_projection_SOURCES = tests/test_opus_projection.c tests/test_opus_common.h
tests_test_opus_projection_LDADD = $(OPUS_OBJ) $(SILK_OBJ) $(CELT_OBJ) $(NE10_LIBS) $(LIBM)
if OPUS_ARM_EXTERNAL_ASM
tests_test_opus_projection_LDADD += libarmasm.la
endif
silk_tests_test_unit_LPC_inv_pred_gain_SOURCES = silk/tests/test_unit_LPC_inv_pred_gain.c
silk_tests_test_unit_LPC_inv_pred_gain_LDADD = $(SILK_OBJ) $(CELT_OBJ) $(NE10_LIBS) $(LIBM)
if OPUS_ARM_EXTERNAL_ASM
silk_tests_test_unit_LPC_inv_pred_gain_LDADD += libarmasm.la
endif
celt_tests_test_unit_cwrs32_SOURCES = celt/tests/test_unit_cwrs32.c celt_tests_test_unit_cwrs32_SOURCES = celt/tests/test_unit_cwrs32.c
celt_tests_test_unit_cwrs32_LDADD = $(LIBM) celt_tests_test_unit_cwrs32_LDADD = $(LIBM)
celt_tests_test_unit_dft_SOURCES = celt/tests/test_unit_dft.c celt_tests_test_unit_dft_SOURCES = celt/tests/test_unit_dft.c
celt_tests_test_unit_dft_LDADD = $(NE10_LIBS) $(LIBM) celt_tests_test_unit_dft_LDADD = $(CELT_OBJ) $(NE10_LIBS) $(LIBM)
if OPUS_ARM_EXTERNAL_ASM if OPUS_ARM_EXTERNAL_ASM
celt_tests_test_unit_dft_LDADD += libarmasm.la celt_tests_test_unit_dft_LDADD += libarmasm.la
endif endif
@ -126,19 +175,19 @@ celt_tests_test_unit_laplace_SOURCES = celt/tests/test_unit_laplace.c
celt_tests_test_unit_laplace_LDADD = $(LIBM) celt_tests_test_unit_laplace_LDADD = $(LIBM)
celt_tests_test_unit_mathops_SOURCES = celt/tests/test_unit_mathops.c celt_tests_test_unit_mathops_SOURCES = celt/tests/test_unit_mathops.c
celt_tests_test_unit_mathops_LDADD = $(NE10_LIBS) $(LIBM) celt_tests_test_unit_mathops_LDADD = $(CELT_OBJ) $(NE10_LIBS) $(LIBM)
if OPUS_ARM_EXTERNAL_ASM if OPUS_ARM_EXTERNAL_ASM
celt_tests_test_unit_mathops_LDADD += libarmasm.la celt_tests_test_unit_mathops_LDADD += libarmasm.la
endif endif
celt_tests_test_unit_mdct_SOURCES = celt/tests/test_unit_mdct.c celt_tests_test_unit_mdct_SOURCES = celt/tests/test_unit_mdct.c
celt_tests_test_unit_mdct_LDADD = $(NE10_LIBS) $(LIBM) celt_tests_test_unit_mdct_LDADD = $(CELT_OBJ) $(NE10_LIBS) $(LIBM)
if OPUS_ARM_EXTERNAL_ASM if OPUS_ARM_EXTERNAL_ASM
celt_tests_test_unit_mdct_LDADD += libarmasm.la celt_tests_test_unit_mdct_LDADD += libarmasm.la
endif endif
celt_tests_test_unit_rotation_SOURCES = celt/tests/test_unit_rotation.c celt_tests_test_unit_rotation_SOURCES = celt/tests/test_unit_rotation.c
celt_tests_test_unit_rotation_LDADD = $(NE10_LIBS) $(LIBM) celt_tests_test_unit_rotation_LDADD = $(CELT_OBJ) $(NE10_LIBS) $(LIBM)
if OPUS_ARM_EXTERNAL_ASM if OPUS_ARM_EXTERNAL_ASM
celt_tests_test_unit_rotation_LDADD += libarmasm.la celt_tests_test_unit_rotation_LDADD += libarmasm.la
endif endif
@ -156,34 +205,32 @@ opus_custom_demo_LDADD = libopus.la $(LIBM)
endif endif
endif endif
EXTRA_DIST = version.mk \ EXTRA_DIST = opus.pc.in \
opus.pc.in \
opus-uninstalled.pc.in \ opus-uninstalled.pc.in \
opus.m4 \ opus.m4 \
Makefile.mips \ Makefile.mips \
Makefile.unix \ Makefile.unix \
CMakeLists.txt \
config.h.cmake.in \
opus_config.cmake \
opus_functions.cmake \
opus_sources.cmake \
OpusConfig.cmake.in \
tests/run_vectors.sh \ tests/run_vectors.sh \
celt/arm/arm2gnu.pl \ celt/arm/arm2gnu.pl \
celt/arm/celt_pitch_xcorr_arm.s \ celt/arm/celt_pitch_xcorr_arm.s \
win32/VS2010/silk_float.vcxproj \ win32/VS2015/opus.vcxproj \
win32/VS2010/celt.vcxproj.filters \ win32/VS2015/test_opus_encode.vcxproj.filters \
win32/VS2010/opus.vcxproj \ win32/VS2015/test_opus_encode.vcxproj \
win32/VS2010/silk_common.vcxproj.filters \ win32/VS2015/opus_demo.vcxproj \
win32/VS2010/silk_float.vcxproj.filters \ win32/VS2015/test_opus_api.vcxproj.filters \
win32/VS2010/test_opus_encode.vcxproj.filters \ win32/VS2015/test_opus_api.vcxproj \
win32/VS2010/silk_common.vcxproj \ win32/VS2015/test_opus_decode.vcxproj.filters \
win32/VS2010/test_opus_encode.vcxproj \ win32/VS2015/opus_demo.vcxproj.filters \
win32/VS2010/opus_demo.vcxproj \ win32/VS2015/opus.vcxproj.filters \
win32/VS2010/test_opus_api.vcxproj.filters \ win32/VS2015/test_opus_decode.vcxproj \
win32/VS2010/test_opus_api.vcxproj \ win32/VS2015/opus.sln \
win32/VS2010/test_opus_decode.vcxproj.filters \ win32/VS2015/common.props \
win32/VS2010/silk_fixed.vcxproj.filters \
win32/VS2010/opus_demo.vcxproj.filters \
win32/VS2010/silk_fixed.vcxproj \
win32/VS2010/opus.vcxproj.filters \
win32/VS2010/test_opus_decode.vcxproj \
win32/VS2010/celt.vcxproj \
win32/VS2010/opus.sln \
win32/genversion.bat \ win32/genversion.bat \
win32/config.h win32/config.h
@ -275,27 +322,30 @@ $(CELT_SOURCES_ARM_ASM:%.s=%-gnu.S): $(top_srcdir)/celt/arm/arm2gnu.pl
OPT_UNIT_TEST_OBJ = $(celt_tests_test_unit_mathops_SOURCES:.c=.o) \ OPT_UNIT_TEST_OBJ = $(celt_tests_test_unit_mathops_SOURCES:.c=.o) \
$(celt_tests_test_unit_rotation_SOURCES:.c=.o) \ $(celt_tests_test_unit_rotation_SOURCES:.c=.o) \
$(celt_tests_test_unit_mdct_SOURCES:.c=.o) \ $(celt_tests_test_unit_mdct_SOURCES:.c=.o) \
$(celt_tests_test_unit_dft_SOURCES:.c=.o) $(celt_tests_test_unit_dft_SOURCES:.c=.o) \
$(silk_tests_test_unit_LPC_inv_pred_gain_SOURCES:.c=.o)
if HAVE_SSE if HAVE_SSE
SSE_OBJ = $(CELT_SOURCES_SSE:.c=.lo) SSE_OBJ = $(CELT_SOURCES_SSE:.c=.lo)
$(SSE_OBJ) $(OPT_UNIT_TEST_OBJ): CFLAGS += $(OPUS_X86_SSE_CFLAGS) $(SSE_OBJ): CFLAGS += $(OPUS_X86_SSE_CFLAGS)
endif endif
if HAVE_SSE2 if HAVE_SSE2
SSE2_OBJ = $(CELT_SOURCES_SSE2:.c=.lo) SSE2_OBJ = $(CELT_SOURCES_SSE2:.c=.lo)
$(SSE2_OBJ) $(OPT_UNIT_TEST_OBJ): CFLAGS += $(OPUS_X86_SSE2_CFLAGS) $(SSE2_OBJ): CFLAGS += $(OPUS_X86_SSE2_CFLAGS)
endif endif
if HAVE_SSE4_1 if HAVE_SSE4_1
SSE4_1_OBJ = $(CELT_SOURCES_SSE4_1:.c=.lo) \ SSE4_1_OBJ = $(CELT_SOURCES_SSE4_1:.c=.lo) \
$(SILK_SOURCES_SSE4_1:.c=.lo) \ $(SILK_SOURCES_SSE4_1:.c=.lo) \
$(SILK_SOURCES_FIXED_SSE4_1:.c=.lo) $(SILK_SOURCES_FIXED_SSE4_1:.c=.lo)
$(SSE4_1_OBJ) $(OPT_UNIT_TEST_OBJ): CFLAGS += $(OPUS_X86_SSE4_1_CFLAGS) $(SSE4_1_OBJ): CFLAGS += $(OPUS_X86_SSE4_1_CFLAGS)
endif endif
if OPUS_ARM_NEON_INTR if HAVE_ARM_NEON_INTR
CELT_ARM_NEON_INTR_OBJ = $(CELT_SOURCES_ARM_NEON_INTR:.c=.lo) ARM_NEON_INTR_OBJ = $(CELT_SOURCES_ARM_NEON_INTR:.c=.lo) \
$(CELT_ARM_NEON_INTR_OBJ) $(OPT_UNIT_TEST_OBJ): CFLAGS += \ $(SILK_SOURCES_ARM_NEON_INTR:.c=.lo) \
$(SILK_SOURCES_FIXED_ARM_NEON_INTR:.c=.lo)
$(ARM_NEON_INTR_OBJ): CFLAGS += \
$(OPUS_ARM_NEON_INTR_CFLAGS) $(NE10_CFLAGS) $(OPUS_ARM_NEON_INTR_CFLAGS) $(NE10_CFLAGS)
endif endif

View File

@ -12,7 +12,7 @@ CFLAGS := -DUSE_ALLOCA $(CFLAGS)
# These options affect performance # These options affect performance
# HAVE_LRINTF: Use C99 intrinsics to speed up float-to-int conversion # HAVE_LRINTF: Use C99 intrinsics to speed up float-to-int conversion
#CFLAGS := -DHAVE_LRINTF $(CFLAGS) CFLAGS := -DHAVE_LRINTF $(CFLAGS)
###################### END OF OPTIONS ###################### ###################### END OF OPTIONS ######################
@ -99,7 +99,7 @@ TESTOPUSAPI_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSAPI_SRCS_C))
TESTOPUSDECODE_SRCS_C = tests/test_opus_decode.c TESTOPUSDECODE_SRCS_C = tests/test_opus_decode.c
TESTOPUSDECODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSDECODE_SRCS_C)) TESTOPUSDECODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSDECODE_SRCS_C))
TESTOPUSENCODE_SRCS_C = tests/test_opus_encode.c TESTOPUSENCODE_SRCS_C = tests/test_opus_encode.c tests/opus_encode_regressions.c
TESTOPUSENCODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSENCODE_SRCS_C)) TESTOPUSENCODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSENCODE_SRCS_C))
TESTOPUSPADDING_SRCS_C = tests/test_opus_padding.c TESTOPUSPADDING_SRCS_C = tests/test_opus_padding.c

View File

@ -97,7 +97,7 @@ TESTOPUSAPI_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSAPI_SRCS_C))
TESTOPUSDECODE_SRCS_C = tests/test_opus_decode.c TESTOPUSDECODE_SRCS_C = tests/test_opus_decode.c
TESTOPUSDECODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSDECODE_SRCS_C)) TESTOPUSDECODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSDECODE_SRCS_C))
TESTOPUSENCODE_SRCS_C = tests/test_opus_encode.c TESTOPUSENCODE_SRCS_C = tests/test_opus_encode.c tests/opus_encode_regressions.c
TESTOPUSENCODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSENCODE_SRCS_C)) TESTOPUSENCODE_OBJS := $(patsubst %.c,%$(OBJSUFFIX),$(TESTOPUSENCODE_SRCS_C))
TESTOPUSPADDING_SRCS_C = tests/test_opus_padding.c TESTOPUSPADDING_SRCS_C = tests/test_opus_padding.c

View File

@ -0,0 +1,19 @@
set(OPUS_VERSION @PROJECT_VERSION@)
set(OPUS_VERSION_STRING @PROJECT_VERSION@)
set(OPUS_VERSION_MAJOR @PROJECT_VERSION_MAJOR@)
set(OPUS_VERSION_MINOR @PROJECT_VERSION_MINOR@)
set(OPUS_VERSION_PATCH @PROJECT_VERSION_PATCH@)
@PACKAGE_INIT@
set_and_check(OPUS_INCLUDE_DIR "@PACKAGE_INCLUDE_INSTALL_DIR@")
set_and_check(OPUS_INCLUDE_DIRS "@PACKAGE_INCLUDE_INSTALL_DIR@")
include(${CMAKE_CURRENT_LIST_DIR}/OpusTargets.cmake)
set(OPUS_LIBRARY Opus::opus)
set(OPUS_LIBRARIES Opus::opus)
check_required_components(Opus)
set(OPUS_FOUND 1)

View File

@ -29,7 +29,7 @@ additional debugging data and cannot support seeking.
Opus stored in files should use the Ogg encapsulation for Opus which is Opus stored in files should use the Ogg encapsulation for Opus which is
described at: described at:
https://wiki.xiph.org/OggOpus https://tools.ietf.org/html/rfc7845
An opus-tools package is available which provides encoding and decoding of An opus-tools package is available which provides encoding and decoding of
Ogg encapsulated Opus files and includes a number of useful features. Ogg encapsulated Opus files and includes a number of useful features.
@ -43,25 +43,43 @@ or on the main Opus website:
To build from a distribution tarball, you only need to do the following: To build from a distribution tarball, you only need to do the following:
% ./configure % ./configure
% make % make
To build from the git repository, the following steps are necessary: To build from the git repository, the following steps are necessary:
0) Set up a development environment:
On an Ubuntu or Debian family Linux distribution:
% sudo apt-get install git autoconf automake libtool gcc make
On a Fedora/Redhat based Linux:
% sudo dnf install git autoconf automake libtool gcc make
Or for older Redhat/Centos Linux releases:
% sudo yum install git autoconf automake libtool gcc make
On Apple macOS, install Xcode and brew.sh, then in the Terminal enter:
% brew install autoconf automake libtool
1) Clone the repository: 1) Clone the repository:
% git clone https://git.xiph.org/opus.git % git clone https://git.xiph.org/opus.git
% cd opus % cd opus
2) Compiling the source 2) Compiling the source
% ./autogen.sh % ./autogen.sh
% ./configure % ./configure
% make % make
3) Install the codec libraries (optional) 3) Install the codec libraries (optional)
% sudo make install % sudo make install
Once you have compiled the codec, there will be a opus_demo executable Once you have compiled the codec, there will be a opus_demo executable
in the top directory. in the top directory.
@ -102,17 +120,18 @@ which SHOULD be run after compiling the package especially the first
time it is run on a new platform. time it is run on a new platform.
To run the integrated tests: To run the integrated tests:
% make check
% make check
There is also collection of standard test vectors which are not There is also collection of standard test vectors which are not
included in this package for size reasons but can be obtained from: included in this package for size reasons but can be obtained from:
https://opus-codec.org/testvectors/opus_testvectors.tar.gz https://opus-codec.org/docs/opus_testvectors-rfc8251.tar.gz
To run compare the code to these test vectors: To run compare the code to these test vectors:
% curl -O https://opus-codec.org/testvectors/opus_testvectors.tar.gz % curl -OL https://opus-codec.org/docs/opus_testvectors-rfc8251.tar.gz
% tar -zxf opus_testvectors.tar.gz % tar -zxf opus_testvectors-rfc8251.tar.gz
% ./tests/run_vectors.sh ./ opus_testvectors 48000 % ./tests/run_vectors.sh ./ opus_newvectors 48000
== Portability notes == == Portability notes ==

View File

@ -1,6 +1,6 @@
# generated automatically by aclocal 1.15 -*- Autoconf -*- # generated automatically by aclocal 1.15.1 -*- Autoconf -*-
# Copyright (C) 1996-2014 Free Software Foundation, Inc. # Copyright (C) 1996-2017 Free Software Foundation, Inc.
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -20,7 +20,7 @@ You have another version of autoconf. It may work, but is not guaranteed to.
If you have problems, you may need to regenerate the build system entirely. If you have problems, you may need to regenerate the build system entirely.
To do so, use the procedure documented by the package, typically 'autoreconf'.])]) To do so, use the procedure documented by the package, typically 'autoreconf'.])])
# Copyright (C) 2002-2014 Free Software Foundation, Inc. # Copyright (C) 2002-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -35,7 +35,7 @@ AC_DEFUN([AM_AUTOMAKE_VERSION],
[am__api_version='1.15' [am__api_version='1.15'
dnl Some users find AM_AUTOMAKE_VERSION and mistake it for a way to dnl Some users find AM_AUTOMAKE_VERSION and mistake it for a way to
dnl require some minimum version. Point them to the right macro. dnl require some minimum version. Point them to the right macro.
m4_if([$1], [1.15], [], m4_if([$1], [1.15.1], [],
[AC_FATAL([Do not call $0, use AM_INIT_AUTOMAKE([$1]).])])dnl [AC_FATAL([Do not call $0, use AM_INIT_AUTOMAKE([$1]).])])dnl
]) ])
@ -51,14 +51,14 @@ m4_define([_AM_AUTOCONF_VERSION], [])
# Call AM_AUTOMAKE_VERSION and AM_AUTOMAKE_VERSION so they can be traced. # Call AM_AUTOMAKE_VERSION and AM_AUTOMAKE_VERSION so they can be traced.
# This function is AC_REQUIREd by AM_INIT_AUTOMAKE. # This function is AC_REQUIREd by AM_INIT_AUTOMAKE.
AC_DEFUN([AM_SET_CURRENT_AUTOMAKE_VERSION], AC_DEFUN([AM_SET_CURRENT_AUTOMAKE_VERSION],
[AM_AUTOMAKE_VERSION([1.15])dnl [AM_AUTOMAKE_VERSION([1.15.1])dnl
m4_ifndef([AC_AUTOCONF_VERSION], m4_ifndef([AC_AUTOCONF_VERSION],
[m4_copy([m4_PACKAGE_VERSION], [AC_AUTOCONF_VERSION])])dnl [m4_copy([m4_PACKAGE_VERSION], [AC_AUTOCONF_VERSION])])dnl
_AM_AUTOCONF_VERSION(m4_defn([AC_AUTOCONF_VERSION]))]) _AM_AUTOCONF_VERSION(m4_defn([AC_AUTOCONF_VERSION]))])
# Figure out how to run the assembler. -*- Autoconf -*- # Figure out how to run the assembler. -*- Autoconf -*-
# Copyright (C) 2001-2014 Free Software Foundation, Inc. # Copyright (C) 2001-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -78,7 +78,7 @@ _AM_IF_OPTION([no-dependencies],, [_AM_DEPENDENCIES([CCAS])])dnl
# AM_AUX_DIR_EXPAND -*- Autoconf -*- # AM_AUX_DIR_EXPAND -*- Autoconf -*-
# Copyright (C) 2001-2014 Free Software Foundation, Inc. # Copyright (C) 2001-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -130,7 +130,7 @@ am_aux_dir=`cd "$ac_aux_dir" && pwd`
# AM_CONDITIONAL -*- Autoconf -*- # AM_CONDITIONAL -*- Autoconf -*-
# Copyright (C) 1997-2014 Free Software Foundation, Inc. # Copyright (C) 1997-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -161,7 +161,7 @@ AC_CONFIG_COMMANDS_PRE(
Usually this means the macro was only invoked conditionally.]]) Usually this means the macro was only invoked conditionally.]])
fi])]) fi])])
# Copyright (C) 1999-2014 Free Software Foundation, Inc. # Copyright (C) 1999-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -352,7 +352,7 @@ _AM_SUBST_NOTMAKE([am__nodep])dnl
# Generate code to set up dependency tracking. -*- Autoconf -*- # Generate code to set up dependency tracking. -*- Autoconf -*-
# Copyright (C) 1999-2014 Free Software Foundation, Inc. # Copyright (C) 1999-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -428,7 +428,7 @@ AC_DEFUN([AM_OUTPUT_DEPENDENCY_COMMANDS],
# Do all the work for Automake. -*- Autoconf -*- # Do all the work for Automake. -*- Autoconf -*-
# Copyright (C) 1996-2014 Free Software Foundation, Inc. # Copyright (C) 1996-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -625,7 +625,7 @@ for _am_header in $config_headers :; do
done done
echo "timestamp for $_am_arg" >`AS_DIRNAME(["$_am_arg"])`/stamp-h[]$_am_stamp_count]) echo "timestamp for $_am_arg" >`AS_DIRNAME(["$_am_arg"])`/stamp-h[]$_am_stamp_count])
# Copyright (C) 2001-2014 Free Software Foundation, Inc. # Copyright (C) 2001-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -646,7 +646,7 @@ if test x"${install_sh+set}" != xset; then
fi fi
AC_SUBST([install_sh])]) AC_SUBST([install_sh])])
# Copyright (C) 2003-2014 Free Software Foundation, Inc. # Copyright (C) 2003-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -668,7 +668,7 @@ AC_SUBST([am__leading_dot])])
# Add --enable-maintainer-mode option to configure. -*- Autoconf -*- # Add --enable-maintainer-mode option to configure. -*- Autoconf -*-
# From Jim Meyering # From Jim Meyering
# Copyright (C) 1996-2014 Free Software Foundation, Inc. # Copyright (C) 1996-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -703,7 +703,7 @@ AC_MSG_CHECKING([whether to enable maintainer-specific portions of Makefiles])
# Check to see how 'make' treats includes. -*- Autoconf -*- # Check to see how 'make' treats includes. -*- Autoconf -*-
# Copyright (C) 2001-2014 Free Software Foundation, Inc. # Copyright (C) 2001-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -753,7 +753,7 @@ rm -f confinc confmf
# Fake the existence of programs that GNU maintainers use. -*- Autoconf -*- # Fake the existence of programs that GNU maintainers use. -*- Autoconf -*-
# Copyright (C) 1997-2014 Free Software Foundation, Inc. # Copyright (C) 1997-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -792,7 +792,7 @@ fi
# Helper functions for option handling. -*- Autoconf -*- # Helper functions for option handling. -*- Autoconf -*-
# Copyright (C) 2001-2014 Free Software Foundation, Inc. # Copyright (C) 2001-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -821,7 +821,7 @@ AC_DEFUN([_AM_SET_OPTIONS],
AC_DEFUN([_AM_IF_OPTION], AC_DEFUN([_AM_IF_OPTION],
[m4_ifset(_AM_MANGLE_OPTION([$1]), [$2], [$3])]) [m4_ifset(_AM_MANGLE_OPTION([$1]), [$2], [$3])])
# Copyright (C) 1999-2014 Free Software Foundation, Inc. # Copyright (C) 1999-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -868,7 +868,7 @@ AC_LANG_POP([C])])
# For backward compatibility. # For backward compatibility.
AC_DEFUN_ONCE([AM_PROG_CC_C_O], [AC_REQUIRE([AC_PROG_CC])]) AC_DEFUN_ONCE([AM_PROG_CC_C_O], [AC_REQUIRE([AC_PROG_CC])])
# Copyright (C) 2001-2014 Free Software Foundation, Inc. # Copyright (C) 2001-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -887,7 +887,7 @@ AC_DEFUN([AM_RUN_LOG],
# Check to make sure that the build environment is sane. -*- Autoconf -*- # Check to make sure that the build environment is sane. -*- Autoconf -*-
# Copyright (C) 1996-2014 Free Software Foundation, Inc. # Copyright (C) 1996-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -968,7 +968,7 @@ AC_CONFIG_COMMANDS_PRE(
rm -f conftest.file rm -f conftest.file
]) ])
# Copyright (C) 2009-2014 Free Software Foundation, Inc. # Copyright (C) 2009-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -1028,7 +1028,7 @@ AC_SUBST([AM_BACKSLASH])dnl
_AM_SUBST_NOTMAKE([AM_BACKSLASH])dnl _AM_SUBST_NOTMAKE([AM_BACKSLASH])dnl
]) ])
# Copyright (C) 2001-2014 Free Software Foundation, Inc. # Copyright (C) 2001-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -1056,7 +1056,7 @@ fi
INSTALL_STRIP_PROGRAM="\$(install_sh) -c -s" INSTALL_STRIP_PROGRAM="\$(install_sh) -c -s"
AC_SUBST([INSTALL_STRIP_PROGRAM])]) AC_SUBST([INSTALL_STRIP_PROGRAM])])
# Copyright (C) 2006-2014 Free Software Foundation, Inc. # Copyright (C) 2006-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -1075,7 +1075,7 @@ AC_DEFUN([AM_SUBST_NOTMAKE], [_AM_SUBST_NOTMAKE($@)])
# Check how to create a tarball. -*- Autoconf -*- # Check how to create a tarball. -*- Autoconf -*-
# Copyright (C) 2004-2014 Free Software Foundation, Inc. # Copyright (C) 2004-2017 Free Software Foundation, Inc.
# #
# This file is free software; the Free Software Foundation # This file is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it, # gives unlimited permission to copy and/or distribute it,
@ -1207,6 +1207,7 @@ AC_SUBST([am__untar])
]) # _AM_PROG_TAR ]) # _AM_PROG_TAR
m4_include([m4/as-gcc-inline-assembly.m4]) m4_include([m4/as-gcc-inline-assembly.m4])
m4_include([m4/ax_add_fortify_source.m4])
m4_include([m4/libtool.m4]) m4_include([m4/libtool.m4])
m4_include([m4/ltoptions.m4]) m4_include([m4/ltoptions.m4])
m4_include([m4/ltsugar.m4]) m4_include([m4/ltsugar.m4])

View File

@ -58,12 +58,12 @@
# define S_MUL(a,b) MULT16_32_Q15(b, a) # define S_MUL(a,b) MULT16_32_Q15(b, a)
# define C_MUL(m,a,b) \ # define C_MUL(m,a,b) \
do{ (m).r = SUB32(S_MUL((a).r,(b).r) , S_MUL((a).i,(b).i)); \ do{ (m).r = SUB32_ovflw(S_MUL((a).r,(b).r) , S_MUL((a).i,(b).i)); \
(m).i = ADD32(S_MUL((a).r,(b).i) , S_MUL((a).i,(b).r)); }while(0) (m).i = ADD32_ovflw(S_MUL((a).r,(b).i) , S_MUL((a).i,(b).r)); }while(0)
# define C_MULC(m,a,b) \ # define C_MULC(m,a,b) \
do{ (m).r = ADD32(S_MUL((a).r,(b).r) , S_MUL((a).i,(b).i)); \ do{ (m).r = ADD32_ovflw(S_MUL((a).r,(b).r) , S_MUL((a).i,(b).i)); \
(m).i = SUB32(S_MUL((a).i,(b).r) , S_MUL((a).r,(b).i)); }while(0) (m).i = SUB32_ovflw(S_MUL((a).i,(b).r) , S_MUL((a).r,(b).i)); }while(0)
# define C_MULBYSCALAR( c, s ) \ # define C_MULBYSCALAR( c, s ) \
do{ (c).r = S_MUL( (c).r , s ) ;\ do{ (c).r = S_MUL( (c).r , s ) ;\
@ -77,17 +77,17 @@
DIVSCALAR( (c).i , div); }while (0) DIVSCALAR( (c).i , div); }while (0)
#define C_ADD( res, a,b)\ #define C_ADD( res, a,b)\
do {(res).r=ADD32((a).r,(b).r); (res).i=ADD32((a).i,(b).i); \ do {(res).r=ADD32_ovflw((a).r,(b).r); (res).i=ADD32_ovflw((a).i,(b).i); \
}while(0) }while(0)
#define C_SUB( res, a,b)\ #define C_SUB( res, a,b)\
do {(res).r=SUB32((a).r,(b).r); (res).i=SUB32((a).i,(b).i); \ do {(res).r=SUB32_ovflw((a).r,(b).r); (res).i=SUB32_ovflw((a).i,(b).i); \
}while(0) }while(0)
#define C_ADDTO( res , a)\ #define C_ADDTO( res , a)\
do {(res).r = ADD32((res).r, (a).r); (res).i = ADD32((res).i,(a).i);\ do {(res).r = ADD32_ovflw((res).r, (a).r); (res).i = ADD32_ovflw((res).i,(a).i);\
}while(0) }while(0)
#define C_SUBFROM( res , a)\ #define C_SUBFROM( res , a)\
do {(res).r = ADD32((res).r,(a).r); (res).i = SUB32((res).i,(a).i); \ do {(res).r = ADD32_ovflw((res).r,(a).r); (res).i = SUB32_ovflw((res).i,(a).i); \
}while(0) }while(0)
#if defined(OPUS_ARM_INLINE_ASM) #if defined(OPUS_ARM_INLINE_ASM)

View File

@ -46,25 +46,50 @@
# endif # endif
# endif # endif
#if OPUS_GNUC_PREREQ(3, 0)
#define opus_likely(x) (__builtin_expect(!!(x), 1))
#define opus_unlikely(x) (__builtin_expect(!!(x), 0))
#else
#define opus_likely(x) (!!(x))
#define opus_unlikely(x) (!!(x))
#endif
#define CELT_SIG_SCALE 32768.f #define CELT_SIG_SCALE 32768.f
#define celt_fatal(str) _celt_fatal(str, __FILE__, __LINE__); #define CELT_FATAL(str) celt_fatal(str, __FILE__, __LINE__);
#ifdef ENABLE_ASSERTIONS
#if defined(ENABLE_ASSERTIONS) || defined(ENABLE_HARDENING)
#ifdef __GNUC__
__attribute__((noreturn))
#endif
void celt_fatal(const char *str, const char *file, int line);
#if defined(CELT_C) && !defined(OVERRIDE_celt_fatal)
#include <stdio.h> #include <stdio.h>
#include <stdlib.h> #include <stdlib.h>
#ifdef __GNUC__ #ifdef __GNUC__
__attribute__((noreturn)) __attribute__((noreturn))
#endif #endif
static OPUS_INLINE void _celt_fatal(const char *str, const char *file, int line) void celt_fatal(const char *str, const char *file, int line)
{ {
fprintf (stderr, "Fatal (internal) error in %s, line %d: %s\n", file, line, str); fprintf (stderr, "Fatal (internal) error in %s, line %d: %s\n", file, line, str);
abort(); abort();
} }
#define celt_assert(cond) {if (!(cond)) {celt_fatal("assertion failed: " #cond);}} #endif
#define celt_assert2(cond, message) {if (!(cond)) {celt_fatal("assertion failed: " #cond "\n" message);}}
#define celt_assert(cond) {if (!(cond)) {CELT_FATAL("assertion failed: " #cond);}}
#define celt_assert2(cond, message) {if (!(cond)) {CELT_FATAL("assertion failed: " #cond "\n" message);}}
#define MUST_SUCCEED(call) celt_assert((call) == OPUS_OK)
#else #else
#define celt_assert(cond) #define celt_assert(cond)
#define celt_assert2(cond, message) #define celt_assert2(cond, message)
#define MUST_SUCCEED(call) do {if((call) != OPUS_OK) {RESTORE_STACK; return OPUS_INTERNAL_ERROR;} } while (0)
#endif
#if defined(ENABLE_ASSERTIONS)
#define celt_sig_assert(cond) {if (!(cond)) {CELT_FATAL("signal assertion failed: " #cond);}}
#else
#define celt_sig_assert(cond)
#endif #endif
#define IMUL32(a,b) ((a)*(b)) #define IMUL32(a,b) ((a)*(b))
@ -78,20 +103,35 @@ static OPUS_INLINE void _celt_fatal(const char *str, const char *file, int line)
#define UADD32(a,b) ((a)+(b)) #define UADD32(a,b) ((a)+(b))
#define USUB32(a,b) ((a)-(b)) #define USUB32(a,b) ((a)-(b))
/* Set this if opus_int64 is a native type of the CPU. */
/* Assume that all LP64 architectures have fast 64-bit types; also x86_64
(which can be ILP32 for x32) and Win64 (which is LLP64). */
#if defined(__x86_64__) || defined(__LP64__) || defined(_WIN64)
#define OPUS_FAST_INT64 1
#else
#define OPUS_FAST_INT64 0
#endif
#define PRINT_MIPS(file) #define PRINT_MIPS(file)
#ifdef FIXED_POINT #ifdef FIXED_POINT
typedef opus_int16 opus_val16; typedef opus_int16 opus_val16;
typedef opus_int32 opus_val32; typedef opus_int32 opus_val32;
typedef opus_int64 opus_val64;
typedef opus_val32 celt_sig; typedef opus_val32 celt_sig;
typedef opus_val16 celt_norm; typedef opus_val16 celt_norm;
typedef opus_val32 celt_ener; typedef opus_val32 celt_ener;
#define celt_isnan(x) 0
#define Q15ONE 32767 #define Q15ONE 32767
#define SIG_SHIFT 12 #define SIG_SHIFT 12
/* Safe saturation value for 32-bit signals. Should be less than
2^31*(1-0.85) to avoid blowing up on DC at deemphasis.*/
#define SIG_SAT (300000000)
#define NORM_SCALING 16384 #define NORM_SCALING 16384
@ -118,7 +158,9 @@ static OPUS_INLINE opus_int16 SAT16(opus_int32 x) {
#include "fixed_generic.h" #include "fixed_generic.h"
#ifdef OPUS_ARM_INLINE_EDSP #ifdef OPUS_ARM_PRESUME_AARCH64_NEON_INTR
#include "arm/fixed_arm64.h"
#elif defined (OPUS_ARM_INLINE_EDSP)
#include "arm/fixed_armv5e.h" #include "arm/fixed_armv5e.h"
#elif defined (OPUS_ARM_INLINE_ASM) #elif defined (OPUS_ARM_INLINE_ASM)
#include "arm/fixed_armv4.h" #include "arm/fixed_armv4.h"
@ -136,6 +178,7 @@ static OPUS_INLINE opus_int16 SAT16(opus_int32 x) {
typedef float opus_val16; typedef float opus_val16;
typedef float opus_val32; typedef float opus_val32;
typedef float opus_val64;
typedef float celt_sig; typedef float celt_sig;
typedef float celt_norm; typedef float celt_norm;
@ -175,6 +218,7 @@ static OPUS_INLINE int celt_isnan(float x)
#define NEG16(x) (-(x)) #define NEG16(x) (-(x))
#define NEG32(x) (-(x)) #define NEG32(x) (-(x))
#define NEG32_ovflw(x) (-(x))
#define EXTRACT16(x) (x) #define EXTRACT16(x) (x)
#define EXTEND32(x) (x) #define EXTEND32(x) (x)
#define SHR16(a,shift) (a) #define SHR16(a,shift) (a)
@ -191,6 +235,7 @@ static OPUS_INLINE int celt_isnan(float x)
#define SATURATE16(x) (x) #define SATURATE16(x) (x)
#define ROUND16(a,shift) (a) #define ROUND16(a,shift) (a)
#define SROUND16(a,shift) (a)
#define HALF16(x) (.5f*(x)) #define HALF16(x) (.5f*(x))
#define HALF32(x) (.5f*(x)) #define HALF32(x) (.5f*(x))
@ -198,6 +243,8 @@ static OPUS_INLINE int celt_isnan(float x)
#define SUB16(a,b) ((a)-(b)) #define SUB16(a,b) ((a)-(b))
#define ADD32(a,b) ((a)+(b)) #define ADD32(a,b) ((a)+(b))
#define SUB32(a,b) ((a)-(b)) #define SUB32(a,b) ((a)-(b))
#define ADD32_ovflw(a,b) ((a)+(b))
#define SUB32_ovflw(a,b) ((a)-(b))
#define MULT16_16_16(a,b) ((a)*(b)) #define MULT16_16_16(a,b) ((a)*(b))
#define MULT16_16(a,b) ((opus_val32)(a)*(opus_val32)(b)) #define MULT16_16(a,b) ((opus_val32)(a)*(opus_val32)(b))
#define MAC16_16(c,a,b) ((c)+(opus_val32)(a)*(opus_val32)(b)) #define MAC16_16(c,a,b) ((c)+(opus_val32)(a)*(opus_val32)(b))
@ -232,9 +279,9 @@ static OPUS_INLINE int celt_isnan(float x)
#ifndef GLOBAL_STACK_SIZE #ifndef GLOBAL_STACK_SIZE
#ifdef FIXED_POINT #ifdef FIXED_POINT
#define GLOBAL_STACK_SIZE 100000 #define GLOBAL_STACK_SIZE 120000
#else #else
#define GLOBAL_STACK_SIZE 100000 #define GLOBAL_STACK_SIZE 120000
#endif #endif
#endif #endif

View File

@ -164,11 +164,11 @@ while (<>) {
$prefix = ""; $prefix = "";
if ($proc) if ($proc)
{ {
$prefix = $prefix.sprintf("\t.type\t%s, %%function; ",$proc) unless ($apple); $prefix = $prefix.sprintf("\t.type\t%s, %%function", $proc) unless ($apple);
# Make sure we $prefix isn't empty here (for the $apple case). # Make sure we $prefix isn't empty here (for the $apple case).
# We handle mangling the label here, make sure it doesn't match # We handle mangling the label here, make sure it doesn't match
# the label handling below (if $prefix would be empty). # the label handling below (if $prefix would be empty).
$prefix = "; "; $prefix = $prefix."; ";
push(@proc_stack, $proc); push(@proc_stack, $proc);
s/^[A-Za-z_\.]\w+/$symprefix$&:/; s/^[A-Za-z_\.]\w+/$symprefix$&:/;
} }

View File

@ -35,18 +35,40 @@
#if defined(OPUS_HAVE_RTCD) #if defined(OPUS_HAVE_RTCD)
# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)
opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *x, const opus_val16 *y, int N) = {
celt_inner_prod_c, /* ARMv4 */
celt_inner_prod_c, /* EDSP */
celt_inner_prod_c, /* Media */
celt_inner_prod_neon /* NEON */
};
void (*const DUAL_INNER_PROD_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02,
int N, opus_val32 *xy1, opus_val32 *xy2) = {
dual_inner_prod_c, /* ARMv4 */
dual_inner_prod_c, /* EDSP */
dual_inner_prod_c, /* Media */
dual_inner_prod_neon /* NEON */
};
# endif
# if defined(FIXED_POINT) # if defined(FIXED_POINT)
# if ((defined(OPUS_ARM_MAY_HAVE_NEON) && !defined(OPUS_ARM_PRESUME_NEON)) || \
(defined(OPUS_ARM_MAY_HAVE_MEDIA) && !defined(OPUS_ARM_PRESUME_MEDIA)) || \
(defined(OPUS_ARM_MAY_HAVE_EDSP) && !defined(OPUS_ARM_PRESUME_EDSP)))
opus_val32 (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *, opus_val32 (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
const opus_val16 *, opus_val32 *, int , int) = { const opus_val16 *, opus_val32 *, int, int, int) = {
celt_pitch_xcorr_c, /* ARMv4 */ celt_pitch_xcorr_c, /* ARMv4 */
MAY_HAVE_EDSP(celt_pitch_xcorr), /* EDSP */ MAY_HAVE_EDSP(celt_pitch_xcorr), /* EDSP */
MAY_HAVE_MEDIA(celt_pitch_xcorr), /* Media */ MAY_HAVE_MEDIA(celt_pitch_xcorr), /* Media */
MAY_HAVE_NEON(celt_pitch_xcorr) /* NEON */ MAY_HAVE_NEON(celt_pitch_xcorr) /* NEON */
}; };
# endif
# else /* !FIXED_POINT */ # else /* !FIXED_POINT */
# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) # if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)
void (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *, void (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
const opus_val16 *, opus_val32 *, int, int) = { const opus_val16 *, opus_val32 *, int, int, int) = {
celt_pitch_xcorr_c, /* ARMv4 */ celt_pitch_xcorr_c, /* ARMv4 */
celt_pitch_xcorr_c, /* EDSP */ celt_pitch_xcorr_c, /* EDSP */
celt_pitch_xcorr_c, /* Media */ celt_pitch_xcorr_c, /* Media */
@ -55,6 +77,23 @@ void (*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
# endif # endif
# endif /* FIXED_POINT */ # endif /* FIXED_POINT */
#if defined(FIXED_POINT) && defined(OPUS_HAVE_RTCD) && \
defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR)
void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])(
const opus_val16 *x,
const opus_val16 *y,
opus_val32 sum[4],
int len
) = {
xcorr_kernel_c, /* ARMv4 */
xcorr_kernel_c, /* EDSP */
xcorr_kernel_c, /* Media */
xcorr_kernel_neon_fixed, /* Neon */
};
#endif
# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR) # if defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
# if defined(HAVE_ARM_NE10) # if defined(HAVE_ARM_NE10)
# if defined(CUSTOM_MODES) # if defined(CUSTOM_MODES)

View File

@ -37,11 +37,12 @@
#include "cpu_support.h" #include "cpu_support.h"
#include "os_support.h" #include "os_support.h"
#include "opus_types.h" #include "opus_types.h"
#include "arch.h"
#define OPUS_CPU_ARM_V4 (1) #define OPUS_CPU_ARM_V4_FLAG (1<<OPUS_ARCH_ARM_V4)
#define OPUS_CPU_ARM_EDSP (1<<1) #define OPUS_CPU_ARM_EDSP_FLAG (1<<OPUS_ARCH_ARM_EDSP)
#define OPUS_CPU_ARM_MEDIA (1<<2) #define OPUS_CPU_ARM_MEDIA_FLAG (1<<OPUS_ARCH_ARM_MEDIA)
#define OPUS_CPU_ARM_NEON (1<<3) #define OPUS_CPU_ARM_NEON_FLAG (1<<OPUS_ARCH_ARM_NEON)
#if defined(_MSC_VER) #if defined(_MSC_VER)
/*For GetExceptionCode() and EXCEPTION_ILLEGAL_INSTRUCTION.*/ /*For GetExceptionCode() and EXCEPTION_ILLEGAL_INSTRUCTION.*/
@ -55,20 +56,22 @@ static OPUS_INLINE opus_uint32 opus_cpu_capabilities(void){
/* MSVC has no OPUS_INLINE __asm support for ARM, but it does let you __emit /* MSVC has no OPUS_INLINE __asm support for ARM, but it does let you __emit
* instructions via their assembled hex code. * instructions via their assembled hex code.
* All of these instructions should be essentially nops. */ * All of these instructions should be essentially nops. */
# if defined(OPUS_ARM_MAY_HAVE_EDSP) # if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_MEDIA) \
|| defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
__try{ __try{
/*PLD [r13]*/ /*PLD [r13]*/
__emit(0xF5DDF000); __emit(0xF5DDF000);
flags|=OPUS_CPU_ARM_EDSP; flags|=OPUS_CPU_ARM_EDSP_FLAG;
} }
__except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){
/*Ignore exception.*/ /*Ignore exception.*/
} }
# if defined(OPUS_ARM_MAY_HAVE_MEDIA) # if defined(OPUS_ARM_MAY_HAVE_MEDIA) \
|| defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
__try{ __try{
/*SHADD8 r3,r3,r3*/ /*SHADD8 r3,r3,r3*/
__emit(0xE6333F93); __emit(0xE6333F93);
flags|=OPUS_CPU_ARM_MEDIA; flags|=OPUS_CPU_ARM_MEDIA_FLAG;
} }
__except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){
/*Ignore exception.*/ /*Ignore exception.*/
@ -77,7 +80,7 @@ static OPUS_INLINE opus_uint32 opus_cpu_capabilities(void){
__try{ __try{
/*VORR q0,q0,q0*/ /*VORR q0,q0,q0*/
__emit(0xF2200150); __emit(0xF2200150);
flags|=OPUS_CPU_ARM_NEON; flags|=OPUS_CPU_ARM_NEON_FLAG;
} }
__except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){ __except(GetExceptionCode()==EXCEPTION_ILLEGAL_INSTRUCTION){
/*Ignore exception.*/ /*Ignore exception.*/
@ -107,26 +110,26 @@ opus_uint32 opus_cpu_capabilities(void)
while(fgets(buf, 512, cpuinfo) != NULL) while(fgets(buf, 512, cpuinfo) != NULL)
{ {
# if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) # if defined(OPUS_ARM_MAY_HAVE_EDSP) || defined(OPUS_ARM_MAY_HAVE_MEDIA) \
|| defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
/* Search for edsp and neon flag */ /* Search for edsp and neon flag */
if(memcmp(buf, "Features", 8) == 0) if(memcmp(buf, "Features", 8) == 0)
{ {
char *p; char *p;
# if defined(OPUS_ARM_MAY_HAVE_EDSP)
p = strstr(buf, " edsp"); p = strstr(buf, " edsp");
if(p != NULL && (p[5] == ' ' || p[5] == '\n')) if(p != NULL && (p[5] == ' ' || p[5] == '\n'))
flags |= OPUS_CPU_ARM_EDSP; flags |= OPUS_CPU_ARM_EDSP_FLAG;
# endif
# if defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR) # if defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
p = strstr(buf, " neon"); p = strstr(buf, " neon");
if(p != NULL && (p[5] == ' ' || p[5] == '\n')) if(p != NULL && (p[5] == ' ' || p[5] == '\n'))
flags |= OPUS_CPU_ARM_NEON; flags |= OPUS_CPU_ARM_NEON_FLAG;
# endif # endif
} }
# endif # endif
# if defined(OPUS_ARM_MAY_HAVE_MEDIA) # if defined(OPUS_ARM_MAY_HAVE_MEDIA) \
|| defined(OPUS_ARM_MAY_HAVE_NEON) || defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
/* Search for media capabilities (>= ARMv6) */ /* Search for media capabilities (>= ARMv6) */
if(memcmp(buf, "CPU architecture:", 17) == 0) if(memcmp(buf, "CPU architecture:", 17) == 0)
{ {
@ -134,7 +137,7 @@ opus_uint32 opus_cpu_capabilities(void)
version = atoi(buf+17); version = atoi(buf+17);
if(version >= 6) if(version >= 6)
flags |= OPUS_CPU_ARM_MEDIA; flags |= OPUS_CPU_ARM_MEDIA_FLAG;
} }
# endif # endif
} }
@ -156,18 +159,26 @@ int opus_select_arch(void)
opus_uint32 flags = opus_cpu_capabilities(); opus_uint32 flags = opus_cpu_capabilities();
int arch = 0; int arch = 0;
if(!(flags & OPUS_CPU_ARM_EDSP)) if(!(flags & OPUS_CPU_ARM_EDSP_FLAG)) {
/* Asserts ensure arch values are sequential */
celt_assert(arch == OPUS_ARCH_ARM_V4);
return arch; return arch;
}
arch++; arch++;
if(!(flags & OPUS_CPU_ARM_MEDIA)) if(!(flags & OPUS_CPU_ARM_MEDIA_FLAG)) {
celt_assert(arch == OPUS_ARCH_ARM_EDSP);
return arch; return arch;
}
arch++; arch++;
if(!(flags & OPUS_CPU_ARM_NEON)) if(!(flags & OPUS_CPU_ARM_NEON_FLAG)) {
celt_assert(arch == OPUS_ARCH_ARM_MEDIA);
return arch; return arch;
}
arch++; arch++;
celt_assert(arch == OPUS_ARCH_ARM_NEON);
return arch; return arch;
} }

View File

@ -66,6 +66,12 @@
# if defined(OPUS_HAVE_RTCD) # if defined(OPUS_HAVE_RTCD)
int opus_select_arch(void); int opus_select_arch(void);
#define OPUS_ARCH_ARM_V4 (0)
#define OPUS_ARCH_ARM_EDSP (1)
#define OPUS_ARCH_ARM_MEDIA (2)
#define OPUS_ARCH_ARM_NEON (3)
# endif # endif
#endif #endif

View File

@ -1,7 +1,7 @@
/* Copyright (c) 2015 Xiph.Org Foundation /* Copyright (c) 2015 Xiph.Org Foundation
Written by Viswanath Puttagunta */ Written by Viswanath Puttagunta */
/** /**
@file celt_ne10_fft.c @file celt_fft_ne10.c
@brief ARM Neon optimizations for fft using NE10 library @brief ARM Neon optimizations for fft using NE10 library
*/ */
@ -36,7 +36,6 @@
#endif #endif
#endif #endif
#include <NE10_init.h>
#include <NE10_dsp.h> #include <NE10_dsp.h>
#include "os_support.h" #include "os_support.h"
#include "kiss_fft.h" #include "kiss_fft.h"

View File

@ -1,7 +1,7 @@
/* Copyright (c) 2015 Xiph.Org Foundation /* Copyright (c) 2015 Xiph.Org Foundation
Written by Viswanath Puttagunta */ Written by Viswanath Puttagunta */
/** /**
@file celt_ne10_mdct.c @file celt_mdct_ne10.c
@brief ARM Neon optimizations for mdct using NE10 library @brief ARM Neon optimizations for mdct using NE10 library
*/ */

View File

@ -37,7 +37,66 @@
#include <arm_neon.h> #include <arm_neon.h>
#include "../pitch.h" #include "../pitch.h"
#if !defined(FIXED_POINT) #if defined(FIXED_POINT)
void xcorr_kernel_neon_fixed(const opus_val16 * x, const opus_val16 * y, opus_val32 sum[4], int len)
{
int j;
int32x4_t a = vld1q_s32(sum);
/* Load y[0...3] */
/* This requires len>0 to always be valid (which we assert in the C code). */
int16x4_t y0 = vld1_s16(y);
y += 4;
for (j = 0; j + 8 <= len; j += 8)
{
/* Load x[0...7] */
int16x8_t xx = vld1q_s16(x);
int16x4_t x0 = vget_low_s16(xx);
int16x4_t x4 = vget_high_s16(xx);
/* Load y[4...11] */
int16x8_t yy = vld1q_s16(y);
int16x4_t y4 = vget_low_s16(yy);
int16x4_t y8 = vget_high_s16(yy);
int32x4_t a0 = vmlal_lane_s16(a, y0, x0, 0);
int32x4_t a1 = vmlal_lane_s16(a0, y4, x4, 0);
int16x4_t y1 = vext_s16(y0, y4, 1);
int16x4_t y5 = vext_s16(y4, y8, 1);
int32x4_t a2 = vmlal_lane_s16(a1, y1, x0, 1);
int32x4_t a3 = vmlal_lane_s16(a2, y5, x4, 1);
int16x4_t y2 = vext_s16(y0, y4, 2);
int16x4_t y6 = vext_s16(y4, y8, 2);
int32x4_t a4 = vmlal_lane_s16(a3, y2, x0, 2);
int32x4_t a5 = vmlal_lane_s16(a4, y6, x4, 2);
int16x4_t y3 = vext_s16(y0, y4, 3);
int16x4_t y7 = vext_s16(y4, y8, 3);
int32x4_t a6 = vmlal_lane_s16(a5, y3, x0, 3);
int32x4_t a7 = vmlal_lane_s16(a6, y7, x4, 3);
y0 = y8;
a = a7;
x += 8;
y += 8;
}
for (; j < len; j++)
{
int16x4_t x0 = vld1_dup_s16(x); /* load next x */
int32x4_t a0 = vmlal_s16(a, y0, x0);
int16x4_t y4 = vld1_dup_s16(y); /* load next y */
y0 = vext_s16(y0, y4, 1);
a = a0;
x++;
y++;
}
vst1q_s32(sum, a);
}
#else
/* /*
* Function: xcorr_kernel_neon_float * Function: xcorr_kernel_neon_float
* --------------------------------- * ---------------------------------
@ -132,121 +191,21 @@ static void xcorr_kernel_neon_float(const float32_t *x, const float32_t *y,
vst1q_f32(sum, SUMM); vst1q_f32(sum, SUMM);
} }
/*
* Function: xcorr_kernel_neon_float_process1
* ---------------------------------
* Computes single correlation values and stores in *sum
*/
static void xcorr_kernel_neon_float_process1(const float32_t *x,
const float32_t *y, float32_t *sum, int len) {
float32x4_t XX[4];
float32x4_t YY[4];
float32x2_t XX_2;
float32x2_t YY_2;
float32x4_t SUMM;
float32x2_t SUMM_2[2];
const float32_t *xi = x;
const float32_t *yi = y;
SUMM = vdupq_n_f32(0);
/* Work on 16 values per iteration */
while (len >= 16) {
XX[0] = vld1q_f32(xi);
xi += 4;
XX[1] = vld1q_f32(xi);
xi += 4;
XX[2] = vld1q_f32(xi);
xi += 4;
XX[3] = vld1q_f32(xi);
xi += 4;
YY[0] = vld1q_f32(yi);
yi += 4;
YY[1] = vld1q_f32(yi);
yi += 4;
YY[2] = vld1q_f32(yi);
yi += 4;
YY[3] = vld1q_f32(yi);
yi += 4;
SUMM = vmlaq_f32(SUMM, YY[0], XX[0]);
SUMM = vmlaq_f32(SUMM, YY[1], XX[1]);
SUMM = vmlaq_f32(SUMM, YY[2], XX[2]);
SUMM = vmlaq_f32(SUMM, YY[3], XX[3]);
len -= 16;
}
/* Work on 8 values */
if (len >= 8) {
XX[0] = vld1q_f32(xi);
xi += 4;
XX[1] = vld1q_f32(xi);
xi += 4;
YY[0] = vld1q_f32(yi);
yi += 4;
YY[1] = vld1q_f32(yi);
yi += 4;
SUMM = vmlaq_f32(SUMM, YY[0], XX[0]);
SUMM = vmlaq_f32(SUMM, YY[1], XX[1]);
len -= 8;
}
/* Work on 4 values */
if (len >= 4) {
XX[0] = vld1q_f32(xi);
xi += 4;
YY[0] = vld1q_f32(yi);
yi += 4;
SUMM = vmlaq_f32(SUMM, YY[0], XX[0]);
len -= 4;
}
/* Start accumulating results */
SUMM_2[0] = vget_low_f32(SUMM);
if (len >= 2) {
/* While at it, consume 2 more values if available */
XX_2 = vld1_f32(xi);
xi += 2;
YY_2 = vld1_f32(yi);
yi += 2;
SUMM_2[0] = vmla_f32(SUMM_2[0], YY_2, XX_2);
len -= 2;
}
SUMM_2[1] = vget_high_f32(SUMM);
SUMM_2[0] = vadd_f32(SUMM_2[0], SUMM_2[1]);
SUMM_2[0] = vpadd_f32(SUMM_2[0], SUMM_2[0]);
/* Ok, now we have result accumulated in SUMM_2[0].0 */
if (len > 0) {
/* Case when you have one value left */
XX_2 = vld1_dup_f32(xi);
YY_2 = vld1_dup_f32(yi);
SUMM_2[0] = vmla_f32(SUMM_2[0], XX_2, YY_2);
}
vst1_lane_f32(sum, SUMM_2[0], 0);
}
void celt_pitch_xcorr_float_neon(const opus_val16 *_x, const opus_val16 *_y, void celt_pitch_xcorr_float_neon(const opus_val16 *_x, const opus_val16 *_y,
opus_val32 *xcorr, int len, int max_pitch) { opus_val32 *xcorr, int len, int max_pitch, int arch) {
int i; int i;
(void)arch;
celt_assert(max_pitch > 0); celt_assert(max_pitch > 0);
celt_assert((((unsigned char *)_x-(unsigned char *)NULL)&3)==0); celt_sig_assert((((unsigned char *)_x-(unsigned char *)NULL)&3)==0);
for (i = 0; i < (max_pitch-3); i += 4) { for (i = 0; i < (max_pitch-3); i += 4) {
xcorr_kernel_neon_float((const float32_t *)_x, (const float32_t *)_y+i, xcorr_kernel_neon_float((const float32_t *)_x, (const float32_t *)_y+i,
(float32_t *)xcorr+i, len); (float32_t *)xcorr+i, len);
} }
/* In case max_pitch isn't multiple of 4 /* In case max_pitch isn't a multiple of 4, do non-unrolled version. */
* compute single correlation value per iteration
*/
for (; i < max_pitch; i++) { for (; i < max_pitch; i++) {
xcorr_kernel_neon_float_process1((const float32_t *)_x, xcorr[i] = celt_inner_prod_neon(_x, _y+i, len);
(const float32_t *)_y+i, (float32_t *)xcorr+i, len);
} }
} }
#endif #endif

View File

@ -44,7 +44,7 @@
.if OPUS_ARM_MAY_HAVE_NEON .if OPUS_ARM_MAY_HAVE_NEON
@ Compute sum[k]=sum(x[j]*y[j+k],j=0...len-1), k=0...3 @ Compute sum[k]=sum(x[j]*y[j+k],j=0...len-1), k=0...3
; xcorr_kernel_neon: @ PROC .type xcorr_kernel_neon, %function; xcorr_kernel_neon: @ PROC
xcorr_kernel_neon_start: xcorr_kernel_neon_start:
@ input: @ input:
@ r3 = int len @ r3 = int len
@ -156,8 +156,8 @@ xcorr_kernel_neon_process1:
.size xcorr_kernel_neon, .-xcorr_kernel_neon @ ENDP .size xcorr_kernel_neon, .-xcorr_kernel_neon @ ENDP
@ opus_val32 celt_pitch_xcorr_neon(opus_val16 *_x, opus_val16 *_y, @ opus_val32 celt_pitch_xcorr_neon(opus_val16 *_x, opus_val16 *_y,
@ opus_val32 *xcorr, int len, int max_pitch) @ opus_val32 *xcorr, int len, int max_pitch, int arch)
; celt_pitch_xcorr_neon: @ PROC .type celt_pitch_xcorr_neon, %function; celt_pitch_xcorr_neon: @ PROC
@ input: @ input:
@ r0 = opus_val16 *_x @ r0 = opus_val16 *_x
@ r1 = opus_val16 *_y @ r1 = opus_val16 *_y
@ -171,6 +171,8 @@ xcorr_kernel_neon_process1:
@ r6 = int max_pitch @ r6 = int max_pitch
@ r12 = int j @ r12 = int j
@ q15 = int maxcorr[4] (q15 is not used by xcorr_kernel_neon()) @ q15 = int maxcorr[4] (q15 is not used by xcorr_kernel_neon())
@ ignored:
@ int arch
STMFD sp!, {r4-r6, lr} STMFD sp!, {r4-r6, lr}
LDR r6, [sp, #16] LDR r6, [sp, #16]
VMOV.S32 q15, #1 VMOV.S32 q15, #1
@ -260,7 +262,7 @@ celt_pitch_xcorr_neon_done:
@ This will get used on ARMv7 devices without NEON, so it has been optimized @ This will get used on ARMv7 devices without NEON, so it has been optimized
@ to take advantage of dual-issuing where possible. @ to take advantage of dual-issuing where possible.
; xcorr_kernel_edsp: @ PROC .type xcorr_kernel_edsp, %function; xcorr_kernel_edsp: @ PROC
xcorr_kernel_edsp_start: xcorr_kernel_edsp_start:
@ input: @ input:
@ r3 = int len @ r3 = int len
@ -344,7 +346,7 @@ xcorr_kernel_edsp_done:
LDMFD sp!, {r2,r4,r5,pc} LDMFD sp!, {r2,r4,r5,pc}
.size xcorr_kernel_edsp, .-xcorr_kernel_edsp @ ENDP .size xcorr_kernel_edsp, .-xcorr_kernel_edsp @ ENDP
; celt_pitch_xcorr_edsp: @ PROC .type celt_pitch_xcorr_edsp, %function; celt_pitch_xcorr_edsp: @ PROC
@ input: @ input:
@ r0 = opus_val16 *_x (must be 32-bit aligned) @ r0 = opus_val16 *_x (must be 32-bit aligned)
@ r1 = opus_val16 *_y (only needs to be 16-bit aligned) @ r1 = opus_val16 *_y (only needs to be 16-bit aligned)
@ -361,6 +363,8 @@ xcorr_kernel_edsp_done:
@ r9 = opus_val32 sum3 @ r9 = opus_val32 sum3
@ r1 = int max_pitch @ r1 = int max_pitch
@ r12 = int j @ r12 = int j
@ ignored:
@ int arch
STMFD sp!, {r4-r11, lr} STMFD sp!, {r4-r11, lr}
MOV r5, r1 MOV r5, r1
LDR r1, [sp, #36] LDR r1, [sp, #36]

View File

@ -153,7 +153,7 @@ xcorr_kernel_neon_process1
ENDP ENDP
; opus_val32 celt_pitch_xcorr_neon(opus_val16 *_x, opus_val16 *_y, ; opus_val32 celt_pitch_xcorr_neon(opus_val16 *_x, opus_val16 *_y,
; opus_val32 *xcorr, int len, int max_pitch) ; opus_val32 *xcorr, int len, int max_pitch, int arch)
celt_pitch_xcorr_neon PROC celt_pitch_xcorr_neon PROC
; input: ; input:
; r0 = opus_val16 *_x ; r0 = opus_val16 *_x
@ -168,6 +168,8 @@ celt_pitch_xcorr_neon PROC
; r6 = int max_pitch ; r6 = int max_pitch
; r12 = int j ; r12 = int j
; q15 = int maxcorr[4] (q15 is not used by xcorr_kernel_neon()) ; q15 = int maxcorr[4] (q15 is not used by xcorr_kernel_neon())
; ignored:
; int arch
STMFD sp!, {r4-r6, lr} STMFD sp!, {r4-r6, lr}
LDR r6, [sp, #16] LDR r6, [sp, #16]
VMOV.S32 q15, #1 VMOV.S32 q15, #1
@ -358,6 +360,8 @@ celt_pitch_xcorr_edsp PROC
; r9 = opus_val32 sum3 ; r9 = opus_val32 sum3
; r1 = int max_pitch ; r1 = int max_pitch
; r12 = int j ; r12 = int j
; ignored:
; int arch
STMFD sp!, {r4-r11, lr} STMFD sp!, {r4-r11, lr}
MOV r5, r1 MOV r5, r1
LDR r1, [sp, #36] LDR r1, [sp, #36]

View File

@ -34,7 +34,6 @@
#if !defined(FFT_ARM_H) #if !defined(FFT_ARM_H)
#define FFT_ARM_H #define FFT_ARM_H
#include "config.h"
#include "kiss_fft.h" #include "kiss_fft.h"
#if defined(HAVE_ARM_NE10) #if defined(HAVE_ARM_NE10)

View File

@ -1,5 +1,4 @@
/* Copyright (c) 2008-2011 Octasic Inc. /* Copyright (C) 2015 Vidyo */
Written by Jean-Marc Valin */
/* /*
Redistribution and use in source and binary forms, with or without Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions modification, are permitted provided that the following conditions
@ -15,8 +14,8 @@
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
@ -25,19 +24,12 @@
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/ */
#ifndef _MLP_H_ #ifndef FIXED_ARM64_H
#define _MLP_H_ #define FIXED_ARM64_H
#include "arch.h" #include <arm_neon.h>
typedef struct { #undef SIG2WORD16
int layers; #define SIG2WORD16(x) (vqmovns_s32(PSHR32((x), SIG_SHIFT)))
const int *topo;
const float *weights;
} MLP;
extern const MLP net; #endif
void mlp_process(const MLP *m, const float *in, float *out);
#endif /* _MLP_H_ */

View File

@ -37,7 +37,7 @@ static OPUS_INLINE opus_val32 MULT16_32_Q16_armv4(opus_val16 a, opus_val32 b)
"#MULT16_32_Q16\n\t" "#MULT16_32_Q16\n\t"
"smull %0, %1, %2, %3\n\t" "smull %0, %1, %2, %3\n\t"
: "=&r"(rd_lo), "=&r"(rd_hi) : "=&r"(rd_lo), "=&r"(rd_hi)
: "%r"(b),"r"(a<<16) : "%r"(b),"r"(SHL32(a,16))
); );
return rd_hi; return rd_hi;
} }
@ -54,10 +54,10 @@ static OPUS_INLINE opus_val32 MULT16_32_Q15_armv4(opus_val16 a, opus_val32 b)
"#MULT16_32_Q15\n\t" "#MULT16_32_Q15\n\t"
"smull %0, %1, %2, %3\n\t" "smull %0, %1, %2, %3\n\t"
: "=&r"(rd_lo), "=&r"(rd_hi) : "=&r"(rd_lo), "=&r"(rd_hi)
: "%r"(b), "r"(a<<16) : "%r"(b), "r"(SHL32(a,16))
); );
/*We intentionally don't OR in the high bit of rd_lo for speed.*/ /*We intentionally don't OR in the high bit of rd_lo for speed.*/
return rd_hi<<1; return SHL32(rd_hi,1);
} }
#define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv4(a, b)) #define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv4(a, b))

View File

@ -59,7 +59,7 @@ static OPUS_INLINE opus_val32 MULT16_32_Q15_armv5e(opus_val16 a, opus_val32 b)
: "=r"(res) : "=r"(res)
: "r"(b), "r"(a) : "r"(b), "r"(a)
); );
return res<<1; return SHL32(res,1);
} }
#define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv5e(a, b)) #define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv5e(a, b))
@ -76,7 +76,7 @@ static OPUS_INLINE opus_val32 MAC16_32_Q15_armv5e(opus_val32 c, opus_val16 a,
"#MAC16_32_Q15\n\t" "#MAC16_32_Q15\n\t"
"smlawb %0, %1, %2, %3;\n" "smlawb %0, %1, %2, %3;\n"
: "=r"(res) : "=r"(res)
: "r"(b<<1), "r"(a), "r"(c) : "r"(SHL32(b,1)), "r"(a), "r"(c)
); );
return res; return res;
} }

View File

@ -33,7 +33,6 @@
#if !defined(MDCT_ARM_H) #if !defined(MDCT_ARM_H)
#define MDCT_ARM_H #define MDCT_ARM_H
#include "config.h"
#include "mdct.h" #include "mdct.h"
#if defined(HAVE_ARM_NE10) #if defined(HAVE_ARM_NE10)

View File

@ -0,0 +1,160 @@
/* Copyright (c) 2010 Xiph.Org Foundation
* Copyright (c) 2013 Parrot */
/*
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#if !defined(PITCH_ARM_H)
# define PITCH_ARM_H
# include "armcpu.h"
# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
opus_val32 celt_inner_prod_neon(const opus_val16 *x, const opus_val16 *y, int N);
void dual_inner_prod_neon(const opus_val16 *x, const opus_val16 *y01,
const opus_val16 *y02, int N, opus_val32 *xy1, opus_val32 *xy2);
# if !defined(OPUS_HAVE_RTCD) && defined(OPUS_ARM_PRESUME_NEON)
# define OVERRIDE_CELT_INNER_PROD (1)
# define OVERRIDE_DUAL_INNER_PROD (1)
# define celt_inner_prod(x, y, N, arch) ((void)(arch), PRESUME_NEON(celt_inner_prod)(x, y, N))
# define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) ((void)(arch), PRESUME_NEON(dual_inner_prod)(x, y01, y02, N, xy1, xy2))
# endif
# endif
# if !defined(OVERRIDE_CELT_INNER_PROD)
# if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR))
extern opus_val32 (*const CELT_INNER_PROD_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *x, const opus_val16 *y, int N);
# define OVERRIDE_CELT_INNER_PROD (1)
# define celt_inner_prod(x, y, N, arch) ((*CELT_INNER_PROD_IMPL[(arch)&OPUS_ARCHMASK])(x, y, N))
# elif defined(OPUS_ARM_PRESUME_NEON_INTR)
# define OVERRIDE_CELT_INNER_PROD (1)
# define celt_inner_prod(x, y, N, arch) ((void)(arch), celt_inner_prod_neon(x, y, N))
# endif
# endif
# if !defined(OVERRIDE_DUAL_INNER_PROD)
# if defined(OPUS_HAVE_RTCD) && (defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR))
extern void (*const DUAL_INNER_PROD_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *x,
const opus_val16 *y01, const opus_val16 *y02, int N, opus_val32 *xy1, opus_val32 *xy2);
# define OVERRIDE_DUAL_INNER_PROD (1)
# define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) ((*DUAL_INNER_PROD_IMPL[(arch)&OPUS_ARCHMASK])(x, y01, y02, N, xy1, xy2))
# elif defined(OPUS_ARM_PRESUME_NEON_INTR)
# define OVERRIDE_DUAL_INNER_PROD (1)
# define dual_inner_prod(x, y01, y02, N, xy1, xy2, arch) ((void)(arch), dual_inner_prod_neon(x, y01, y02, N, xy1, xy2))
# endif
# endif
# if defined(FIXED_POINT)
# if defined(OPUS_ARM_MAY_HAVE_NEON)
opus_val32 celt_pitch_xcorr_neon(const opus_val16 *_x, const opus_val16 *_y,
opus_val32 *xcorr, int len, int max_pitch, int arch);
# endif
# if defined(OPUS_ARM_MAY_HAVE_MEDIA)
# define celt_pitch_xcorr_media MAY_HAVE_EDSP(celt_pitch_xcorr)
# endif
# if defined(OPUS_ARM_MAY_HAVE_EDSP)
opus_val32 celt_pitch_xcorr_edsp(const opus_val16 *_x, const opus_val16 *_y,
opus_val32 *xcorr, int len, int max_pitch, int arch);
# endif
# if defined(OPUS_HAVE_RTCD) && \
((defined(OPUS_ARM_MAY_HAVE_NEON) && !defined(OPUS_ARM_PRESUME_NEON)) || \
(defined(OPUS_ARM_MAY_HAVE_MEDIA) && !defined(OPUS_ARM_PRESUME_MEDIA)) || \
(defined(OPUS_ARM_MAY_HAVE_EDSP) && !defined(OPUS_ARM_PRESUME_EDSP)))
extern opus_val32
(*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
const opus_val16 *, opus_val32 *, int, int, int);
# define OVERRIDE_PITCH_XCORR (1)
# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
((*CELT_PITCH_XCORR_IMPL[(arch)&OPUS_ARCHMASK])(_x, _y, \
xcorr, len, max_pitch, arch))
# elif defined(OPUS_ARM_PRESUME_EDSP) || \
defined(OPUS_ARM_PRESUME_MEDIA) || \
defined(OPUS_ARM_PRESUME_NEON)
# define OVERRIDE_PITCH_XCORR (1)
# define celt_pitch_xcorr (PRESUME_NEON(celt_pitch_xcorr))
# endif
# if defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
void xcorr_kernel_neon_fixed(
const opus_val16 *x,
const opus_val16 *y,
opus_val32 sum[4],
int len);
# endif
# if defined(OPUS_HAVE_RTCD) && \
(defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR))
extern void (*const XCORR_KERNEL_IMPL[OPUS_ARCHMASK + 1])(
const opus_val16 *x,
const opus_val16 *y,
opus_val32 sum[4],
int len);
# define OVERRIDE_XCORR_KERNEL (1)
# define xcorr_kernel(x, y, sum, len, arch) \
((*XCORR_KERNEL_IMPL[(arch) & OPUS_ARCHMASK])(x, y, sum, len))
# elif defined(OPUS_ARM_PRESUME_NEON_INTR)
# define OVERRIDE_XCORR_KERNEL (1)
# define xcorr_kernel(x, y, sum, len, arch) \
((void)arch, xcorr_kernel_neon_fixed(x, y, sum, len))
# endif
#else /* Start !FIXED_POINT */
/* Float case */
#if defined(OPUS_ARM_MAY_HAVE_NEON_INTR)
void celt_pitch_xcorr_float_neon(const opus_val16 *_x, const opus_val16 *_y,
opus_val32 *xcorr, int len, int max_pitch, int arch);
#endif
# if defined(OPUS_HAVE_RTCD) && \
(defined(OPUS_ARM_MAY_HAVE_NEON_INTR) && !defined(OPUS_ARM_PRESUME_NEON_INTR))
extern void
(*const CELT_PITCH_XCORR_IMPL[OPUS_ARCHMASK+1])(const opus_val16 *,
const opus_val16 *, opus_val32 *, int, int, int);
# define OVERRIDE_PITCH_XCORR (1)
# define celt_pitch_xcorr(_x, _y, xcorr, len, max_pitch, arch) \
((*CELT_PITCH_XCORR_IMPL[(arch)&OPUS_ARCHMASK])(_x, _y, \
xcorr, len, max_pitch, arch))
# elif defined(OPUS_ARM_PRESUME_NEON_INTR)
# define OVERRIDE_PITCH_XCORR (1)
# define celt_pitch_xcorr celt_pitch_xcorr_float_neon
# endif
#endif /* end !FIXED_POINT */
#endif

View File

@ -0,0 +1,290 @@
/***********************************************************************
Copyright (c) 2017 Google Inc.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
- Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name of Internet Society, IETF or IETF Trust, nor the
names of specific contributors, may be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
***********************************************************************/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <arm_neon.h>
#include "pitch.h"
#ifdef FIXED_POINT
opus_val32 celt_inner_prod_neon(const opus_val16 *x, const opus_val16 *y, int N)
{
int i;
opus_val32 xy;
int16x8_t x_s16x8, y_s16x8;
int32x4_t xy_s32x4 = vdupq_n_s32(0);
int64x2_t xy_s64x2;
int64x1_t xy_s64x1;
for (i = 0; i < N - 7; i += 8) {
x_s16x8 = vld1q_s16(&x[i]);
y_s16x8 = vld1q_s16(&y[i]);
xy_s32x4 = vmlal_s16(xy_s32x4, vget_low_s16 (x_s16x8), vget_low_s16 (y_s16x8));
xy_s32x4 = vmlal_s16(xy_s32x4, vget_high_s16(x_s16x8), vget_high_s16(y_s16x8));
}
if (N - i >= 4) {
const int16x4_t x_s16x4 = vld1_s16(&x[i]);
const int16x4_t y_s16x4 = vld1_s16(&y[i]);
xy_s32x4 = vmlal_s16(xy_s32x4, x_s16x4, y_s16x4);
i += 4;
}
xy_s64x2 = vpaddlq_s32(xy_s32x4);
xy_s64x1 = vadd_s64(vget_low_s64(xy_s64x2), vget_high_s64(xy_s64x2));
xy = vget_lane_s32(vreinterpret_s32_s64(xy_s64x1), 0);
for (; i < N; i++) {
xy = MAC16_16(xy, x[i], y[i]);
}
#ifdef OPUS_CHECK_ASM
celt_assert(celt_inner_prod_c(x, y, N) == xy);
#endif
return xy;
}
void dual_inner_prod_neon(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02,
int N, opus_val32 *xy1, opus_val32 *xy2)
{
int i;
opus_val32 xy01, xy02;
int16x8_t x_s16x8, y01_s16x8, y02_s16x8;
int32x4_t xy01_s32x4 = vdupq_n_s32(0);
int32x4_t xy02_s32x4 = vdupq_n_s32(0);
int64x2_t xy01_s64x2, xy02_s64x2;
int64x1_t xy01_s64x1, xy02_s64x1;
for (i = 0; i < N - 7; i += 8) {
x_s16x8 = vld1q_s16(&x[i]);
y01_s16x8 = vld1q_s16(&y01[i]);
y02_s16x8 = vld1q_s16(&y02[i]);
xy01_s32x4 = vmlal_s16(xy01_s32x4, vget_low_s16 (x_s16x8), vget_low_s16 (y01_s16x8));
xy02_s32x4 = vmlal_s16(xy02_s32x4, vget_low_s16 (x_s16x8), vget_low_s16 (y02_s16x8));
xy01_s32x4 = vmlal_s16(xy01_s32x4, vget_high_s16(x_s16x8), vget_high_s16(y01_s16x8));
xy02_s32x4 = vmlal_s16(xy02_s32x4, vget_high_s16(x_s16x8), vget_high_s16(y02_s16x8));
}
if (N - i >= 4) {
const int16x4_t x_s16x4 = vld1_s16(&x[i]);
const int16x4_t y01_s16x4 = vld1_s16(&y01[i]);
const int16x4_t y02_s16x4 = vld1_s16(&y02[i]);
xy01_s32x4 = vmlal_s16(xy01_s32x4, x_s16x4, y01_s16x4);
xy02_s32x4 = vmlal_s16(xy02_s32x4, x_s16x4, y02_s16x4);
i += 4;
}
xy01_s64x2 = vpaddlq_s32(xy01_s32x4);
xy02_s64x2 = vpaddlq_s32(xy02_s32x4);
xy01_s64x1 = vadd_s64(vget_low_s64(xy01_s64x2), vget_high_s64(xy01_s64x2));
xy02_s64x1 = vadd_s64(vget_low_s64(xy02_s64x2), vget_high_s64(xy02_s64x2));
xy01 = vget_lane_s32(vreinterpret_s32_s64(xy01_s64x1), 0);
xy02 = vget_lane_s32(vreinterpret_s32_s64(xy02_s64x1), 0);
for (; i < N; i++) {
xy01 = MAC16_16(xy01, x[i], y01[i]);
xy02 = MAC16_16(xy02, x[i], y02[i]);
}
*xy1 = xy01;
*xy2 = xy02;
#ifdef OPUS_CHECK_ASM
{
opus_val32 xy1_c, xy2_c;
dual_inner_prod_c(x, y01, y02, N, &xy1_c, &xy2_c);
celt_assert(xy1_c == *xy1);
celt_assert(xy2_c == *xy2);
}
#endif
}
#else /* !FIXED_POINT */
/* ========================================================================== */
#ifdef OPUS_CHECK_ASM
/* This part of code simulates floating-point NEON operations. */
/* celt_inner_prod_neon_float_c_simulation() simulates the floating-point */
/* operations of celt_inner_prod_neon(), and both functions should have bit */
/* exact output. */
static opus_val32 celt_inner_prod_neon_float_c_simulation(const opus_val16 *x, const opus_val16 *y, int N)
{
int i;
opus_val32 xy, xy0 = 0, xy1 = 0, xy2 = 0, xy3 = 0;
for (i = 0; i < N - 3; i += 4) {
xy0 = MAC16_16(xy0, x[i + 0], y[i + 0]);
xy1 = MAC16_16(xy1, x[i + 1], y[i + 1]);
xy2 = MAC16_16(xy2, x[i + 2], y[i + 2]);
xy3 = MAC16_16(xy3, x[i + 3], y[i + 3]);
}
xy0 += xy2;
xy1 += xy3;
xy = xy0 + xy1;
for (; i < N; i++) {
xy = MAC16_16(xy, x[i], y[i]);
}
return xy;
}
/* dual_inner_prod_neon_float_c_simulation() simulates the floating-point */
/* operations of dual_inner_prod_neon(), and both functions should have bit */
/* exact output. */
static void dual_inner_prod_neon_float_c_simulation(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02,
int N, opus_val32 *xy1, opus_val32 *xy2)
{
int i;
opus_val32 xy01, xy02, xy01_0 = 0, xy01_1 = 0, xy01_2 = 0, xy01_3 = 0, xy02_0 = 0, xy02_1 = 0, xy02_2 = 0, xy02_3 = 0;
for (i = 0; i < N - 3; i += 4) {
xy01_0 = MAC16_16(xy01_0, x[i + 0], y01[i + 0]);
xy01_1 = MAC16_16(xy01_1, x[i + 1], y01[i + 1]);
xy01_2 = MAC16_16(xy01_2, x[i + 2], y01[i + 2]);
xy01_3 = MAC16_16(xy01_3, x[i + 3], y01[i + 3]);
xy02_0 = MAC16_16(xy02_0, x[i + 0], y02[i + 0]);
xy02_1 = MAC16_16(xy02_1, x[i + 1], y02[i + 1]);
xy02_2 = MAC16_16(xy02_2, x[i + 2], y02[i + 2]);
xy02_3 = MAC16_16(xy02_3, x[i + 3], y02[i + 3]);
}
xy01_0 += xy01_2;
xy02_0 += xy02_2;
xy01_1 += xy01_3;
xy02_1 += xy02_3;
xy01 = xy01_0 + xy01_1;
xy02 = xy02_0 + xy02_1;
for (; i < N; i++) {
xy01 = MAC16_16(xy01, x[i], y01[i]);
xy02 = MAC16_16(xy02, x[i], y02[i]);
}
*xy1 = xy01;
*xy2 = xy02;
}
#endif /* OPUS_CHECK_ASM */
/* ========================================================================== */
opus_val32 celt_inner_prod_neon(const opus_val16 *x, const opus_val16 *y, int N)
{
int i;
opus_val32 xy;
float32x4_t xy_f32x4 = vdupq_n_f32(0);
float32x2_t xy_f32x2;
for (i = 0; i < N - 7; i += 8) {
float32x4_t x_f32x4, y_f32x4;
x_f32x4 = vld1q_f32(&x[i]);
y_f32x4 = vld1q_f32(&y[i]);
xy_f32x4 = vmlaq_f32(xy_f32x4, x_f32x4, y_f32x4);
x_f32x4 = vld1q_f32(&x[i + 4]);
y_f32x4 = vld1q_f32(&y[i + 4]);
xy_f32x4 = vmlaq_f32(xy_f32x4, x_f32x4, y_f32x4);
}
if (N - i >= 4) {
const float32x4_t x_f32x4 = vld1q_f32(&x[i]);
const float32x4_t y_f32x4 = vld1q_f32(&y[i]);
xy_f32x4 = vmlaq_f32(xy_f32x4, x_f32x4, y_f32x4);
i += 4;
}
xy_f32x2 = vadd_f32(vget_low_f32(xy_f32x4), vget_high_f32(xy_f32x4));
xy_f32x2 = vpadd_f32(xy_f32x2, xy_f32x2);
xy = vget_lane_f32(xy_f32x2, 0);
for (; i < N; i++) {
xy = MAC16_16(xy, x[i], y[i]);
}
#ifdef OPUS_CHECK_ASM
celt_assert(ABS32(celt_inner_prod_neon_float_c_simulation(x, y, N) - xy) <= VERY_SMALL);
#endif
return xy;
}
void dual_inner_prod_neon(const opus_val16 *x, const opus_val16 *y01, const opus_val16 *y02,
int N, opus_val32 *xy1, opus_val32 *xy2)
{
int i;
opus_val32 xy01, xy02;
float32x4_t xy01_f32x4 = vdupq_n_f32(0);
float32x4_t xy02_f32x4 = vdupq_n_f32(0);
float32x2_t xy01_f32x2, xy02_f32x2;
for (i = 0; i < N - 7; i += 8) {
float32x4_t x_f32x4, y01_f32x4, y02_f32x4;
x_f32x4 = vld1q_f32(&x[i]);
y01_f32x4 = vld1q_f32(&y01[i]);
y02_f32x4 = vld1q_f32(&y02[i]);
xy01_f32x4 = vmlaq_f32(xy01_f32x4, x_f32x4, y01_f32x4);
xy02_f32x4 = vmlaq_f32(xy02_f32x4, x_f32x4, y02_f32x4);
x_f32x4 = vld1q_f32(&x[i + 4]);
y01_f32x4 = vld1q_f32(&y01[i + 4]);
y02_f32x4 = vld1q_f32(&y02[i + 4]);
xy01_f32x4 = vmlaq_f32(xy01_f32x4, x_f32x4, y01_f32x4);
xy02_f32x4 = vmlaq_f32(xy02_f32x4, x_f32x4, y02_f32x4);
}
if (N - i >= 4) {
const float32x4_t x_f32x4 = vld1q_f32(&x[i]);
const float32x4_t y01_f32x4 = vld1q_f32(&y01[i]);
const float32x4_t y02_f32x4 = vld1q_f32(&y02[i]);
xy01_f32x4 = vmlaq_f32(xy01_f32x4, x_f32x4, y01_f32x4);
xy02_f32x4 = vmlaq_f32(xy02_f32x4, x_f32x4, y02_f32x4);
i += 4;
}
xy01_f32x2 = vadd_f32(vget_low_f32(xy01_f32x4), vget_high_f32(xy01_f32x4));
xy02_f32x2 = vadd_f32(vget_low_f32(xy02_f32x4), vget_high_f32(xy02_f32x4));
xy01_f32x2 = vpadd_f32(xy01_f32x2, xy01_f32x2);
xy02_f32x2 = vpadd_f32(xy02_f32x2, xy02_f32x2);
xy01 = vget_lane_f32(xy01_f32x2, 0);
xy02 = vget_lane_f32(xy02_f32x2, 0);
for (; i < N; i++) {
xy01 = MAC16_16(xy01, x[i], y01[i]);
xy02 = MAC16_16(xy02, x[i], y02[i]);
}
*xy1 = xy01;
*xy2 = xy02;
#ifdef OPUS_CHECK_ASM
{
opus_val32 xy1_c, xy2_c;
dual_inner_prod_neon_float_c_simulation(x, y01, y02, N, &xy1_c, &xy2_c);
celt_assert(ABS32(xy1_c - *xy1) <= VERY_SMALL);
celt_assert(ABS32(xy2_c - *xy2) <= VERY_SMALL);
}
#endif
}
#endif /* FIXED_POINT */

View File

@ -65,19 +65,19 @@ opus_uint32 celt_lcg_rand(opus_uint32 seed)
/* This is a cos() approximation designed to be bit-exact on any platform. Bit exactness /* This is a cos() approximation designed to be bit-exact on any platform. Bit exactness
with this approximation is important because it has an impact on the bit allocation */ with this approximation is important because it has an impact on the bit allocation */
static opus_int16 bitexact_cos(opus_int16 x) opus_int16 bitexact_cos(opus_int16 x)
{ {
opus_int32 tmp; opus_int32 tmp;
opus_int16 x2; opus_int16 x2;
tmp = (4096+((opus_int32)(x)*(x)))>>13; tmp = (4096+((opus_int32)(x)*(x)))>>13;
celt_assert(tmp<=32767); celt_sig_assert(tmp<=32767);
x2 = tmp; x2 = tmp;
x2 = (32767-x2) + FRAC_MUL16(x2, (-7651 + FRAC_MUL16(x2, (8277 + FRAC_MUL16(-626, x2))))); x2 = (32767-x2) + FRAC_MUL16(x2, (-7651 + FRAC_MUL16(x2, (8277 + FRAC_MUL16(-626, x2)))));
celt_assert(x2<=32766); celt_sig_assert(x2<=32766);
return 1+x2; return 1+x2;
} }
static int bitexact_log2tan(int isin,int icos) int bitexact_log2tan(int isin,int icos)
{ {
int lc; int lc;
int ls; int ls;
@ -92,10 +92,11 @@ static int bitexact_log2tan(int isin,int icos)
#ifdef FIXED_POINT #ifdef FIXED_POINT
/* Compute the amplitude (sqrt energy) in each of the bands */ /* Compute the amplitude (sqrt energy) in each of the bands */
void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM) void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM, int arch)
{ {
int i, c, N; int i, c, N;
const opus_int16 *eBands = m->eBands; const opus_int16 *eBands = m->eBands;
(void)arch;
N = m->shortMdctSize<<LM; N = m->shortMdctSize<<LM;
c=0; do { c=0; do {
for (i=0;i<end;i++) for (i=0;i<end;i++)
@ -155,7 +156,7 @@ void normalise_bands(const CELTMode *m, const celt_sig * OPUS_RESTRICT freq, cel
#else /* FIXED_POINT */ #else /* FIXED_POINT */
/* Compute the amplitude (sqrt energy) in each of the bands */ /* Compute the amplitude (sqrt energy) in each of the bands */
void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM) void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM, int arch)
{ {
int i, c, N; int i, c, N;
const opus_int16 *eBands = m->eBands; const opus_int16 *eBands = m->eBands;
@ -164,7 +165,7 @@ void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *band
for (i=0;i<end;i++) for (i=0;i<end;i++)
{ {
opus_val32 sum; opus_val32 sum;
sum = 1e-27f + celt_inner_prod_c(&X[c*N+(eBands[i]<<LM)], &X[c*N+(eBands[i]<<LM)], (eBands[i+1]-eBands[i])<<LM); sum = 1e-27f + celt_inner_prod(&X[c*N+(eBands[i]<<LM)], &X[c*N+(eBands[i]<<LM)], (eBands[i+1]-eBands[i])<<LM, arch);
bandE[i+c*m->nbEBands] = celt_sqrt(sum); bandE[i+c*m->nbEBands] = celt_sqrt(sum);
/*printf ("%f ", bandE[i+c*m->nbEBands]);*/ /*printf ("%f ", bandE[i+c*m->nbEBands]);*/
} }
@ -224,9 +225,9 @@ void denormalise_bands(const CELTMode *m, const celt_norm * OPUS_RESTRICT X,
#endif #endif
j=M*eBands[i]; j=M*eBands[i];
band_end = M*eBands[i+1]; band_end = M*eBands[i+1];
lg = ADD16(bandLogE[i], SHL16((opus_val16)eMeans[i],6)); lg = SATURATE16(ADD32(bandLogE[i], SHL32((opus_val32)eMeans[i],6)));
#ifndef FIXED_POINT #ifndef FIXED_POINT
g = celt_exp2(lg); g = celt_exp2(MIN32(32.f, lg));
#else #else
/* Handle the integer part of the log energy */ /* Handle the integer part of the log energy */
shift = 16-(lg>>DB_SHIFT); shift = 16-(lg>>DB_SHIFT);
@ -241,12 +242,12 @@ void denormalise_bands(const CELTMode *m, const celt_norm * OPUS_RESTRICT X,
/* Handle extreme gains with negative shift. */ /* Handle extreme gains with negative shift. */
if (shift<0) if (shift<0)
{ {
/* For shift < -2 we'd be likely to overflow, so we're capping /* For shift <= -2 and g > 16384 we'd be likely to overflow, so we're
the gain here. This shouldn't happen unless the bitstream is capping the gain here, which is equivalent to a cap of 18 on lg.
already corrupted. */ This shouldn't trigger unless the bitstream is already corrupted. */
if (shift < -2) if (shift <= -2)
{ {
g = 32767; g = 16384;
shift = -2; shift = -2;
} }
do { do {
@ -281,7 +282,7 @@ void anti_collapse(const CELTMode *m, celt_norm *X_, unsigned char *collapse_mas
N0 = m->eBands[i+1]-m->eBands[i]; N0 = m->eBands[i+1]-m->eBands[i];
/* depth in 1/8 bits */ /* depth in 1/8 bits */
celt_assert(pulses[i]>=0); celt_sig_assert(pulses[i]>=0);
depth = celt_udiv(1+pulses[i], (m->eBands[i+1]-m->eBands[i]))>>LM; depth = celt_udiv(1+pulses[i], (m->eBands[i+1]-m->eBands[i]))>>LM;
#ifdef FIXED_POINT #ifdef FIXED_POINT
@ -360,6 +361,30 @@ void anti_collapse(const CELTMode *m, celt_norm *X_, unsigned char *collapse_mas
} }
} }
/* Compute the weights to use for optimizing normalized distortion across
channels. We use the amplitude to weight square distortion, which means
that we use the square root of the value we would have been using if we
wanted to minimize the MSE in the non-normalized domain. This roughly
corresponds to some quick-and-dirty perceptual experiments I ran to
measure inter-aural masking (there doesn't seem to be any published data
on the topic). */
static void compute_channel_weights(celt_ener Ex, celt_ener Ey, opus_val16 w[2])
{
celt_ener minE;
#ifdef FIXED_POINT
int shift;
#endif
minE = MIN32(Ex, Ey);
/* Adjustment to make the weights a bit more conservative. */
Ex = ADD32(Ex, minE/3);
Ey = ADD32(Ey, minE/3);
#ifdef FIXED_POINT
shift = celt_ilog2(EPSILON+MAX32(Ex, Ey))-14;
#endif
w[0] = VSHR32(Ex, shift);
w[1] = VSHR32(Ey, shift);
}
static void intensity_stereo(const CELTMode *m, celt_norm * OPUS_RESTRICT X, const celt_norm * OPUS_RESTRICT Y, const celt_ener *bandE, int bandID, int N) static void intensity_stereo(const CELTMode *m, celt_norm * OPUS_RESTRICT X, const celt_norm * OPUS_RESTRICT Y, const celt_ener *bandE, int bandID, int N)
{ {
int i = bandID; int i = bandID;
@ -414,7 +439,7 @@ static void stereo_merge(celt_norm * OPUS_RESTRICT X, celt_norm * OPUS_RESTRICT
/* Compensating for the mid normalization */ /* Compensating for the mid normalization */
xp = MULT16_32_Q15(mid, xp); xp = MULT16_32_Q15(mid, xp);
/* mid and side are in Q15, not Q14 like X and Y */ /* mid and side are in Q15, not Q14 like X and Y */
mid2 = SHR32(mid, 1); mid2 = SHR16(mid, 1);
El = MULT16_16(mid2, mid2) + side - 2*xp; El = MULT16_16(mid2, mid2) + side - 2*xp;
Er = MULT16_16(mid2, mid2) + side + 2*xp; Er = MULT16_16(mid2, mid2) + side + 2*xp;
if (Er < QCONST32(6e-4f, 28) || El < QCONST32(6e-4f, 28)) if (Er < QCONST32(6e-4f, 28) || El < QCONST32(6e-4f, 28))
@ -453,7 +478,7 @@ static void stereo_merge(celt_norm * OPUS_RESTRICT X, celt_norm * OPUS_RESTRICT
/* Decide whether we should spread the pulses in the current frame */ /* Decide whether we should spread the pulses in the current frame */
int spreading_decision(const CELTMode *m, const celt_norm *X, int *average, int spreading_decision(const CELTMode *m, const celt_norm *X, int *average,
int last_decision, int *hf_average, int *tapset_decision, int update_hf, int last_decision, int *hf_average, int *tapset_decision, int update_hf,
int end, int C, int M) int end, int C, int M, const int *spread_weight)
{ {
int i, c, N0; int i, c, N0;
int sum = 0, nbBands=0; int sum = 0, nbBands=0;
@ -494,8 +519,8 @@ int spreading_decision(const CELTMode *m, const celt_norm *X, int *average,
if (i>m->nbEBands-4) if (i>m->nbEBands-4)
hf_sum += celt_udiv(32*(tcount[1]+tcount[0]), N); hf_sum += celt_udiv(32*(tcount[1]+tcount[0]), N);
tmp = (2*tcount[2] >= N) + (2*tcount[1] >= N) + (2*tcount[0] >= N); tmp = (2*tcount[2] >= N) + (2*tcount[1] >= N) + (2*tcount[0] >= N);
sum += tmp*256; sum += tmp*spread_weight[i];
nbBands++; nbBands+=spread_weight[i];
} }
} while (++c<C); } while (++c<C);
@ -519,7 +544,7 @@ int spreading_decision(const CELTMode *m, const celt_norm *X, int *average,
/*printf("%d %d %d\n", hf_sum, *hf_average, *tapset_decision);*/ /*printf("%d %d %d\n", hf_sum, *hf_average, *tapset_decision);*/
celt_assert(nbBands>0); /* end has to be non-zero */ celt_assert(nbBands>0); /* end has to be non-zero */
celt_assert(sum>=0); celt_assert(sum>=0);
sum = celt_udiv(sum, nbBands); sum = celt_udiv((opus_int32)sum<<8, nbBands);
/* Recursive averaging */ /* Recursive averaging */
sum = (sum+*average)>>1; sum = (sum+*average)>>1;
*average = sum; *average = sum;
@ -647,6 +672,7 @@ static int compute_qn(int N, int b, int offset, int pulse_cap, int stereo)
struct band_ctx { struct band_ctx {
int encode; int encode;
int resynth;
const CELTMode *m; const CELTMode *m;
int i; int i;
int intensity; int intensity;
@ -657,6 +683,9 @@ struct band_ctx {
const celt_ener *bandE; const celt_ener *bandE;
opus_uint32 seed; opus_uint32 seed;
int arch; int arch;
int theta_round;
int disable_inv;
int avoid_split_noise;
}; };
struct split_ctx { struct split_ctx {
@ -714,8 +743,35 @@ static void compute_theta(struct band_ctx *ctx, struct split_ctx *sctx,
if (qn!=1) if (qn!=1)
{ {
if (encode) if (encode)
itheta = (itheta*qn+8192)>>14; {
if (!stereo || ctx->theta_round == 0)
{
itheta = (itheta*(opus_int32)qn+8192)>>14;
if (!stereo && ctx->avoid_split_noise && itheta > 0 && itheta < qn)
{
/* Check if the selected value of theta will cause the bit allocation
to inject noise on one side. If so, make sure the energy of that side
is zero. */
int unquantized = celt_udiv((opus_int32)itheta*16384, qn);
imid = bitexact_cos((opus_int16)unquantized);
iside = bitexact_cos((opus_int16)(16384-unquantized));
delta = FRAC_MUL16((N-1)<<7,bitexact_log2tan(iside,imid));
if (delta > *b)
itheta = qn;
else if (delta < -*b)
itheta = 0;
}
} else {
int down;
/* Bias quantization towards itheta=0 and itheta=16384. */
int bias = itheta > 8192 ? 32767/qn : -32767/qn;
down = IMIN(qn-1, IMAX(0, (itheta*(opus_int32)qn + bias)>>14));
if (ctx->theta_round < 0)
itheta = down;
else
itheta = down+1;
}
}
/* Entropy coding of the angle. We use a uniform pdf for the /* Entropy coding of the angle. We use a uniform pdf for the
time split, a step for stereo, and a triangular one for the rest. */ time split, a step for stereo, and a triangular one for the rest. */
if (stereo && N>2) if (stereo && N>2)
@ -793,7 +849,7 @@ static void compute_theta(struct band_ctx *ctx, struct split_ctx *sctx,
} else if (stereo) { } else if (stereo) {
if (encode) if (encode)
{ {
inv = itheta > 8192; inv = itheta > 8192 && !ctx->disable_inv;
if (inv) if (inv)
{ {
int j; int j;
@ -810,6 +866,9 @@ static void compute_theta(struct band_ctx *ctx, struct split_ctx *sctx,
inv = ec_dec_bit_logp(ec, 2); inv = ec_dec_bit_logp(ec, 2);
} else } else
inv = 0; inv = 0;
/* inv flag override to avoid problems with downmixing. */
if (ctx->disable_inv)
inv = 0;
itheta = 0; itheta = 0;
} }
qalloc = ec_tell_frac(ec) - tell; qalloc = ec_tell_frac(ec) - tell;
@ -845,11 +904,6 @@ static void compute_theta(struct band_ctx *ctx, struct split_ctx *sctx,
static unsigned quant_band_n1(struct band_ctx *ctx, celt_norm *X, celt_norm *Y, int b, static unsigned quant_band_n1(struct band_ctx *ctx, celt_norm *X, celt_norm *Y, int b,
celt_norm *lowband_out) celt_norm *lowband_out)
{ {
#ifdef RESYNTH
int resynth = 1;
#else
int resynth = !ctx->encode;
#endif
int c; int c;
int stereo; int stereo;
celt_norm *x = X; celt_norm *x = X;
@ -874,7 +928,7 @@ static unsigned quant_band_n1(struct band_ctx *ctx, celt_norm *X, celt_norm *Y,
ctx->remaining_bits -= 1<<BITRES; ctx->remaining_bits -= 1<<BITRES;
b-=1<<BITRES; b-=1<<BITRES;
} }
if (resynth) if (ctx->resynth)
x[0] = sign ? -NORM_SCALING : NORM_SCALING; x[0] = sign ? -NORM_SCALING : NORM_SCALING;
x = Y; x = Y;
} while (++c<1+stereo); } while (++c<1+stereo);
@ -899,11 +953,6 @@ static unsigned quant_partition(struct band_ctx *ctx, celt_norm *X,
int B0=B; int B0=B;
opus_val16 mid=0, side=0; opus_val16 mid=0, side=0;
unsigned cm=0; unsigned cm=0;
#ifdef RESYNTH
int resynth = 1;
#else
int resynth = !ctx->encode;
#endif
celt_norm *Y=NULL; celt_norm *Y=NULL;
int encode; int encode;
const CELTMode *m; const CELTMode *m;
@ -935,8 +984,7 @@ static unsigned quant_partition(struct band_ctx *ctx, celt_norm *X,
fill = (fill&1)|(fill<<1); fill = (fill&1)|(fill<<1);
B = (B+1)>>1; B = (B+1)>>1;
compute_theta(ctx, &sctx, X, Y, N, &b, B, B0, compute_theta(ctx, &sctx, X, Y, N, &b, B, B0, LM, 0, &fill);
LM, 0, &fill);
imid = sctx.imid; imid = sctx.imid;
iside = sctx.iside; iside = sctx.iside;
delta = sctx.delta; delta = sctx.delta;
@ -970,24 +1018,20 @@ static unsigned quant_partition(struct band_ctx *ctx, celt_norm *X,
rebalance = ctx->remaining_bits; rebalance = ctx->remaining_bits;
if (mbits >= sbits) if (mbits >= sbits)
{ {
cm = quant_partition(ctx, X, N, mbits, B, cm = quant_partition(ctx, X, N, mbits, B, lowband, LM,
lowband, LM,
MULT16_16_P15(gain,mid), fill); MULT16_16_P15(gain,mid), fill);
rebalance = mbits - (rebalance-ctx->remaining_bits); rebalance = mbits - (rebalance-ctx->remaining_bits);
if (rebalance > 3<<BITRES && itheta!=0) if (rebalance > 3<<BITRES && itheta!=0)
sbits += rebalance - (3<<BITRES); sbits += rebalance - (3<<BITRES);
cm |= quant_partition(ctx, Y, N, sbits, B, cm |= quant_partition(ctx, Y, N, sbits, B, next_lowband2, LM,
next_lowband2, LM,
MULT16_16_P15(gain,side), fill>>B)<<(B0>>1); MULT16_16_P15(gain,side), fill>>B)<<(B0>>1);
} else { } else {
cm = quant_partition(ctx, Y, N, sbits, B, cm = quant_partition(ctx, Y, N, sbits, B, next_lowband2, LM,
next_lowband2, LM,
MULT16_16_P15(gain,side), fill>>B)<<(B0>>1); MULT16_16_P15(gain,side), fill>>B)<<(B0>>1);
rebalance = sbits - (rebalance-ctx->remaining_bits); rebalance = sbits - (rebalance-ctx->remaining_bits);
if (rebalance > 3<<BITRES && itheta!=16384) if (rebalance > 3<<BITRES && itheta!=16384)
mbits += rebalance - (3<<BITRES); mbits += rebalance - (3<<BITRES);
cm |= quant_partition(ctx, X, N, mbits, B, cm |= quant_partition(ctx, X, N, mbits, B, lowband, LM,
lowband, LM,
MULT16_16_P15(gain,mid), fill); MULT16_16_P15(gain,mid), fill);
} }
} else { } else {
@ -1012,18 +1056,14 @@ static unsigned quant_partition(struct band_ctx *ctx, celt_norm *X,
/* Finally do the actual quantization */ /* Finally do the actual quantization */
if (encode) if (encode)
{ {
cm = alg_quant(X, N, K, spread, B, ec cm = alg_quant(X, N, K, spread, B, ec, gain, ctx->resynth, ctx->arch);
#ifdef RESYNTH
, gain
#endif
);
} else { } else {
cm = alg_unquant(X, N, K, spread, B, ec, gain); cm = alg_unquant(X, N, K, spread, B, ec, gain);
} }
} else { } else {
/* If there's no pulse, fill the band anyway */ /* If there's no pulse, fill the band anyway */
int j; int j;
if (resynth) if (ctx->resynth)
{ {
unsigned cm_mask; unsigned cm_mask;
/* B can be as large as 16, so this shift might overflow an int on a /* B can be as large as 16, so this shift might overflow an int on a
@ -1080,11 +1120,6 @@ static unsigned quant_band(struct band_ctx *ctx, celt_norm *X,
int recombine=0; int recombine=0;
int longBlocks; int longBlocks;
unsigned cm=0; unsigned cm=0;
#ifdef RESYNTH
int resynth = 1;
#else
int resynth = !ctx->encode;
#endif
int k; int k;
int encode; int encode;
int tf_change; int tf_change;
@ -1151,11 +1186,10 @@ static unsigned quant_band(struct band_ctx *ctx, celt_norm *X,
deinterleave_hadamard(lowband, N_B>>recombine, B0<<recombine, longBlocks); deinterleave_hadamard(lowband, N_B>>recombine, B0<<recombine, longBlocks);
} }
cm = quant_partition(ctx, X, N, b, B, lowband, cm = quant_partition(ctx, X, N, b, B, lowband, LM, gain, fill);
LM, gain, fill);
/* This code is used by the decoder and by the resynthesis-enabled encoder */ /* This code is used by the decoder and by the resynthesis-enabled encoder */
if (resynth) if (ctx->resynth)
{ {
/* Undo the sample reorganization going from time order to frequency order */ /* Undo the sample reorganization going from time order to frequency order */
if (B0>1) if (B0>1)
@ -1208,11 +1242,6 @@ static unsigned quant_band_stereo(struct band_ctx *ctx, celt_norm *X, celt_norm
int inv = 0; int inv = 0;
opus_val16 mid=0, side=0; opus_val16 mid=0, side=0;
unsigned cm=0; unsigned cm=0;
#ifdef RESYNTH
int resynth = 1;
#else
int resynth = !ctx->encode;
#endif
int mbits, sbits, delta; int mbits, sbits, delta;
int itheta; int itheta;
int qalloc; int qalloc;
@ -1232,8 +1261,7 @@ static unsigned quant_band_stereo(struct band_ctx *ctx, celt_norm *X, celt_norm
orig_fill = fill; orig_fill = fill;
compute_theta(ctx, &sctx, X, Y, N, &b, B, B, compute_theta(ctx, &sctx, X, Y, N, &b, B, B, LM, 1, &fill);
LM, 1, &fill);
inv = sctx.inv; inv = sctx.inv;
imid = sctx.imid; imid = sctx.imid;
iside = sctx.iside; iside = sctx.iside;
@ -1281,13 +1309,13 @@ static unsigned quant_band_stereo(struct band_ctx *ctx, celt_norm *X, celt_norm
sign = 1-2*sign; sign = 1-2*sign;
/* We use orig_fill here because we want to fold the side, but if /* We use orig_fill here because we want to fold the side, but if
itheta==16384, we'll have cleared the low bits of fill. */ itheta==16384, we'll have cleared the low bits of fill. */
cm = quant_band(ctx, x2, N, mbits, B, lowband, cm = quant_band(ctx, x2, N, mbits, B, lowband, LM, lowband_out, Q15ONE,
LM, lowband_out, Q15ONE, lowband_scratch, orig_fill); lowband_scratch, orig_fill);
/* We don't split N=2 bands, so cm is either 1 or 0 (for a fold-collapse), /* We don't split N=2 bands, so cm is either 1 or 0 (for a fold-collapse),
and there's no need to worry about mixing with the other channel. */ and there's no need to worry about mixing with the other channel. */
y2[0] = -sign*x2[1]; y2[0] = -sign*x2[1];
y2[1] = sign*x2[0]; y2[1] = sign*x2[0];
if (resynth) if (ctx->resynth)
{ {
celt_norm tmp; celt_norm tmp;
X[0] = MULT16_16_Q15(mid, X[0]); X[0] = MULT16_16_Q15(mid, X[0]);
@ -1314,38 +1342,32 @@ static unsigned quant_band_stereo(struct band_ctx *ctx, celt_norm *X, celt_norm
{ {
/* In stereo mode, we do not apply a scaling to the mid because we need the normalized /* In stereo mode, we do not apply a scaling to the mid because we need the normalized
mid for folding later. */ mid for folding later. */
cm = quant_band(ctx, X, N, mbits, B, cm = quant_band(ctx, X, N, mbits, B, lowband, LM, lowband_out, Q15ONE,
lowband, LM, lowband_out, lowband_scratch, fill);
Q15ONE, lowband_scratch, fill);
rebalance = mbits - (rebalance-ctx->remaining_bits); rebalance = mbits - (rebalance-ctx->remaining_bits);
if (rebalance > 3<<BITRES && itheta!=0) if (rebalance > 3<<BITRES && itheta!=0)
sbits += rebalance - (3<<BITRES); sbits += rebalance - (3<<BITRES);
/* For a stereo split, the high bits of fill are always zero, so no /* For a stereo split, the high bits of fill are always zero, so no
folding will be done to the side. */ folding will be done to the side. */
cm |= quant_band(ctx, Y, N, sbits, B, cm |= quant_band(ctx, Y, N, sbits, B, NULL, LM, NULL, side, NULL, fill>>B);
NULL, LM, NULL,
side, NULL, fill>>B);
} else { } else {
/* For a stereo split, the high bits of fill are always zero, so no /* For a stereo split, the high bits of fill are always zero, so no
folding will be done to the side. */ folding will be done to the side. */
cm = quant_band(ctx, Y, N, sbits, B, cm = quant_band(ctx, Y, N, sbits, B, NULL, LM, NULL, side, NULL, fill>>B);
NULL, LM, NULL,
side, NULL, fill>>B);
rebalance = sbits - (rebalance-ctx->remaining_bits); rebalance = sbits - (rebalance-ctx->remaining_bits);
if (rebalance > 3<<BITRES && itheta!=16384) if (rebalance > 3<<BITRES && itheta!=16384)
mbits += rebalance - (3<<BITRES); mbits += rebalance - (3<<BITRES);
/* In stereo mode, we do not apply a scaling to the mid because we need the normalized /* In stereo mode, we do not apply a scaling to the mid because we need the normalized
mid for folding later. */ mid for folding later. */
cm |= quant_band(ctx, X, N, mbits, B, cm |= quant_band(ctx, X, N, mbits, B, lowband, LM, lowband_out, Q15ONE,
lowband, LM, lowband_out, lowband_scratch, fill);
Q15ONE, lowband_scratch, fill);
} }
} }
/* This code is used by the decoder and by the resynthesis-enabled encoder */ /* This code is used by the decoder and by the resynthesis-enabled encoder */
if (resynth) if (ctx->resynth)
{ {
if (N!=2) if (N!=2)
stereo_merge(X, Y, mid, N, ctx->arch); stereo_merge(X, Y, mid, N, ctx->arch);
@ -1359,19 +1381,38 @@ static unsigned quant_band_stereo(struct band_ctx *ctx, celt_norm *X, celt_norm
return cm; return cm;
} }
static void special_hybrid_folding(const CELTMode *m, celt_norm *norm, celt_norm *norm2, int start, int M, int dual_stereo)
{
int n1, n2;
const opus_int16 * OPUS_RESTRICT eBands = m->eBands;
n1 = M*(eBands[start+1]-eBands[start]);
n2 = M*(eBands[start+2]-eBands[start+1]);
/* Duplicate enough of the first band folding data to be able to fold the second band.
Copies no data for CELT-only mode. */
OPUS_COPY(&norm[n1], &norm[2*n1 - n2], n2-n1);
if (dual_stereo)
OPUS_COPY(&norm2[n1], &norm2[2*n1 - n2], n2-n1);
}
void quant_all_bands(int encode, const CELTMode *m, int start, int end, void quant_all_bands(int encode, const CELTMode *m, int start, int end,
celt_norm *X_, celt_norm *Y_, unsigned char *collapse_masks, celt_norm *X_, celt_norm *Y_, unsigned char *collapse_masks,
const celt_ener *bandE, int *pulses, int shortBlocks, int spread, const celt_ener *bandE, int *pulses, int shortBlocks, int spread,
int dual_stereo, int intensity, int *tf_res, opus_int32 total_bits, int dual_stereo, int intensity, int *tf_res, opus_int32 total_bits,
opus_int32 balance, ec_ctx *ec, int LM, int codedBands, opus_int32 balance, ec_ctx *ec, int LM, int codedBands,
opus_uint32 *seed, int arch) opus_uint32 *seed, int complexity, int arch, int disable_inv)
{ {
int i; int i;
opus_int32 remaining_bits; opus_int32 remaining_bits;
const opus_int16 * OPUS_RESTRICT eBands = m->eBands; const opus_int16 * OPUS_RESTRICT eBands = m->eBands;
celt_norm * OPUS_RESTRICT norm, * OPUS_RESTRICT norm2; celt_norm * OPUS_RESTRICT norm, * OPUS_RESTRICT norm2;
VARDECL(celt_norm, _norm); VARDECL(celt_norm, _norm);
VARDECL(celt_norm, _lowband_scratch);
VARDECL(celt_norm, X_save);
VARDECL(celt_norm, Y_save);
VARDECL(celt_norm, X_save2);
VARDECL(celt_norm, Y_save2);
VARDECL(celt_norm, norm_save2);
int resynth_alloc;
celt_norm *lowband_scratch; celt_norm *lowband_scratch;
int B; int B;
int M; int M;
@ -1379,10 +1420,11 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
int update_lowband = 1; int update_lowband = 1;
int C = Y_ != NULL ? 2 : 1; int C = Y_ != NULL ? 2 : 1;
int norm_offset; int norm_offset;
int theta_rdo = encode && Y_!=NULL && !dual_stereo && complexity>=8;
#ifdef RESYNTH #ifdef RESYNTH
int resynth = 1; int resynth = 1;
#else #else
int resynth = !encode; int resynth = !encode || theta_rdo;
#endif #endif
struct band_ctx ctx; struct band_ctx ctx;
SAVE_STACK; SAVE_STACK;
@ -1395,9 +1437,24 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
ALLOC(_norm, C*(M*eBands[m->nbEBands-1]-norm_offset), celt_norm); ALLOC(_norm, C*(M*eBands[m->nbEBands-1]-norm_offset), celt_norm);
norm = _norm; norm = _norm;
norm2 = norm + M*eBands[m->nbEBands-1]-norm_offset; norm2 = norm + M*eBands[m->nbEBands-1]-norm_offset;
/* We can use the last band as scratch space because we don't need that
scratch space for the last band. */ /* For decoding, we can use the last band as scratch space because we don't need that
scratch space for the last band and we don't care about the data there until we're
decoding the last band. */
if (encode && resynth)
resynth_alloc = M*(eBands[m->nbEBands]-eBands[m->nbEBands-1]);
else
resynth_alloc = ALLOC_NONE;
ALLOC(_lowband_scratch, resynth_alloc, celt_norm);
if (encode && resynth)
lowband_scratch = _lowband_scratch;
else
lowband_scratch = X_+M*eBands[m->nbEBands-1]; lowband_scratch = X_+M*eBands[m->nbEBands-1];
ALLOC(X_save, resynth_alloc, celt_norm);
ALLOC(Y_save, resynth_alloc, celt_norm);
ALLOC(X_save2, resynth_alloc, celt_norm);
ALLOC(Y_save2, resynth_alloc, celt_norm);
ALLOC(norm_save2, resynth_alloc, celt_norm);
lowband_offset = 0; lowband_offset = 0;
ctx.bandE = bandE; ctx.bandE = bandE;
@ -1408,6 +1465,11 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
ctx.seed = *seed; ctx.seed = *seed;
ctx.spread = spread; ctx.spread = spread;
ctx.arch = arch; ctx.arch = arch;
ctx.disable_inv = disable_inv;
ctx.resynth = resynth;
ctx.theta_round = 0;
/* Avoid injecting noise in the first band on transients. */
ctx.avoid_split_noise = B > 1;
for (i=start;i<end;i++) for (i=start;i<end;i++)
{ {
opus_int32 tell; opus_int32 tell;
@ -1430,6 +1492,7 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
else else
Y = NULL; Y = NULL;
N = M*eBands[i+1]-M*eBands[i]; N = M*eBands[i+1]-M*eBands[i];
celt_assert(N > 0);
tell = ec_tell_frac(ec); tell = ec_tell_frac(ec);
/* Compute how many bits we want to allocate to this band */ /* Compute how many bits we want to allocate to this band */
@ -1445,8 +1508,15 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
b = 0; b = 0;
} }
#ifndef DISABLE_UPDATE_DRAFT
if (resynth && (M*eBands[i]-N >= M*eBands[start] || i==start+1) && (update_lowband || lowband_offset==0))
lowband_offset = i;
if (i == start+1)
special_hybrid_folding(m, norm, norm2, start, M, dual_stereo);
#else
if (resynth && M*eBands[i]-N >= M*eBands[start] && (update_lowband || lowband_offset==0)) if (resynth && M*eBands[i]-N >= M*eBands[start] && (update_lowband || lowband_offset==0))
lowband_offset = i; lowband_offset = i;
#endif
tf_change = tf_res[i]; tf_change = tf_res[i];
ctx.tf_change = tf_change; ctx.tf_change = tf_change;
@ -1457,7 +1527,7 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
Y = norm; Y = norm;
lowband_scratch = NULL; lowband_scratch = NULL;
} }
if (i==end-1) if (last && !theta_rdo)
lowband_scratch = NULL; lowband_scratch = NULL;
/* Get a conservative estimate of the collapse_mask's for the bands we're /* Get a conservative estimate of the collapse_mask's for the bands we're
@ -1472,7 +1542,11 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
fold_start = lowband_offset; fold_start = lowband_offset;
while(M*eBands[--fold_start] > effective_lowband+norm_offset); while(M*eBands[--fold_start] > effective_lowband+norm_offset);
fold_end = lowband_offset-1; fold_end = lowband_offset-1;
#ifndef DISABLE_UPDATE_DRAFT
while(++fold_end < i && M*eBands[fold_end] < effective_lowband+norm_offset+N);
#else
while(M*eBands[++fold_end] < effective_lowband+norm_offset+N); while(M*eBands[++fold_end] < effective_lowband+norm_offset+N);
#endif
x_cm = y_cm = 0; x_cm = y_cm = 0;
fold_i = fold_start; do { fold_i = fold_start; do {
x_cm |= collapse_masks[fold_i*C+0]; x_cm |= collapse_masks[fold_i*C+0];
@ -1505,9 +1579,75 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
} else { } else {
if (Y!=NULL) if (Y!=NULL)
{ {
if (theta_rdo && i < intensity)
{
ec_ctx ec_save, ec_save2;
struct band_ctx ctx_save, ctx_save2;
opus_val32 dist0, dist1;
unsigned cm, cm2;
int nstart_bytes, nend_bytes, save_bytes;
unsigned char *bytes_buf;
unsigned char bytes_save[1275];
opus_val16 w[2];
compute_channel_weights(bandE[i], bandE[i+m->nbEBands], w);
/* Make a copy. */
cm = x_cm|y_cm;
ec_save = *ec;
ctx_save = ctx;
OPUS_COPY(X_save, X, N);
OPUS_COPY(Y_save, Y, N);
/* Encode and round down. */
ctx.theta_round = -1;
x_cm = quant_band_stereo(&ctx, X, Y, N, b, B,
effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, cm);
dist0 = MULT16_32_Q15(w[0], celt_inner_prod(X_save, X, N, arch)) + MULT16_32_Q15(w[1], celt_inner_prod(Y_save, Y, N, arch));
/* Save first result. */
cm2 = x_cm;
ec_save2 = *ec;
ctx_save2 = ctx;
OPUS_COPY(X_save2, X, N);
OPUS_COPY(Y_save2, Y, N);
if (!last)
OPUS_COPY(norm_save2, norm+M*eBands[i]-norm_offset, N);
nstart_bytes = ec_save.offs;
nend_bytes = ec_save.storage;
bytes_buf = ec_save.buf+nstart_bytes;
save_bytes = nend_bytes-nstart_bytes;
OPUS_COPY(bytes_save, bytes_buf, save_bytes);
/* Restore */
*ec = ec_save;
ctx = ctx_save;
OPUS_COPY(X, X_save, N);
OPUS_COPY(Y, Y_save, N);
#ifndef DISABLE_UPDATE_DRAFT
if (i == start+1)
special_hybrid_folding(m, norm, norm2, start, M, dual_stereo);
#endif
/* Encode and round up. */
ctx.theta_round = 1;
x_cm = quant_band_stereo(&ctx, X, Y, N, b, B,
effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, cm);
dist1 = MULT16_32_Q15(w[0], celt_inner_prod(X_save, X, N, arch)) + MULT16_32_Q15(w[1], celt_inner_prod(Y_save, Y, N, arch));
if (dist0 >= dist1) {
x_cm = cm2;
*ec = ec_save2;
ctx = ctx_save2;
OPUS_COPY(X, X_save2, N);
OPUS_COPY(Y, Y_save2, N);
if (!last)
OPUS_COPY(norm+M*eBands[i]-norm_offset, norm_save2, N);
OPUS_COPY(bytes_buf, bytes_save, save_bytes);
}
} else {
ctx.theta_round = 0;
x_cm = quant_band_stereo(&ctx, X, Y, N, b, B, x_cm = quant_band_stereo(&ctx, X, Y, N, b, B,
effective_lowband != -1 ? norm+effective_lowband : NULL, LM, effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, x_cm|y_cm); last?NULL:norm+M*eBands[i]-norm_offset, lowband_scratch, x_cm|y_cm);
}
} else { } else {
x_cm = quant_band(&ctx, X, N, b, B, x_cm = quant_band(&ctx, X, N, b, B,
effective_lowband != -1 ? norm+effective_lowband : NULL, LM, effective_lowband != -1 ? norm+effective_lowband : NULL, LM,
@ -1521,6 +1661,9 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
/* Update the folding position only as long as we have 1 bit/sample depth. */ /* Update the folding position only as long as we have 1 bit/sample depth. */
update_lowband = b>(N<<BITRES); update_lowband = b>(N<<BITRES);
/* We only need to avoid noise on a split for the first band. After that, we
have folding. */
ctx.avoid_split_noise = 0;
} }
*seed = ctx.seed; *seed = ctx.seed;

View File

@ -36,12 +36,15 @@
#include "entdec.h" #include "entdec.h"
#include "rate.h" #include "rate.h"
opus_int16 bitexact_cos(opus_int16 x);
int bitexact_log2tan(int isin,int icos);
/** Compute the amplitude (sqrt energy) in each of the bands /** Compute the amplitude (sqrt energy) in each of the bands
* @param m Mode data * @param m Mode data
* @param X Spectrum * @param X Spectrum
* @param bandE Square root of the energy for each band (returned) * @param bandE Square root of the energy for each band (returned)
*/ */
void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM); void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM, int arch);
/*void compute_noise_energies(const CELTMode *m, const celt_sig *X, const opus_val16 *tonality, celt_ener *bandE);*/ /*void compute_noise_energies(const CELTMode *m, const celt_sig *X, const opus_val16 *tonality, celt_ener *bandE);*/
@ -69,7 +72,7 @@ void denormalise_bands(const CELTMode *m, const celt_norm * OPUS_RESTRICT X,
int spreading_decision(const CELTMode *m, const celt_norm *X, int *average, int spreading_decision(const CELTMode *m, const celt_norm *X, int *average,
int last_decision, int *hf_average, int *tapset_decision, int update_hf, int last_decision, int *hf_average, int *tapset_decision, int update_hf,
int end, int C, int M); int end, int C, int M, const int *spread_weight);
#ifdef MEASURE_NORM_MSE #ifdef MEASURE_NORM_MSE
void measure_norm_mse(const CELTMode *m, float *X, float *X0, float *bandE, float *bandE0, int M, int N, int C); void measure_norm_mse(const CELTMode *m, float *X, float *X0, float *bandE, float *bandE0, int M, int N, int C);
@ -105,7 +108,7 @@ void quant_all_bands(int encode, const CELTMode *m, int start, int end,
const celt_ener *bandE, int *pulses, int shortBlocks, int spread, const celt_ener *bandE, int *pulses, int shortBlocks, int spread,
int dual_stereo, int intensity, int *tf_res, opus_int32 total_bits, int dual_stereo, int intensity, int *tf_res, opus_int32 total_bits,
opus_int32 balance, ec_ctx *ec, int M, int codedBands, opus_uint32 *seed, opus_int32 balance, ec_ctx *ec, int M, int codedBands, opus_uint32 *seed,
int arch); int complexity, int arch, int disable_inv);
void anti_collapse(const CELTMode *m, celt_norm *X_, void anti_collapse(const CELTMode *m, celt_norm *X_,
unsigned char *collapse_masks, int LM, int C, int size, int start, unsigned char *collapse_masks, int LM, int C, int size, int start,

View File

@ -111,26 +111,31 @@ void comb_filter_const_c(opus_val32 *y, opus_val32 *x, int T, int N,
t = MAC16_32_Q16(x[i], g10, x2); t = MAC16_32_Q16(x[i], g10, x2);
t = MAC16_32_Q16(t, g11, ADD32(x1,x3)); t = MAC16_32_Q16(t, g11, ADD32(x1,x3));
t = MAC16_32_Q16(t, g12, ADD32(x0,x4)); t = MAC16_32_Q16(t, g12, ADD32(x0,x4));
t = SATURATE(t, SIG_SAT);
y[i] = t; y[i] = t;
x4=SHL32(x[i-T+3],1); x4=SHL32(x[i-T+3],1);
t = MAC16_32_Q16(x[i+1], g10, x1); t = MAC16_32_Q16(x[i+1], g10, x1);
t = MAC16_32_Q16(t, g11, ADD32(x0,x2)); t = MAC16_32_Q16(t, g11, ADD32(x0,x2));
t = MAC16_32_Q16(t, g12, ADD32(x4,x3)); t = MAC16_32_Q16(t, g12, ADD32(x4,x3));
t = SATURATE(t, SIG_SAT);
y[i+1] = t; y[i+1] = t;
x3=SHL32(x[i-T+4],1); x3=SHL32(x[i-T+4],1);
t = MAC16_32_Q16(x[i+2], g10, x0); t = MAC16_32_Q16(x[i+2], g10, x0);
t = MAC16_32_Q16(t, g11, ADD32(x4,x1)); t = MAC16_32_Q16(t, g11, ADD32(x4,x1));
t = MAC16_32_Q16(t, g12, ADD32(x3,x2)); t = MAC16_32_Q16(t, g12, ADD32(x3,x2));
t = SATURATE(t, SIG_SAT);
y[i+2] = t; y[i+2] = t;
x2=SHL32(x[i-T+5],1); x2=SHL32(x[i-T+5],1);
t = MAC16_32_Q16(x[i+3], g10, x4); t = MAC16_32_Q16(x[i+3], g10, x4);
t = MAC16_32_Q16(t, g11, ADD32(x3,x0)); t = MAC16_32_Q16(t, g11, ADD32(x3,x0));
t = MAC16_32_Q16(t, g12, ADD32(x2,x1)); t = MAC16_32_Q16(t, g12, ADD32(x2,x1));
t = SATURATE(t, SIG_SAT);
y[i+3] = t; y[i+3] = t;
x1=SHL32(x[i-T+6],1); x1=SHL32(x[i-T+6],1);
t = MAC16_32_Q16(x[i+4], g10, x3); t = MAC16_32_Q16(x[i+4], g10, x3);
t = MAC16_32_Q16(t, g11, ADD32(x2,x4)); t = MAC16_32_Q16(t, g11, ADD32(x2,x4));
t = MAC16_32_Q16(t, g12, ADD32(x1,x0)); t = MAC16_32_Q16(t, g12, ADD32(x1,x0));
t = SATURATE(t, SIG_SAT);
y[i+4] = t; y[i+4] = t;
} }
#ifdef CUSTOM_MODES #ifdef CUSTOM_MODES
@ -141,6 +146,7 @@ void comb_filter_const_c(opus_val32 *y, opus_val32 *x, int T, int N,
t = MAC16_32_Q16(x[i], g10, x2); t = MAC16_32_Q16(x[i], g10, x2);
t = MAC16_32_Q16(t, g11, ADD32(x1,x3)); t = MAC16_32_Q16(t, g11, ADD32(x1,x3));
t = MAC16_32_Q16(t, g12, ADD32(x0,x4)); t = MAC16_32_Q16(t, g12, ADD32(x0,x4));
t = SATURATE(t, SIG_SAT);
y[i] = t; y[i] = t;
x4=x3; x4=x3;
x3=x2; x3=x2;
@ -169,6 +175,7 @@ void comb_filter_const_c(opus_val32 *y, opus_val32 *x, int T, int N,
+ MULT16_32_Q15(g10,x2) + MULT16_32_Q15(g10,x2)
+ MULT16_32_Q15(g11,ADD32(x1,x3)) + MULT16_32_Q15(g11,ADD32(x1,x3))
+ MULT16_32_Q15(g12,ADD32(x0,x4)); + MULT16_32_Q15(g12,ADD32(x0,x4));
y[i] = SATURATE(y[i], SIG_SAT);
x4=x3; x4=x3;
x3=x2; x3=x2;
x2=x1; x2=x1;
@ -200,6 +207,10 @@ void comb_filter(opus_val32 *y, opus_val32 *x, int T0, int T1, int N,
OPUS_MOVE(y, x, N); OPUS_MOVE(y, x, N);
return; return;
} }
/* When the gain is zero, T0 and/or T1 is set to zero. We need
to have then be at least 2 to avoid processing garbage data. */
T0 = IMAX(T0, COMBFILTER_MINPERIOD);
T1 = IMAX(T1, COMBFILTER_MINPERIOD);
g00 = MULT16_16_P15(g0, gains[tapset0][0]); g00 = MULT16_16_P15(g0, gains[tapset0][0]);
g01 = MULT16_16_P15(g0, gains[tapset0][1]); g01 = MULT16_16_P15(g0, gains[tapset0][1]);
g02 = MULT16_16_P15(g0, gains[tapset0][2]); g02 = MULT16_16_P15(g0, gains[tapset0][2]);
@ -225,6 +236,7 @@ void comb_filter(opus_val32 *y, opus_val32 *x, int T0, int T1, int N,
+ MULT16_32_Q15(MULT16_16_Q15(f,g10),x2) + MULT16_32_Q15(MULT16_16_Q15(f,g10),x2)
+ MULT16_32_Q15(MULT16_16_Q15(f,g11),ADD32(x1,x3)) + MULT16_32_Q15(MULT16_16_Q15(f,g11),ADD32(x1,x3))
+ MULT16_32_Q15(MULT16_16_Q15(f,g12),ADD32(x0,x4)); + MULT16_32_Q15(MULT16_16_Q15(f,g12),ADD32(x0,x4));
y[i] = SATURATE(y[i], SIG_SAT);
x4=x3; x4=x3;
x3=x2; x3=x2;
x2=x1; x2=x1;
@ -244,11 +256,16 @@ void comb_filter(opus_val32 *y, opus_val32 *x, int T0, int T1, int N,
} }
#endif /* OVERRIDE_comb_filter */ #endif /* OVERRIDE_comb_filter */
/* TF change table. Positive values mean better frequency resolution (longer
effective window), whereas negative values mean better time resolution
(shorter effective window). The second index is computed as:
4*isTransient + 2*tf_select + per_band_flag */
const signed char tf_select_table[4][8] = { const signed char tf_select_table[4][8] = {
{0, -1, 0, -1, 0,-1, 0,-1}, /*isTransient=0 isTransient=1 */
{0, -1, 0, -2, 1, 0, 1,-1}, {0, -1, 0, -1, 0,-1, 0,-1}, /* 2.5 ms */
{0, -2, 0, -3, 2, 0, 1,-1}, {0, -1, 0, -2, 1, 0, 1,-1}, /* 5 ms */
{0, -2, 0, -3, 3, 0, 1,-1}, {0, -2, 0, -3, 2, 0, 1,-1}, /* 10 ms */
{0, -2, 0, -3, 3, 0, 1,-1}, /* 20 ms */
}; };

View File

@ -50,6 +50,8 @@ extern "C" {
#define CELTDecoder OpusCustomDecoder #define CELTDecoder OpusCustomDecoder
#define CELTMode OpusCustomMode #define CELTMode OpusCustomMode
#define LEAK_BANDS 19
typedef struct { typedef struct {
int valid; int valid;
float tonality; float tonality;
@ -57,17 +59,27 @@ typedef struct {
float noisiness; float noisiness;
float activity; float activity;
float music_prob; float music_prob;
float music_prob_min;
float music_prob_max;
int bandwidth; int bandwidth;
}AnalysisInfo; float activity_probability;
float max_pitch_ratio;
/* Store as Q6 char to save space. */
unsigned char leak_boost[LEAK_BANDS];
} AnalysisInfo;
typedef struct {
int signalType;
int offset;
} SILKInfo;
#define __celt_check_mode_ptr_ptr(ptr) ((ptr) + ((ptr) - (const CELTMode**)(ptr))) #define __celt_check_mode_ptr_ptr(ptr) ((ptr) + ((ptr) - (const CELTMode**)(ptr)))
#define __celt_check_analysis_ptr(ptr) ((ptr) + ((ptr) - (const AnalysisInfo*)(ptr))) #define __celt_check_analysis_ptr(ptr) ((ptr) + ((ptr) - (const AnalysisInfo*)(ptr)))
/* Encoder/decoder Requests */ #define __celt_check_silkinfo_ptr(ptr) ((ptr) + ((ptr) - (const SILKInfo*)(ptr)))
/* Expose this option again when variable framesize actually works */ /* Encoder/decoder Requests */
#define OPUS_FRAMESIZE_VARIABLE 5010 /**< Optimize the frame size dynamically */
#define CELT_SET_PREDICTION_REQUEST 10002 #define CELT_SET_PREDICTION_REQUEST 10002
@ -116,6 +128,9 @@ typedef struct {
#define OPUS_SET_ENERGY_MASK_REQUEST 10026 #define OPUS_SET_ENERGY_MASK_REQUEST 10026
#define OPUS_SET_ENERGY_MASK(x) OPUS_SET_ENERGY_MASK_REQUEST, __opus_check_val16_ptr(x) #define OPUS_SET_ENERGY_MASK(x) OPUS_SET_ENERGY_MASK_REQUEST, __opus_check_val16_ptr(x)
#define CELT_SET_SILK_INFO_REQUEST 10028
#define CELT_SET_SILK_INFO(x) CELT_SET_SILK_INFO_REQUEST, __celt_check_silkinfo_ptr(x)
/* Encoder stuff */ /* Encoder stuff */
int celt_encoder_get_size(int channels); int celt_encoder_get_size(int channels);
@ -194,6 +209,13 @@ static OPUS_INLINE int fromOpus(unsigned char c)
extern const signed char tf_select_table[4][8]; extern const signed char tf_select_table[4][8];
#if defined(ENABLE_HARDENING) || defined(ENABLE_ASSERTIONS)
void validate_celt_decoder(CELTDecoder *st);
#define VALIDATE_CELT_DECODER(st) validate_celt_decoder(st)
#else
#define VALIDATE_CELT_DECODER(st)
#endif
int resampling_factor(opus_int32 rate); int resampling_factor(opus_int32 rate);
void celt_preemphasis(const opus_val16 * OPUS_RESTRICT pcmp, celt_sig * OPUS_RESTRICT inp, void celt_preemphasis(const opus_val16 * OPUS_RESTRICT pcmp, celt_sig * OPUS_RESTRICT inp,

View File

@ -51,6 +51,14 @@
#include "celt_lpc.h" #include "celt_lpc.h"
#include "vq.h" #include "vq.h"
/* The maximum pitch lag to allow in the pitch-based PLC. It's possible to save
CPU time in the PLC pitch search by making this smaller than MAX_PERIOD. The
current value corresponds to a pitch of 66.67 Hz. */
#define PLC_PITCH_LAG_MAX (720)
/* The minimum pitch lag to allow in the pitch-based PLC. This corresponds to a
pitch of 480 Hz. */
#define PLC_PITCH_LAG_MIN (100)
#if defined(SMALL_FOOTPRINT) && defined(FIXED_POINT) #if defined(SMALL_FOOTPRINT) && defined(FIXED_POINT)
#define NORM_ALIASING_HACK #define NORM_ALIASING_HACK
#endif #endif
@ -73,6 +81,7 @@ struct OpusCustomDecoder {
int downsample; int downsample;
int start, end; int start, end;
int signalling; int signalling;
int disable_inv;
int arch; int arch;
/* Everything beyond this point gets cleared on a reset */ /* Everything beyond this point gets cleared on a reset */
@ -82,6 +91,7 @@ struct OpusCustomDecoder {
int error; int error;
int last_pitch_index; int last_pitch_index;
int loss_count; int loss_count;
int skip_plc;
int postfilter_period; int postfilter_period;
int postfilter_period_old; int postfilter_period_old;
opus_val16 postfilter_gain; opus_val16 postfilter_gain;
@ -99,6 +109,38 @@ struct OpusCustomDecoder {
/* opus_val16 backgroundLogE[], Size = 2*mode->nbEBands */ /* opus_val16 backgroundLogE[], Size = 2*mode->nbEBands */
}; };
#if defined(ENABLE_HARDENING) || defined(ENABLE_ASSERTIONS)
/* Make basic checks on the CELT state to ensure we don't end
up writing all over memory. */
void validate_celt_decoder(CELTDecoder *st)
{
#ifndef CUSTOM_MODES
celt_assert(st->mode == opus_custom_mode_create(48000, 960, NULL));
celt_assert(st->overlap == 120);
#endif
celt_assert(st->channels == 1 || st->channels == 2);
celt_assert(st->stream_channels == 1 || st->stream_channels == 2);
celt_assert(st->downsample > 0);
celt_assert(st->start == 0 || st->start == 17);
celt_assert(st->start < st->end);
celt_assert(st->end <= 21);
#ifdef OPUS_ARCHMASK
celt_assert(st->arch >= 0);
celt_assert(st->arch <= OPUS_ARCHMASK);
#endif
celt_assert(st->last_pitch_index <= PLC_PITCH_LAG_MAX);
celt_assert(st->last_pitch_index >= PLC_PITCH_LAG_MIN || st->last_pitch_index == 0);
celt_assert(st->postfilter_period < MAX_PERIOD);
celt_assert(st->postfilter_period >= COMBFILTER_MINPERIOD || st->postfilter_period == 0);
celt_assert(st->postfilter_period_old < MAX_PERIOD);
celt_assert(st->postfilter_period_old >= COMBFILTER_MINPERIOD || st->postfilter_period_old == 0);
celt_assert(st->postfilter_tapset <= 2);
celt_assert(st->postfilter_tapset >= 0);
celt_assert(st->postfilter_tapset_old <= 2);
celt_assert(st->postfilter_tapset_old >= 0);
}
#endif
int celt_decoder_get_size(int channels) int celt_decoder_get_size(int channels)
{ {
const CELTMode *mode = opus_custom_mode_create(48000, 960, NULL); const CELTMode *mode = opus_custom_mode_create(48000, 960, NULL);
@ -162,10 +204,13 @@ OPUS_CUSTOM_NOSTATIC int opus_custom_decoder_init(CELTDecoder *st, const CELTMod
st->start = 0; st->start = 0;
st->end = st->mode->effEBands; st->end = st->mode->effEBands;
st->signalling = 1; st->signalling = 1;
#ifndef DISABLE_UPDATE_DRAFT
st->disable_inv = channels == 1;
#else
st->disable_inv = 0;
#endif
st->arch = opus_select_arch(); st->arch = opus_select_arch();
st->loss_count = 0;
opus_custom_decoder_ctl(st, OPUS_RESET_STATE); opus_custom_decoder_ctl(st, OPUS_RESET_STATE);
return OPUS_OK; return OPUS_OK;
@ -178,6 +223,36 @@ void opus_custom_decoder_destroy(CELTDecoder *st)
} }
#endif /* CUSTOM_MODES */ #endif /* CUSTOM_MODES */
#ifndef CUSTOM_MODES
/* Special case for stereo with no downsampling and no accumulation. This is
quite common and we can make it faster by processing both channels in the
same loop, reducing overhead due to the dependency loop in the IIR filter. */
static void deemphasis_stereo_simple(celt_sig *in[], opus_val16 *pcm, int N, const opus_val16 coef0,
celt_sig *mem)
{
celt_sig * OPUS_RESTRICT x0;
celt_sig * OPUS_RESTRICT x1;
celt_sig m0, m1;
int j;
x0=in[0];
x1=in[1];
m0 = mem[0];
m1 = mem[1];
for (j=0;j<N;j++)
{
celt_sig tmp0, tmp1;
/* Add VERY_SMALL to x[] first to reduce dependency chain. */
tmp0 = x0[j] + VERY_SMALL + m0;
tmp1 = x1[j] + VERY_SMALL + m1;
m0 = MULT16_32_Q15(coef0, tmp0);
m1 = MULT16_32_Q15(coef0, tmp1);
pcm[2*j ] = SCALEOUT(SIG2WORD16(tmp0));
pcm[2*j+1] = SCALEOUT(SIG2WORD16(tmp1));
}
mem[0] = m0;
mem[1] = m1;
}
#endif
#ifndef RESYNTH #ifndef RESYNTH
static static
@ -191,6 +266,14 @@ void deemphasis(celt_sig *in[], opus_val16 *pcm, int N, int C, int downsample, c
opus_val16 coef0; opus_val16 coef0;
VARDECL(celt_sig, scratch); VARDECL(celt_sig, scratch);
SAVE_STACK; SAVE_STACK;
#ifndef CUSTOM_MODES
/* Short version for common case. */
if (downsample == 1 && C == 2 && !accum)
{
deemphasis_stereo_simple(in, pcm, N, coef[0], mem);
return;
}
#endif
#ifndef FIXED_POINT #ifndef FIXED_POINT
(void)accum; (void)accum;
celt_assert(accum==0); celt_assert(accum==0);
@ -226,7 +309,7 @@ void deemphasis(celt_sig *in[], opus_val16 *pcm, int N, int C, int downsample, c
/* Shortcut for the standard (non-custom modes) case */ /* Shortcut for the standard (non-custom modes) case */
for (j=0;j<N;j++) for (j=0;j<N;j++)
{ {
celt_sig tmp = x[j] + m + VERY_SMALL; celt_sig tmp = x[j] + VERY_SMALL + m;
m = MULT16_32_Q15(coef0, tmp); m = MULT16_32_Q15(coef0, tmp);
scratch[j] = tmp; scratch[j] = tmp;
} }
@ -247,7 +330,7 @@ void deemphasis(celt_sig *in[], opus_val16 *pcm, int N, int C, int downsample, c
{ {
for (j=0;j<N;j++) for (j=0;j<N;j++)
{ {
celt_sig tmp = x[j] + m + VERY_SMALL; celt_sig tmp = x[j] + VERY_SMALL + m;
m = MULT16_32_Q15(coef0, tmp); m = MULT16_32_Q15(coef0, tmp);
y[j*C] = SCALEOUT(SIG2WORD16(tmp)); y[j*C] = SCALEOUT(SIG2WORD16(tmp));
} }
@ -334,7 +417,7 @@ void celt_synthesis(const CELTMode *mode, celt_norm *X, celt_sig * out_syn[],
denormalise_bands(mode, X+N, freq2, oldBandE+nbEBands, start, effEnd, M, denormalise_bands(mode, X+N, freq2, oldBandE+nbEBands, start, effEnd, M,
downsample, silence); downsample, silence);
for (i=0;i<N;i++) for (i=0;i<N;i++)
freq[i] = HALF32(ADD32(freq[i],freq2[i])); freq[i] = ADD32(HALF32(freq[i]), HALF32(freq2[i]));
for (b=0;b<B;b++) for (b=0;b<B;b++)
clt_mdct_backward(&mode->mdct, &freq[b], out_syn[0]+NB*b, mode->window, overlap, shift, B, arch); clt_mdct_backward(&mode->mdct, &freq[b], out_syn[0]+NB*b, mode->window, overlap, shift, B, arch);
} else { } else {
@ -346,6 +429,12 @@ void celt_synthesis(const CELTMode *mode, celt_norm *X, celt_sig * out_syn[],
clt_mdct_backward(&mode->mdct, &freq[b], out_syn[c]+NB*b, mode->window, overlap, shift, B, arch); clt_mdct_backward(&mode->mdct, &freq[b], out_syn[c]+NB*b, mode->window, overlap, shift, B, arch);
} while (++c<CC); } while (++c<CC);
} }
/* Saturate IMDCT output so that we can't overflow in the pitch postfilter
or in the */
c=0; do {
for (i=0;i<N;i++)
out_syn[c][i] = SATURATE(out_syn[c][i], SIG_SAT);
} while (++c<CC);
RESTORE_STACK; RESTORE_STACK;
} }
@ -388,14 +477,6 @@ static void tf_decode(int start, int end, int isTransient, int *tf_res, int LM,
} }
} }
/* The maximum pitch lag to allow in the pitch-based PLC. It's possible to save
CPU time in the PLC pitch search by making this smaller than MAX_PERIOD. The
current value corresponds to a pitch of 66.67 Hz. */
#define PLC_PITCH_LAG_MAX (720)
/* The minimum pitch lag to allow in the pitch-based PLC. This corresponds to a
pitch of 480 Hz. */
#define PLC_PITCH_LAG_MIN (100)
static int celt_plc_pitch_search(celt_sig *decode_mem[2], int C, int arch) static int celt_plc_pitch_search(celt_sig *decode_mem[2], int C, int arch)
{ {
int pitch_index; int pitch_index;
@ -447,7 +528,7 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
loss_count = st->loss_count; loss_count = st->loss_count;
start = st->start; start = st->start;
noise_based = loss_count >= 5 || start != 0; noise_based = loss_count >= 5 || start != 0 || st->skip_plc;
if (noise_based) if (noise_based)
{ {
/* Noise-based PLC/CNG */ /* Noise-based PLC/CNG */
@ -505,12 +586,15 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
celt_synthesis(mode, X, out_syn, oldBandE, start, effEnd, C, C, 0, LM, st->downsample, 0, st->arch); celt_synthesis(mode, X, out_syn, oldBandE, start, effEnd, C, C, 0, LM, st->downsample, 0, st->arch);
} else { } else {
int exc_length;
/* Pitch-based PLC */ /* Pitch-based PLC */
const opus_val16 *window; const opus_val16 *window;
opus_val16 *exc;
opus_val16 fade = Q15ONE; opus_val16 fade = Q15ONE;
int pitch_index; int pitch_index;
VARDECL(opus_val32, etmp); VARDECL(opus_val32, etmp);
VARDECL(opus_val16, exc); VARDECL(opus_val16, _exc);
VARDECL(opus_val16, fir_tmp);
if (loss_count == 0) if (loss_count == 0)
{ {
@ -520,8 +604,14 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
fade = QCONST16(.8f,15); fade = QCONST16(.8f,15);
} }
/* We want the excitation for 2 pitch periods in order to look for a
decaying signal, but we can't get more than MAX_PERIOD. */
exc_length = IMIN(2*pitch_index, MAX_PERIOD);
ALLOC(etmp, overlap, opus_val32); ALLOC(etmp, overlap, opus_val32);
ALLOC(exc, MAX_PERIOD, opus_val16); ALLOC(_exc, MAX_PERIOD+LPC_ORDER, opus_val16);
ALLOC(fir_tmp, exc_length, opus_val16);
exc = _exc+LPC_ORDER;
window = mode->window; window = mode->window;
c=0; do { c=0; do {
opus_val16 decay; opus_val16 decay;
@ -530,13 +620,11 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
celt_sig *buf; celt_sig *buf;
int extrapolation_offset; int extrapolation_offset;
int extrapolation_len; int extrapolation_len;
int exc_length;
int j; int j;
buf = decode_mem[c]; buf = decode_mem[c];
for (i=0;i<MAX_PERIOD;i++) { for (i=0;i<MAX_PERIOD+LPC_ORDER;i++)
exc[i] = ROUND16(buf[DECODE_BUFFER_SIZE-MAX_PERIOD+i], SIG_SHIFT); exc[i-LPC_ORDER] = ROUND16(buf[DECODE_BUFFER_SIZE-MAX_PERIOD-LPC_ORDER+i], SIG_SHIFT);
}
if (loss_count == 0) if (loss_count == 0)
{ {
@ -562,22 +650,32 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
#endif #endif
} }
_celt_lpc(lpc+c*LPC_ORDER, ac, LPC_ORDER); _celt_lpc(lpc+c*LPC_ORDER, ac, LPC_ORDER);
#ifdef FIXED_POINT
/* For fixed-point, apply bandwidth expansion until we can guarantee that
no overflow can happen in the IIR filter. This means:
32768*sum(abs(filter)) < 2^31 */
while (1) {
opus_val16 tmp=Q15ONE;
opus_val32 sum=QCONST16(1., SIG_SHIFT);
for (i=0;i<LPC_ORDER;i++)
sum += ABS16(lpc[c*LPC_ORDER+i]);
if (sum < 65535) break;
for (i=0;i<LPC_ORDER;i++)
{
tmp = MULT16_16_Q15(QCONST16(.99f,15), tmp);
lpc[c*LPC_ORDER+i] = MULT16_16_Q15(lpc[c*LPC_ORDER+i], tmp);
}
}
#endif
} }
/* We want the excitation for 2 pitch periods in order to look for a
decaying signal, but we can't get more than MAX_PERIOD. */
exc_length = IMIN(2*pitch_index, MAX_PERIOD);
/* Initialize the LPC history with the samples just before the start /* Initialize the LPC history with the samples just before the start
of the region for which we're computing the excitation. */ of the region for which we're computing the excitation. */
{ {
opus_val16 lpc_mem[LPC_ORDER]; /* Compute the excitation for exc_length samples before the loss. We need the copy
for (i=0;i<LPC_ORDER;i++) because celt_fir() cannot filter in-place. */
{
lpc_mem[i] =
ROUND16(buf[DECODE_BUFFER_SIZE-exc_length-1-i], SIG_SHIFT);
}
/* Compute the excitation for exc_length samples before the loss. */
celt_fir(exc+MAX_PERIOD-exc_length, lpc+c*LPC_ORDER, celt_fir(exc+MAX_PERIOD-exc_length, lpc+c*LPC_ORDER,
exc+MAX_PERIOD-exc_length, exc_length, LPC_ORDER, lpc_mem, st->arch); fir_tmp, exc_length, LPC_ORDER, st->arch);
OPUS_COPY(exc+MAX_PERIOD-exc_length, fir_tmp, exc_length);
} }
/* Check if the waveform is decaying, and if so how fast. /* Check if the waveform is decaying, and if so how fast.
@ -631,9 +729,8 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
tmp = ROUND16( tmp = ROUND16(
buf[DECODE_BUFFER_SIZE-MAX_PERIOD-N+extrapolation_offset+j], buf[DECODE_BUFFER_SIZE-MAX_PERIOD-N+extrapolation_offset+j],
SIG_SHIFT); SIG_SHIFT);
S1 += SHR32(MULT16_16(tmp, tmp), 8); S1 += SHR32(MULT16_16(tmp, tmp), 10);
} }
{ {
opus_val16 lpc_mem[LPC_ORDER]; opus_val16 lpc_mem[LPC_ORDER];
/* Copy the last decoded samples (prior to the overlap region) to /* Copy the last decoded samples (prior to the overlap region) to
@ -645,6 +742,10 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
celt_iir(buf+DECODE_BUFFER_SIZE-N, lpc+c*LPC_ORDER, celt_iir(buf+DECODE_BUFFER_SIZE-N, lpc+c*LPC_ORDER,
buf+DECODE_BUFFER_SIZE-N, extrapolation_len, LPC_ORDER, buf+DECODE_BUFFER_SIZE-N, extrapolation_len, LPC_ORDER,
lpc_mem, st->arch); lpc_mem, st->arch);
#ifdef FIXED_POINT
for (i=0; i < extrapolation_len; i++)
buf[DECODE_BUFFER_SIZE-N+i] = SATURATE(buf[DECODE_BUFFER_SIZE-N+i], SIG_SAT);
#endif
} }
/* Check if the synthesis energy is higher than expected, which can /* Check if the synthesis energy is higher than expected, which can
@ -655,7 +756,7 @@ static void celt_decode_lost(CELTDecoder * OPUS_RESTRICT st, int N, int LM)
for (i=0;i<extrapolation_len;i++) for (i=0;i<extrapolation_len;i++)
{ {
opus_val16 tmp = ROUND16(buf[DECODE_BUFFER_SIZE-N+i], SIG_SHIFT); opus_val16 tmp = ROUND16(buf[DECODE_BUFFER_SIZE-N+i], SIG_SHIFT);
S2 += SHR32(MULT16_16(tmp, tmp), 8); S2 += SHR32(MULT16_16(tmp, tmp), 10);
} }
/* This checks for an "explosion" in the synthesis. */ /* This checks for an "explosion" in the synthesis. */
#ifdef FIXED_POINT #ifdef FIXED_POINT
@ -763,6 +864,7 @@ int celt_decode_with_ec(CELTDecoder * OPUS_RESTRICT st, const unsigned char *dat
const opus_int16 *eBands; const opus_int16 *eBands;
ALLOC_STACK; ALLOC_STACK;
VALIDATE_CELT_DECODER(st);
mode = st->mode; mode = st->mode;
nbEBands = mode->nbEBands; nbEBands = mode->nbEBands;
overlap = mode->overlap; overlap = mode->overlap;
@ -832,6 +934,10 @@ int celt_decode_with_ec(CELTDecoder * OPUS_RESTRICT st, const unsigned char *dat
return frame_size/st->downsample; return frame_size/st->downsample;
} }
/* Check if there are at least two packets received consecutively before
* turning on the pitch-based PLC */
st->skip_plc = st->loss_count != 0;
if (dec == NULL) if (dec == NULL)
{ {
ec_dec_init(&_dec,(unsigned char*)data,len); ec_dec_init(&_dec,(unsigned char*)data,len);
@ -953,7 +1059,7 @@ int celt_decode_with_ec(CELTDecoder * OPUS_RESTRICT st, const unsigned char *dat
ALLOC(pulses, nbEBands, int); ALLOC(pulses, nbEBands, int);
ALLOC(fine_priority, nbEBands, int); ALLOC(fine_priority, nbEBands, int);
codedBands = compute_allocation(mode, start, end, offsets, cap, codedBands = clt_compute_allocation(mode, start, end, offsets, cap,
alloc_trim, &intensity, &dual_stereo, bits, &balance, pulses, alloc_trim, &intensity, &dual_stereo, bits, &balance, pulses,
fine_quant, fine_priority, C, LM, dec, 0, 0, 0); fine_quant, fine_priority, C, LM, dec, 0, 0, 0);
@ -976,7 +1082,8 @@ int celt_decode_with_ec(CELTDecoder * OPUS_RESTRICT st, const unsigned char *dat
quant_all_bands(0, mode, start, end, X, C==2 ? X+N : NULL, collapse_masks, quant_all_bands(0, mode, start, end, X, C==2 ? X+N : NULL, collapse_masks,
NULL, pulses, shortBlocks, spread_decision, dual_stereo, intensity, tf_res, NULL, pulses, shortBlocks, spread_decision, dual_stereo, intensity, tf_res,
len*(8<<BITRES)-anti_collapse_rsv, balance, dec, LM, codedBands, &st->rng, st->arch); len*(8<<BITRES)-anti_collapse_rsv, balance, dec, LM, codedBands, &st->rng, 0,
st->arch, st->disable_inv);
if (anti_collapse_rsv > 0) if (anti_collapse_rsv > 0)
{ {
@ -1198,6 +1305,7 @@ int opus_custom_decoder_ctl(CELTDecoder * OPUS_RESTRICT st, int request, ...)
((char*)&st->DECODER_RESET_START - (char*)st)); ((char*)&st->DECODER_RESET_START - (char*)st));
for (i=0;i<2*st->mode->nbEBands;i++) for (i=0;i<2*st->mode->nbEBands;i++)
oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT); oldLogE[i]=oldLogE2[i]=-QCONST16(28.f,DB_SHIFT);
st->skip_plc = 1;
} }
break; break;
case OPUS_GET_PITCH_REQUEST: case OPUS_GET_PITCH_REQUEST:
@ -1230,6 +1338,26 @@ int opus_custom_decoder_ctl(CELTDecoder * OPUS_RESTRICT st, int request, ...)
*value=st->rng; *value=st->rng;
} }
break; break;
case OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST:
{
opus_int32 value = va_arg(ap, opus_int32);
if(value<0 || value>1)
{
goto bad_arg;
}
st->disable_inv = value;
}
break;
case OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
if (!value)
{
goto bad_arg;
}
*value = st->disable_inv;
}
break;
default: default:
goto bad_request; goto bad_request;
} }

View File

@ -73,8 +73,8 @@ struct OpusCustomEncoder {
int constrained_vbr; /* If zero, VBR can do whatever it likes with the rate */ int constrained_vbr; /* If zero, VBR can do whatever it likes with the rate */
int loss_rate; int loss_rate;
int lsb_depth; int lsb_depth;
int variable_duration;
int lfe; int lfe;
int disable_inv;
int arch; int arch;
/* Everything beyond this point gets cleared on a reset */ /* Everything beyond this point gets cleared on a reset */
@ -98,6 +98,7 @@ struct OpusCustomEncoder {
#endif #endif
int consec_transient; int consec_transient;
AnalysisInfo analysis; AnalysisInfo analysis;
SILKInfo silk_info;
opus_val32 preemph_memE[2]; opus_val32 preemph_memE[2];
opus_val32 preemph_memD[2]; opus_val32 preemph_memD[2];
@ -123,6 +124,7 @@ struct OpusCustomEncoder {
/* opus_val16 oldBandE[], Size = channels*mode->nbEBands */ /* opus_val16 oldBandE[], Size = channels*mode->nbEBands */
/* opus_val16 oldLogE[], Size = channels*mode->nbEBands */ /* opus_val16 oldLogE[], Size = channels*mode->nbEBands */
/* opus_val16 oldLogE2[], Size = channels*mode->nbEBands */ /* opus_val16 oldLogE2[], Size = channels*mode->nbEBands */
/* opus_val16 energyError[], Size = channels*mode->nbEBands */
}; };
int celt_encoder_get_size(int channels) int celt_encoder_get_size(int channels)
@ -136,9 +138,10 @@ OPUS_CUSTOM_NOSTATIC int opus_custom_encoder_get_size(const CELTMode *mode, int
int size = sizeof(struct CELTEncoder) int size = sizeof(struct CELTEncoder)
+ (channels*mode->overlap-1)*sizeof(celt_sig) /* celt_sig in_mem[channels*mode->overlap]; */ + (channels*mode->overlap-1)*sizeof(celt_sig) /* celt_sig in_mem[channels*mode->overlap]; */
+ channels*COMBFILTER_MAXPERIOD*sizeof(celt_sig) /* celt_sig prefilter_mem[channels*COMBFILTER_MAXPERIOD]; */ + channels*COMBFILTER_MAXPERIOD*sizeof(celt_sig) /* celt_sig prefilter_mem[channels*COMBFILTER_MAXPERIOD]; */
+ 3*channels*mode->nbEBands*sizeof(opus_val16); /* opus_val16 oldBandE[channels*mode->nbEBands]; */ + 4*channels*mode->nbEBands*sizeof(opus_val16); /* opus_val16 oldBandE[channels*mode->nbEBands]; */
/* opus_val16 oldLogE[channels*mode->nbEBands]; */ /* opus_val16 oldLogE[channels*mode->nbEBands]; */
/* opus_val16 oldLogE2[channels*mode->nbEBands]; */ /* opus_val16 oldLogE2[channels*mode->nbEBands]; */
/* opus_val16 energyError[channels*mode->nbEBands]; */
return size; return size;
} }
@ -178,7 +181,6 @@ static int opus_custom_encoder_init_arch(CELTEncoder *st, const CELTMode *mode,
st->start = 0; st->start = 0;
st->end = st->mode->effEBands; st->end = st->mode->effEBands;
st->signalling = 1; st->signalling = 1;
st->arch = arch; st->arch = arch;
st->constrained_vbr = 1; st->constrained_vbr = 1;
@ -223,7 +225,8 @@ void opus_custom_encoder_destroy(CELTEncoder *st)
static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int C, static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int C,
opus_val16 *tf_estimate, int *tf_chan) opus_val16 *tf_estimate, int *tf_chan, int allow_weak_transients,
int *weak_transient)
{ {
int i; int i;
VARDECL(opus_val16, tmp); VARDECL(opus_val16, tmp);
@ -233,6 +236,12 @@ static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int
int c; int c;
opus_val16 tf_max; opus_val16 tf_max;
int len2; int len2;
/* Forward masking: 6.7 dB/ms. */
#ifdef FIXED_POINT
int forward_shift = 4;
#else
opus_val16 forward_decay = QCONST16(.0625f,15);
#endif
/* Table of 6*64/x, trained on real data to minimize the average error */ /* Table of 6*64/x, trained on real data to minimize the average error */
static const unsigned char inv_table[128] = { static const unsigned char inv_table[128] = {
255,255,156,110, 86, 70, 59, 51, 45, 40, 37, 33, 31, 28, 26, 25, 255,255,156,110, 86, 70, 59, 51, 45, 40, 37, 33, 31, 28, 26, 25,
@ -247,6 +256,19 @@ static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int
SAVE_STACK; SAVE_STACK;
ALLOC(tmp, len, opus_val16); ALLOC(tmp, len, opus_val16);
*weak_transient = 0;
/* For lower bitrates, let's be more conservative and have a forward masking
decay of 3.3 dB/ms. This avoids having to code transients at very low
bitrate (mostly for hybrid), which can result in unstable energy and/or
partial collapse. */
if (allow_weak_transients)
{
#ifdef FIXED_POINT
forward_shift = 5;
#else
forward_decay = QCONST16(.03125f,15);
#endif
}
len2=len/2; len2=len/2;
for (c=0;c<C;c++) for (c=0;c<C;c++)
{ {
@ -269,7 +291,7 @@ static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int
mem0 = mem1 + y - 2*x; mem0 = mem1 + y - 2*x;
mem1 = x - .5f*y; mem1 = x - .5f*y;
#endif #endif
tmp[i] = EXTRACT16(SHR32(y,2)); tmp[i] = SROUND16(y, 2);
/*printf("%f ", tmp[i]);*/ /*printf("%f ", tmp[i]);*/
} }
/*printf("\n");*/ /*printf("\n");*/
@ -280,7 +302,7 @@ static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int
/* Normalize tmp to max range */ /* Normalize tmp to max range */
{ {
int shift=0; int shift=0;
shift = 14-celt_ilog2(1+celt_maxabs16(tmp, len)); shift = 14-celt_ilog2(MAX16(1, celt_maxabs16(tmp, len)));
if (shift!=0) if (shift!=0)
{ {
for (i=0;i<len;i++) for (i=0;i<len;i++)
@ -299,9 +321,9 @@ static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int
mean += x2; mean += x2;
#ifdef FIXED_POINT #ifdef FIXED_POINT
/* FIXME: Use PSHR16() instead */ /* FIXME: Use PSHR16() instead */
tmp[i] = mem0 + PSHR32(x2-mem0,4); tmp[i] = mem0 + PSHR32(x2-mem0,forward_shift);
#else #else
tmp[i] = mem0 + MULT16_16_P15(QCONST16(.0625f,15),x2-mem0); tmp[i] = mem0 + MULT16_16_P15(forward_decay,x2-mem0);
#endif #endif
mem0 = tmp[i]; mem0 = tmp[i];
} }
@ -311,6 +333,7 @@ static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int
/* Backward pass to compute the pre-echo threshold */ /* Backward pass to compute the pre-echo threshold */
for (i=len2-1;i>=0;i--) for (i=len2-1;i>=0;i--)
{ {
/* Backward masking: 13.9 dB/ms. */
#ifdef FIXED_POINT #ifdef FIXED_POINT
/* FIXME: Use PSHR16() instead */ /* FIXME: Use PSHR16() instead */
tmp[i] = mem0 + PSHR32(tmp[i]-mem0,3); tmp[i] = mem0 + PSHR32(tmp[i]-mem0,3);
@ -339,6 +362,12 @@ static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int
/* Compute harmonic mean discarding the unreliable boundaries /* Compute harmonic mean discarding the unreliable boundaries
The data is smooth, so we only take 1/4th of the samples */ The data is smooth, so we only take 1/4th of the samples */
unmask=0; unmask=0;
/* We should never see NaNs here. If we find any, then something really bad happened and we better abort
before it does any damage later on. If these asserts are disabled (no hardening), then the table
lookup a few lines below (id = ...) is likely to crash dur to an out-of-bounds read. DO NOT FIX
that crash on NaN since it could result in a worse issue later on. */
celt_assert(!celt_isnan(tmp[0]));
celt_assert(!celt_isnan(norm));
for (i=12;i<len2-5;i+=4) for (i=12;i<len2-5;i+=4)
{ {
int id; int id;
@ -359,7 +388,12 @@ static int transient_analysis(const opus_val32 * OPUS_RESTRICT in, int len, int
} }
} }
is_transient = mask_metric>200; is_transient = mask_metric>200;
/* For low bitrates, define "weak transients" that need to be
handled differently to avoid partial collapse. */
if (allow_weak_transients && is_transient && mask_metric<600) {
is_transient = 0;
*weak_transient = 1;
}
/* Arbitrary metric for VBR boost */ /* Arbitrary metric for VBR boost */
tf_max = MAX16(0,celt_sqrt(27*mask_metric)-42); tf_max = MAX16(0,celt_sqrt(27*mask_metric)-42);
/* *tf_estimate = 1 + MIN16(1, sqrt(MAX16(0, tf_max-30))/20); */ /* *tf_estimate = 1 + MIN16(1, sqrt(MAX16(0, tf_max-30))/20); */
@ -549,7 +583,7 @@ static opus_val32 l1_metric(const celt_norm *tmp, int N, int LM, opus_val16 bias
static int tf_analysis(const CELTMode *m, int len, int isTransient, static int tf_analysis(const CELTMode *m, int len, int isTransient,
int *tf_res, int lambda, celt_norm *X, int N0, int LM, int *tf_res, int lambda, celt_norm *X, int N0, int LM,
int *tf_sum, opus_val16 tf_estimate, int tf_chan) opus_val16 tf_estimate, int tf_chan, int *importance)
{ {
int i; int i;
VARDECL(int, metric); VARDECL(int, metric);
@ -574,7 +608,6 @@ static int tf_analysis(const CELTMode *m, int len, int isTransient,
ALLOC(path0, len, int); ALLOC(path0, len, int);
ALLOC(path1, len, int); ALLOC(path1, len, int);
*tf_sum = 0;
for (i=0;i<len;i++) for (i=0;i<len;i++)
{ {
int k, N; int k, N;
@ -629,27 +662,26 @@ static int tf_analysis(const CELTMode *m, int len, int isTransient,
metric[i] = 2*best_level; metric[i] = 2*best_level;
else else
metric[i] = -2*best_level; metric[i] = -2*best_level;
*tf_sum += (isTransient ? LM : 0) - metric[i]/2;
/* For bands that can't be split to -1, set the metric to the half-way point to avoid /* For bands that can't be split to -1, set the metric to the half-way point to avoid
biasing the decision */ biasing the decision */
if (narrow && (metric[i]==0 || metric[i]==-2*LM)) if (narrow && (metric[i]==0 || metric[i]==-2*LM))
metric[i]-=1; metric[i]-=1;
/*printf("%d ", metric[i]);*/ /*printf("%d ", metric[i]/2 + (!isTransient)*LM);*/
} }
/*printf("\n");*/ /*printf("\n");*/
/* Search for the optimal tf resolution, including tf_select */ /* Search for the optimal tf resolution, including tf_select */
tf_select = 0; tf_select = 0;
for (sel=0;sel<2;sel++) for (sel=0;sel<2;sel++)
{ {
cost0 = 0; cost0 = importance[0]*abs(metric[0]-2*tf_select_table[LM][4*isTransient+2*sel+0]);
cost1 = isTransient ? 0 : lambda; cost1 = importance[0]*abs(metric[0]-2*tf_select_table[LM][4*isTransient+2*sel+1]) + (isTransient ? 0 : lambda);
for (i=1;i<len;i++) for (i=1;i<len;i++)
{ {
int curr0, curr1; int curr0, curr1;
curr0 = IMIN(cost0, cost1 + lambda); curr0 = IMIN(cost0, cost1 + lambda);
curr1 = IMIN(cost0 + lambda, cost1); curr1 = IMIN(cost0 + lambda, cost1);
cost0 = curr0 + abs(metric[i]-2*tf_select_table[LM][4*isTransient+2*sel+0]); cost0 = curr0 + importance[i]*abs(metric[i]-2*tf_select_table[LM][4*isTransient+2*sel+0]);
cost1 = curr1 + abs(metric[i]-2*tf_select_table[LM][4*isTransient+2*sel+1]); cost1 = curr1 + importance[i]*abs(metric[i]-2*tf_select_table[LM][4*isTransient+2*sel+1]);
} }
cost0 = IMIN(cost0, cost1); cost0 = IMIN(cost0, cost1);
selcost[sel]=cost0; selcost[sel]=cost0;
@ -658,8 +690,8 @@ static int tf_analysis(const CELTMode *m, int len, int isTransient,
* If tests confirm it's useful for non-transients, we could allow it. */ * If tests confirm it's useful for non-transients, we could allow it. */
if (selcost[1]<selcost[0] && isTransient) if (selcost[1]<selcost[0] && isTransient)
tf_select=1; tf_select=1;
cost0 = 0; cost0 = importance[0]*abs(metric[0]-2*tf_select_table[LM][4*isTransient+2*tf_select+0]);
cost1 = isTransient ? 0 : lambda; cost1 = importance[0]*abs(metric[0]-2*tf_select_table[LM][4*isTransient+2*tf_select+1]) + (isTransient ? 0 : lambda);
/* Viterbi forward pass */ /* Viterbi forward pass */
for (i=1;i<len;i++) for (i=1;i<len;i++)
{ {
@ -687,8 +719,8 @@ static int tf_analysis(const CELTMode *m, int len, int isTransient,
curr1 = from1; curr1 = from1;
path1[i]= 1; path1[i]= 1;
} }
cost0 = curr0 + abs(metric[i]-2*tf_select_table[LM][4*isTransient+2*tf_select+0]); cost0 = curr0 + importance[i]*abs(metric[i]-2*tf_select_table[LM][4*isTransient+2*tf_select+0]);
cost1 = curr1 + abs(metric[i]-2*tf_select_table[LM][4*isTransient+2*tf_select+1]); cost1 = curr1 + importance[i]*abs(metric[i]-2*tf_select_table[LM][4*isTransient+2*tf_select+1]);
} }
tf_res[len-1] = cost0 < cost1 ? 0 : 1; tf_res[len-1] = cost0 < cost1 ? 0 : 1;
/* Viterbi backward pass to check the decisions */ /* Viterbi backward pass to check the decisions */
@ -754,7 +786,7 @@ static void tf_encode(int start, int end, int isTransient, int *tf_res, int LM,
static int alloc_trim_analysis(const CELTMode *m, const celt_norm *X, static int alloc_trim_analysis(const CELTMode *m, const celt_norm *X,
const opus_val16 *bandLogE, int end, int LM, int C, int N0, const opus_val16 *bandLogE, int end, int LM, int C, int N0,
AnalysisInfo *analysis, opus_val16 *stereo_saving, opus_val16 tf_estimate, AnalysisInfo *analysis, opus_val16 *stereo_saving, opus_val16 tf_estimate,
int intensity, opus_val16 surround_trim, int arch) int intensity, opus_val16 surround_trim, opus_int32 equiv_rate, int arch)
{ {
int i; int i;
opus_val32 diff=0; opus_val32 diff=0;
@ -762,6 +794,14 @@ static int alloc_trim_analysis(const CELTMode *m, const celt_norm *X,
int trim_index; int trim_index;
opus_val16 trim = QCONST16(5.f, 8); opus_val16 trim = QCONST16(5.f, 8);
opus_val16 logXC, logXC2; opus_val16 logXC, logXC2;
/* At low bitrate, reducing the trim seems to help. At higher bitrates, it's less
clear what's best, so we're keeping it as it was before, at least for now. */
if (equiv_rate < 64000) {
trim = QCONST16(4.f, 8);
} else if (equiv_rate < 80000) {
opus_int32 frac = (equiv_rate-64000) >> 10;
trim = QCONST16(4.f, 8) + QCONST16(1.f/16.f, 8)*frac;
}
if (C==2) if (C==2)
{ {
opus_val16 sum = 0; /* Q10 */ opus_val16 sum = 0; /* Q10 */
@ -809,7 +849,7 @@ static int alloc_trim_analysis(const CELTMode *m, const celt_norm *X,
} while (++c<C); } while (++c<C);
diff /= C*(end-1); diff /= C*(end-1);
/*printf("%f\n", diff);*/ /*printf("%f\n", diff);*/
trim -= MAX16(-QCONST16(2.f, 8), MIN16(QCONST16(2.f, 8), SHR16(diff+QCONST16(1.f, DB_SHIFT),DB_SHIFT-8)/6 )); trim -= MAX32(-QCONST16(2.f, 8), MIN32(QCONST16(2.f, 8), SHR32(diff+QCONST16(1.f, DB_SHIFT),DB_SHIFT-8)/6 ));
trim -= SHR16(surround_trim, DB_SHIFT-8); trim -= SHR16(surround_trim, DB_SHIFT-8);
trim -= 2*SHR16(tf_estimate, 14-8); trim -= 2*SHR16(tf_estimate, 14-8);
#ifndef DISABLE_FLOAT_API #ifndef DISABLE_FLOAT_API
@ -930,7 +970,8 @@ static opus_val16 median_of_3(const opus_val16 *x)
static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16 *bandLogE2, static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16 *bandLogE2,
int nbEBands, int start, int end, int C, int *offsets, int lsb_depth, const opus_int16 *logN, int nbEBands, int start, int end, int C, int *offsets, int lsb_depth, const opus_int16 *logN,
int isTransient, int vbr, int constrained_vbr, const opus_int16 *eBands, int LM, int isTransient, int vbr, int constrained_vbr, const opus_int16 *eBands, int LM,
int effectiveBytes, opus_int32 *tot_boost_, int lfe, opus_val16 *surround_dynalloc) int effectiveBytes, opus_int32 *tot_boost_, int lfe, opus_val16 *surround_dynalloc,
AnalysisInfo *analysis, int *importance, int *spread_weight)
{ {
int i, c; int i, c;
opus_int32 tot_boost=0; opus_int32 tot_boost=0;
@ -956,6 +997,42 @@ static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16
for (i=0;i<end;i++) for (i=0;i<end;i++)
maxDepth = MAX16(maxDepth, bandLogE[c*nbEBands+i]-noise_floor[i]); maxDepth = MAX16(maxDepth, bandLogE[c*nbEBands+i]-noise_floor[i]);
} while (++c<C); } while (++c<C);
{
/* Compute a really simple masking model to avoid taking into account completely masked
bands when computing the spreading decision. */
VARDECL(opus_val16, mask);
VARDECL(opus_val16, sig);
ALLOC(mask, nbEBands, opus_val16);
ALLOC(sig, nbEBands, opus_val16);
for (i=0;i<end;i++)
mask[i] = bandLogE[i]-noise_floor[i];
if (C==2)
{
for (i=0;i<end;i++)
mask[i] = MAX16(mask[i], bandLogE[nbEBands+i]-noise_floor[i]);
}
OPUS_COPY(sig, mask, end);
for (i=1;i<end;i++)
mask[i] = MAX16(mask[i], mask[i-1] - QCONST16(2.f, DB_SHIFT));
for (i=end-2;i>=0;i--)
mask[i] = MAX16(mask[i], mask[i+1] - QCONST16(3.f, DB_SHIFT));
for (i=0;i<end;i++)
{
/* Compute SMR: Mask is never more than 72 dB below the peak and never below the noise floor.*/
opus_val16 smr = sig[i]-MAX16(MAX16(0, maxDepth-QCONST16(12.f, DB_SHIFT)), mask[i]);
/* Clamp SMR to make sure we're not shifting by something negative or too large. */
#ifdef FIXED_POINT
/* FIXME: Use PSHR16() instead */
int shift = -PSHR32(MAX16(-QCONST16(5.f, DB_SHIFT), MIN16(0, smr)), DB_SHIFT);
#else
int shift = IMIN(5, IMAX(0, -(int)floor(.5f + smr)));
#endif
spread_weight[i] = 32 >> shift;
}
/*for (i=0;i<end;i++)
printf("%d ", spread_weight[i]);
printf("\n");*/
}
/* Make sure that dynamic allocation can't make us bust the budget */ /* Make sure that dynamic allocation can't make us bust the budget */
if (effectiveBytes > 50 && LM>=1 && !lfe) if (effectiveBytes > 50 && LM>=1 && !lfe)
{ {
@ -1012,6 +1089,14 @@ static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16
} }
for (i=start;i<end;i++) for (i=start;i<end;i++)
follower[i] = MAX16(follower[i], surround_dynalloc[i]); follower[i] = MAX16(follower[i], surround_dynalloc[i]);
for (i=start;i<end;i++)
{
#ifdef FIXED_POINT
importance[i] = PSHR32(13*celt_exp2(MIN16(follower[i], QCONST16(4.f, DB_SHIFT))), 16);
#else
importance[i] = (int)floor(.5f+13*celt_exp2(MIN16(follower[i], QCONST16(4.f, DB_SHIFT))));
#endif
}
/* For non-transient CBR/CVBR frames, halve the dynalloc contribution */ /* For non-transient CBR/CVBR frames, halve the dynalloc contribution */
if ((!vbr || constrained_vbr)&&!isTransient) if ((!vbr || constrained_vbr)&&!isTransient)
{ {
@ -1020,14 +1105,26 @@ static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16
} }
for (i=start;i<end;i++) for (i=start;i<end;i++)
{ {
int width;
int boost;
int boost_bits;
if (i<8) if (i<8)
follower[i] *= 2; follower[i] *= 2;
if (i>=12) if (i>=12)
follower[i] = HALF16(follower[i]); follower[i] = HALF16(follower[i]);
}
#ifdef DISABLE_FLOAT_API
(void)analysis;
#else
if (analysis->valid)
{
for (i=start;i<IMIN(LEAK_BANDS, end);i++)
follower[i] = follower[i] + QCONST16(1.f/64.f, DB_SHIFT)*analysis->leak_boost[i];
}
#endif
for (i=start;i<end;i++)
{
int width;
int boost;
int boost_bits;
follower[i] = MIN16(follower[i], QCONST16(4, DB_SHIFT)); follower[i] = MIN16(follower[i], QCONST16(4, DB_SHIFT));
width = C*(eBands[i+1]-eBands[i])<<LM; width = C*(eBands[i+1]-eBands[i])<<LM;
@ -1042,11 +1139,11 @@ static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16
boost = (int)SHR32(EXTEND32(follower[i])*width/6,DB_SHIFT); boost = (int)SHR32(EXTEND32(follower[i])*width/6,DB_SHIFT);
boost_bits = boost*6<<BITRES; boost_bits = boost*6<<BITRES;
} }
/* For CBR and non-transient CVBR frames, limit dynalloc to 1/4 of the bits */ /* For CBR and non-transient CVBR frames, limit dynalloc to 2/3 of the bits */
if ((!vbr || (constrained_vbr&&!isTransient)) if ((!vbr || (constrained_vbr&&!isTransient))
&& (tot_boost+boost_bits)>>BITRES>>3 > effectiveBytes/4) && (tot_boost+boost_bits)>>BITRES>>3 > 2*effectiveBytes/3)
{ {
opus_int32 cap = ((effectiveBytes/4)<<BITRES<<3); opus_int32 cap = ((2*effectiveBytes/3)<<BITRES<<3);
offsets[i] = cap-tot_boost; offsets[i] = cap-tot_boost;
tot_boost = cap; tot_boost = cap;
break; break;
@ -1055,6 +1152,9 @@ static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16
tot_boost += boost_bits; tot_boost += boost_bits;
} }
} }
} else {
for (i=start;i<end;i++)
importance[i] = 13;
} }
*tot_boost_ = tot_boost; *tot_boost_ = tot_boost;
RESTORE_STACK; RESTORE_STACK;
@ -1063,7 +1163,7 @@ static opus_val16 dynalloc_analysis(const opus_val16 *bandLogE, const opus_val16
static int run_prefilter(CELTEncoder *st, celt_sig *in, celt_sig *prefilter_mem, int CC, int N, static int run_prefilter(CELTEncoder *st, celt_sig *in, celt_sig *prefilter_mem, int CC, int N,
int prefilter_tapset, int *pitch, opus_val16 *gain, int *qgain, int enabled, int nbAvailableBytes) int prefilter_tapset, int *pitch, opus_val16 *gain, int *qgain, int enabled, int nbAvailableBytes, AnalysisInfo *analysis)
{ {
int c; int c;
VARDECL(celt_sig, _pre); VARDECL(celt_sig, _pre);
@ -1119,7 +1219,12 @@ static int run_prefilter(CELTEncoder *st, celt_sig *in, celt_sig *prefilter_mem,
gain1 = 0; gain1 = 0;
pitch_index = COMBFILTER_MINPERIOD; pitch_index = COMBFILTER_MINPERIOD;
} }
#ifndef DISABLE_FLOAT_API
if (analysis->valid)
gain1 = (opus_val16)(gain1 * analysis->max_pitch_ratio);
#else
(void)analysis;
#endif
/* Gain threshold for enabling the prefilter/postfilter */ /* Gain threshold for enabling the prefilter/postfilter */
pf_threshold = QCONST16(.2f,15); pf_threshold = QCONST16(.2f,15);
@ -1175,10 +1280,10 @@ static int run_prefilter(CELTEncoder *st, celt_sig *in, celt_sig *prefilter_mem,
if (N>COMBFILTER_MAXPERIOD) if (N>COMBFILTER_MAXPERIOD)
{ {
OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, pre[c]+N, COMBFILTER_MAXPERIOD); OPUS_COPY(prefilter_mem+c*COMBFILTER_MAXPERIOD, pre[c]+N, COMBFILTER_MAXPERIOD);
} else { } else {
OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, prefilter_mem+c*COMBFILTER_MAXPERIOD+N, COMBFILTER_MAXPERIOD-N); OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD, prefilter_mem+c*COMBFILTER_MAXPERIOD+N, COMBFILTER_MAXPERIOD-N);
OPUS_MOVE(prefilter_mem+c*COMBFILTER_MAXPERIOD+COMBFILTER_MAXPERIOD-N, pre[c]+COMBFILTER_MAXPERIOD, N); OPUS_COPY(prefilter_mem+c*COMBFILTER_MAXPERIOD+COMBFILTER_MAXPERIOD-N, pre[c]+COMBFILTER_MAXPERIOD, N);
} }
} while (++c<CC); } while (++c<CC);
@ -1193,7 +1298,7 @@ static int compute_vbr(const CELTMode *mode, AnalysisInfo *analysis, opus_int32
int LM, opus_int32 bitrate, int lastCodedBands, int C, int intensity, int LM, opus_int32 bitrate, int lastCodedBands, int C, int intensity,
int constrained_vbr, opus_val16 stereo_saving, int tot_boost, int constrained_vbr, opus_val16 stereo_saving, int tot_boost,
opus_val16 tf_estimate, int pitch_change, opus_val16 maxDepth, opus_val16 tf_estimate, int pitch_change, opus_val16 maxDepth,
int variable_duration, int lfe, int has_surround_mask, opus_val16 surround_masking, int lfe, int has_surround_mask, opus_val16 surround_masking,
opus_val16 temporal_vbr) opus_val16 temporal_vbr)
{ {
/* The target rate in 8th bits per frame */ /* The target rate in 8th bits per frame */
@ -1235,10 +1340,9 @@ static int compute_vbr(const CELTMode *mode, AnalysisInfo *analysis, opus_int32
SHR32(MULT16_16(stereo_saving-QCONST16(0.1f,8),(coded_stereo_dof<<BITRES)),8)); SHR32(MULT16_16(stereo_saving-QCONST16(0.1f,8),(coded_stereo_dof<<BITRES)),8));
} }
/* Boost the rate according to dynalloc (minus the dynalloc average for calibration). */ /* Boost the rate according to dynalloc (minus the dynalloc average for calibration). */
target += tot_boost-(16<<LM); target += tot_boost-(19<<LM);
/* Apply transient boost, compensating for average boost. */ /* Apply transient boost, compensating for average boost. */
tf_calibration = variable_duration==OPUS_FRAMESIZE_VARIABLE ? tf_calibration = QCONST16(0.044f,14);
QCONST16(0.02f,14) : QCONST16(0.04f,14);
target += (opus_int32)SHL32(MULT16_32_Q15(tf_estimate-tf_calibration, target),1); target += (opus_int32)SHL32(MULT16_32_Q15(tf_estimate-tf_calibration, target),1);
#ifndef DISABLE_FLOAT_API #ifndef DISABLE_FLOAT_API
@ -1249,7 +1353,7 @@ static int compute_vbr(const CELTMode *mode, AnalysisInfo *analysis, opus_int32
float tonal; float tonal;
/* Tonality boost (compensating for the average). */ /* Tonality boost (compensating for the average). */
tonal = MAX16(0.f,analysis->tonality-.15f)-0.09f; tonal = MAX16(0.f,analysis->tonality-.15f)-0.12f;
tonal_target = target + (opus_int32)((coded_bins<<BITRES)*1.2f*tonal); tonal_target = target + (opus_int32)((coded_bins<<BITRES)*1.2f*tonal);
if (pitch_change) if (pitch_change)
tonal_target += (opus_int32)((coded_bins<<BITRES)*.8f); tonal_target += (opus_int32)((coded_bins<<BITRES)*.8f);
@ -1279,18 +1383,11 @@ static int compute_vbr(const CELTMode *mode, AnalysisInfo *analysis, opus_int32
/*printf("%f %d\n", maxDepth, floor_depth);*/ /*printf("%f %d\n", maxDepth, floor_depth);*/
} }
if ((!has_surround_mask||lfe) && (constrained_vbr || bitrate<64000)) /* Make VBR less aggressive for constrained VBR because we can't keep a higher bitrate
for long. Needs tuning. */
if ((!has_surround_mask||lfe) && constrained_vbr)
{ {
opus_val16 rate_factor; target = base_target + (opus_int32)MULT16_32_Q15(QCONST16(0.67f, 15), target-base_target);
#ifdef FIXED_POINT
rate_factor = MAX16(0,(bitrate-32000));
#else
rate_factor = MAX16(0,(1.f/32768)*(bitrate-32000));
#endif
if (constrained_vbr)
rate_factor = MIN16(rate_factor, QCONST16(0.67f, 15));
target = base_target + (opus_int32)MULT16_32_Q15(rate_factor, target-base_target);
} }
if (!has_surround_mask && tf_estimate < QCONST16(.2f, 14)) if (!has_surround_mask && tf_estimate < QCONST16(.2f, 14))
@ -1324,11 +1421,13 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
VARDECL(int, pulses); VARDECL(int, pulses);
VARDECL(int, cap); VARDECL(int, cap);
VARDECL(int, offsets); VARDECL(int, offsets);
VARDECL(int, importance);
VARDECL(int, spread_weight);
VARDECL(int, fine_priority); VARDECL(int, fine_priority);
VARDECL(int, tf_res); VARDECL(int, tf_res);
VARDECL(unsigned char, collapse_masks); VARDECL(unsigned char, collapse_masks);
celt_sig *prefilter_mem; celt_sig *prefilter_mem;
opus_val16 *oldBandE, *oldLogE, *oldLogE2; opus_val16 *oldBandE, *oldLogE, *oldLogE2, *energyError;
int shortBlocks=0; int shortBlocks=0;
int isTransient=0; int isTransient=0;
const int CC = st->channels; const int CC = st->channels;
@ -1340,7 +1439,6 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
int end; int end;
int effEnd; int effEnd;
int codedBands; int codedBands;
int tf_sum;
int alloc_trim; int alloc_trim;
int pitch_index=COMBFILTER_MINPERIOD; int pitch_index=COMBFILTER_MINPERIOD;
opus_val16 gain1 = 0; opus_val16 gain1 = 0;
@ -1352,6 +1450,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
opus_int32 total_boost; opus_int32 total_boost;
opus_int32 balance; opus_int32 balance;
opus_int32 tell; opus_int32 tell;
opus_int32 tell0_frac;
int prefilter_tapset=0; int prefilter_tapset=0;
int pf_on; int pf_on;
int anti_collapse_rsv; int anti_collapse_rsv;
@ -1373,7 +1472,10 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
opus_val16 surround_masking=0; opus_val16 surround_masking=0;
opus_val16 temporal_vbr=0; opus_val16 temporal_vbr=0;
opus_val16 surround_trim = 0; opus_val16 surround_trim = 0;
opus_int32 equiv_rate = 510000; opus_int32 equiv_rate;
int hybrid;
int weak_transient = 0;
int enable_tf_analysis;
VARDECL(opus_val16, surround_dynalloc); VARDECL(opus_val16, surround_dynalloc);
ALLOC_STACK; ALLOC_STACK;
@ -1383,6 +1485,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
eBands = mode->eBands; eBands = mode->eBands;
start = st->start; start = st->start;
end = st->end; end = st->end;
hybrid = start != 0;
tf_estimate = 0; tf_estimate = 0;
if (nbCompressedBytes<2 || pcm==NULL) if (nbCompressedBytes<2 || pcm==NULL)
{ {
@ -1406,12 +1509,14 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
oldBandE = (opus_val16*)(st->in_mem+CC*(overlap+COMBFILTER_MAXPERIOD)); oldBandE = (opus_val16*)(st->in_mem+CC*(overlap+COMBFILTER_MAXPERIOD));
oldLogE = oldBandE + CC*nbEBands; oldLogE = oldBandE + CC*nbEBands;
oldLogE2 = oldLogE + CC*nbEBands; oldLogE2 = oldLogE + CC*nbEBands;
energyError = oldLogE2 + CC*nbEBands;
if (enc==NULL) if (enc==NULL)
{ {
tell=1; tell0_frac=tell=1;
nbFilledBytes=0; nbFilledBytes=0;
} else { } else {
tell0_frac=ec_tell_frac(enc);
tell=ec_tell(enc); tell=ec_tell(enc);
nbFilledBytes=(tell+4)>>3; nbFilledBytes=(tell+4)>>3;
} }
@ -1464,10 +1569,11 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
if (st->bitrate!=OPUS_BITRATE_MAX) if (st->bitrate!=OPUS_BITRATE_MAX)
nbCompressedBytes = IMAX(2, IMIN(nbCompressedBytes, nbCompressedBytes = IMAX(2, IMIN(nbCompressedBytes,
(tmp+4*mode->Fs)/(8*mode->Fs)-!!st->signalling)); (tmp+4*mode->Fs)/(8*mode->Fs)-!!st->signalling));
effectiveBytes = nbCompressedBytes; effectiveBytes = nbCompressedBytes - nbFilledBytes;
} }
equiv_rate = ((opus_int32)nbCompressedBytes*8*50 >> (3-LM)) - (40*C+20)*((400>>LM) - 50);
if (st->bitrate != OPUS_BITRATE_MAX) if (st->bitrate != OPUS_BITRATE_MAX)
equiv_rate = st->bitrate - (40*C+20)*((400>>LM) - 50); equiv_rate = IMIN(equiv_rate, st->bitrate - (40*C+20)*((400>>LM) - 50));
if (enc==NULL) if (enc==NULL)
{ {
@ -1555,17 +1661,17 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
{ {
int enabled; int enabled;
int qg; int qg;
enabled = ((st->lfe&&nbAvailableBytes>3) || nbAvailableBytes>12*C) && start==0 && !silence && !st->disable_pf enabled = ((st->lfe&&nbAvailableBytes>3) || nbAvailableBytes>12*C) && !hybrid && !silence && !st->disable_pf
&& st->complexity >= 5 && !(st->consec_transient && LM!=3 && st->variable_duration==OPUS_FRAMESIZE_VARIABLE); && st->complexity >= 5;
prefilter_tapset = st->tapset_decision; prefilter_tapset = st->tapset_decision;
pf_on = run_prefilter(st, in, prefilter_mem, CC, N, prefilter_tapset, &pitch_index, &gain1, &qg, enabled, nbAvailableBytes); pf_on = run_prefilter(st, in, prefilter_mem, CC, N, prefilter_tapset, &pitch_index, &gain1, &qg, enabled, nbAvailableBytes, &st->analysis);
if ((gain1 > QCONST16(.4f,15) || st->prefilter_gain > QCONST16(.4f,15)) && (!st->analysis.valid || st->analysis.tonality > .3) if ((gain1 > QCONST16(.4f,15) || st->prefilter_gain > QCONST16(.4f,15)) && (!st->analysis.valid || st->analysis.tonality > .3)
&& (pitch_index > 1.26*st->prefilter_period || pitch_index < .79*st->prefilter_period)) && (pitch_index > 1.26*st->prefilter_period || pitch_index < .79*st->prefilter_period))
pitch_change = 1; pitch_change = 1;
if (pf_on==0) if (pf_on==0)
{ {
if(start==0 && tell+16<=total_bits) if(!hybrid && tell+16<=total_bits)
ec_enc_bit_logp(enc, 0, 1); ec_enc_bit_logp(enc, 0, 1);
} else { } else {
/*This block is not gated by a total bits check only because /*This block is not gated by a total bits check only because
@ -1586,8 +1692,12 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
shortBlocks = 0; shortBlocks = 0;
if (st->complexity >= 1 && !st->lfe) if (st->complexity >= 1 && !st->lfe)
{ {
/* Reduces the likelihood of energy instability on fricatives at low bitrate
in hybrid mode. It seems like we still want to have real transients on vowels
though (small SILK quantization offset value). */
int allow_weak_transients = hybrid && effectiveBytes<15 && st->silk_info.signalType != 2;
isTransient = transient_analysis(in, N+overlap, CC, isTransient = transient_analysis(in, N+overlap, CC,
&tf_estimate, &tf_chan); &tf_estimate, &tf_chan, allow_weak_transients, &weak_transient);
} }
if (LM>0 && ec_tell(enc)+3<=total_bits) if (LM>0 && ec_tell(enc)+3<=total_bits)
{ {
@ -1607,16 +1717,19 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
if (secondMdct) if (secondMdct)
{ {
compute_mdcts(mode, 0, in, freq, C, CC, LM, st->upsample, st->arch); compute_mdcts(mode, 0, in, freq, C, CC, LM, st->upsample, st->arch);
compute_band_energies(mode, freq, bandE, effEnd, C, LM); compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch);
amp2Log2(mode, effEnd, end, bandE, bandLogE2, C); amp2Log2(mode, effEnd, end, bandE, bandLogE2, C);
for (i=0;i<C*nbEBands;i++) for (i=0;i<C*nbEBands;i++)
bandLogE2[i] += HALF16(SHL16(LM, DB_SHIFT)); bandLogE2[i] += HALF16(SHL16(LM, DB_SHIFT));
} }
compute_mdcts(mode, shortBlocks, in, freq, C, CC, LM, st->upsample, st->arch); compute_mdcts(mode, shortBlocks, in, freq, C, CC, LM, st->upsample, st->arch);
/* This should catch any NaN in the CELT input. Since we're not supposed to see any (they're filtered
at the Opus layer), just abort. */
celt_assert(!celt_isnan(freq[0]) && (C==1 || !celt_isnan(freq[N])));
if (CC==2&&C==1) if (CC==2&&C==1)
tf_chan = 0; tf_chan = 0;
compute_band_energies(mode, freq, bandE, effEnd, C, LM); compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch);
if (st->lfe) if (st->lfe)
{ {
@ -1631,7 +1744,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
ALLOC(surround_dynalloc, C*nbEBands, opus_val16); ALLOC(surround_dynalloc, C*nbEBands, opus_val16);
OPUS_CLEAR(surround_dynalloc, end); OPUS_CLEAR(surround_dynalloc, end);
/* This computes how much masking takes place between surround channels */ /* This computes how much masking takes place between surround channels */
if (start==0&&st->energy_mask&&!st->lfe) if (!hybrid&&st->energy_mask&&!st->lfe)
{ {
int mask_end; int mask_end;
int midband; int midband;
@ -1733,14 +1846,14 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
/* Last chance to catch any transient we might have missed in the /* Last chance to catch any transient we might have missed in the
time-domain analysis */ time-domain analysis */
if (LM>0 && ec_tell(enc)+3<=total_bits && !isTransient && st->complexity>=5 && !st->lfe) if (LM>0 && ec_tell(enc)+3<=total_bits && !isTransient && st->complexity>=5 && !st->lfe && !hybrid)
{ {
if (patch_transient_decision(bandLogE, oldBandE, nbEBands, start, end, C)) if (patch_transient_decision(bandLogE, oldBandE, nbEBands, start, end, C))
{ {
isTransient = 1; isTransient = 1;
shortBlocks = M; shortBlocks = M;
compute_mdcts(mode, shortBlocks, in, freq, C, CC, LM, st->upsample, st->arch); compute_mdcts(mode, shortBlocks, in, freq, C, CC, LM, st->upsample, st->arch);
compute_band_energies(mode, freq, bandE, effEnd, C, LM); compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch);
amp2Log2(mode, effEnd, end, bandE, bandLogE, C); amp2Log2(mode, effEnd, end, bandE, bandLogE, C);
/* Compensate for the scaling of short vs long mdcts */ /* Compensate for the scaling of short vs long mdcts */
for (i=0;i<C*nbEBands;i++) for (i=0;i<C*nbEBands;i++)
@ -1757,31 +1870,59 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
/* Band normalisation */ /* Band normalisation */
normalise_bands(mode, freq, X, bandE, effEnd, C, M); normalise_bands(mode, freq, X, bandE, effEnd, C, M);
enable_tf_analysis = effectiveBytes>=15*C && !hybrid && st->complexity>=2 && !st->lfe;
ALLOC(offsets, nbEBands, int);
ALLOC(importance, nbEBands, int);
ALLOC(spread_weight, nbEBands, int);
maxDepth = dynalloc_analysis(bandLogE, bandLogE2, nbEBands, start, end, C, offsets,
st->lsb_depth, mode->logN, isTransient, st->vbr, st->constrained_vbr,
eBands, LM, effectiveBytes, &tot_boost, st->lfe, surround_dynalloc, &st->analysis, importance, spread_weight);
ALLOC(tf_res, nbEBands, int); ALLOC(tf_res, nbEBands, int);
/* Disable variable tf resolution for hybrid and at very low bitrate */ /* Disable variable tf resolution for hybrid and at very low bitrate */
if (effectiveBytes>=15*C && start==0 && st->complexity>=2 && !st->lfe) if (enable_tf_analysis)
{ {
int lambda; int lambda;
if (effectiveBytes<40) lambda = IMAX(80, 20480/effectiveBytes + 2);
lambda = 12; tf_select = tf_analysis(mode, effEnd, isTransient, tf_res, lambda, X, N, LM, tf_estimate, tf_chan, importance);
else if (effectiveBytes<60)
lambda = 6;
else if (effectiveBytes<100)
lambda = 4;
else
lambda = 3;
lambda*=2;
tf_select = tf_analysis(mode, effEnd, isTransient, tf_res, lambda, X, N, LM, &tf_sum, tf_estimate, tf_chan);
for (i=effEnd;i<end;i++) for (i=effEnd;i<end;i++)
tf_res[i] = tf_res[effEnd-1]; tf_res[i] = tf_res[effEnd-1];
} else if (hybrid && weak_transient)
{
/* For weak transients, we rely on the fact that improving time resolution using
TF on a long window is imperfect and will not result in an energy collapse at
low bitrate. */
for (i=0;i<end;i++)
tf_res[i] = 1;
tf_select=0;
} else if (hybrid && effectiveBytes<15 && st->silk_info.signalType != 2)
{
/* For low bitrate hybrid, we force temporal resolution to 5 ms rather than 2.5 ms. */
for (i=0;i<end;i++)
tf_res[i] = 0;
tf_select=isTransient;
} else { } else {
tf_sum = 0;
for (i=0;i<end;i++) for (i=0;i<end;i++)
tf_res[i] = isTransient; tf_res[i] = isTransient;
tf_select=0; tf_select=0;
} }
ALLOC(error, C*nbEBands, opus_val16); ALLOC(error, C*nbEBands, opus_val16);
c=0;
do {
for (i=start;i<end;i++)
{
/* When the energy is stable, slightly bias energy quantization towards
the previous error to make the gain more stable (a constant offset is
better than fluctuations). */
if (ABS32(SUB32(bandLogE[i+c*nbEBands], oldBandE[i+c*nbEBands])) < QCONST16(2.f, DB_SHIFT))
{
bandLogE[i+c*nbEBands] -= MULT16_16_Q15(energyError[i+c*nbEBands], QCONST16(0.25f, 15));
}
}
} while (++c < C);
quant_coarse_energy(mode, start, end, effEnd, bandLogE, quant_coarse_energy(mode, start, end, effEnd, bandLogE,
oldBandE, total_bits, error, enc, oldBandE, total_bits, error, enc,
C, LM, nbAvailableBytes, st->force_intra, C, LM, nbAvailableBytes, st->force_intra,
@ -1795,7 +1936,15 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
{ {
st->tapset_decision = 0; st->tapset_decision = 0;
st->spread_decision = SPREAD_NORMAL; st->spread_decision = SPREAD_NORMAL;
} else if (shortBlocks || st->complexity < 3 || nbAvailableBytes < 10*C || start != 0) } else if (hybrid)
{
if (st->complexity == 0)
st->spread_decision = SPREAD_NONE;
else if (isTransient)
st->spread_decision = SPREAD_NORMAL;
else
st->spread_decision = SPREAD_AGGRESSIVE;
} else if (shortBlocks || st->complexity < 3 || nbAvailableBytes < 10*C)
{ {
if (st->complexity == 0) if (st->complexity == 0)
st->spread_decision = SPREAD_NONE; st->spread_decision = SPREAD_NONE;
@ -1819,7 +1968,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
{ {
st->spread_decision = spreading_decision(mode, X, st->spread_decision = spreading_decision(mode, X,
&st->tonal_average, st->spread_decision, &st->hf_average, &st->tonal_average, st->spread_decision, &st->hf_average,
&st->tapset_decision, pf_on&&!shortBlocks, effEnd, C, M); &st->tapset_decision, pf_on&&!shortBlocks, effEnd, C, M, spread_weight);
} }
/*printf("%d %d\n", st->tapset_decision, st->spread_decision);*/ /*printf("%d %d\n", st->tapset_decision, st->spread_decision);*/
/*printf("%f %d %f %d\n\n", st->analysis.tonality, st->spread_decision, st->analysis.tonality_slope, st->tapset_decision);*/ /*printf("%f %d %f %d\n\n", st->analysis.tonality, st->spread_decision, st->analysis.tonality_slope, st->tapset_decision);*/
@ -1827,11 +1976,6 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
ec_enc_icdf(enc, st->spread_decision, spread_icdf, 5); ec_enc_icdf(enc, st->spread_decision, spread_icdf, 5);
} }
ALLOC(offsets, nbEBands, int);
maxDepth = dynalloc_analysis(bandLogE, bandLogE2, nbEBands, start, end, C, offsets,
st->lsb_depth, mode->logN, isTransient, st->vbr, st->constrained_vbr,
eBands, LM, effectiveBytes, &tot_boost, st->lfe, surround_dynalloc);
/* For LFE, everything interesting is in the first band */ /* For LFE, everything interesting is in the first band */
if (st->lfe) if (st->lfe)
offsets[0] = IMIN(8, effectiveBytes/3); offsets[0] = IMIN(8, effectiveBytes/3);
@ -1893,12 +2037,15 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
alloc_trim = 5; alloc_trim = 5;
if (tell+(6<<BITRES) <= total_bits - total_boost) if (tell+(6<<BITRES) <= total_bits - total_boost)
{ {
if (st->lfe) if (start > 0 || st->lfe)
{
st->stereo_saving = 0;
alloc_trim = 5; alloc_trim = 5;
else } else {
alloc_trim = alloc_trim_analysis(mode, X, bandLogE, alloc_trim = alloc_trim_analysis(mode, X, bandLogE,
end, LM, C, N, &st->analysis, &st->stereo_saving, tf_estimate, end, LM, C, N, &st->analysis, &st->stereo_saving, tf_estimate,
st->intensity, surround_trim, st->arch); st->intensity, surround_trim, equiv_rate, st->arch);
}
ec_enc_icdf(enc, alloc_trim, trim_icdf, 7); ec_enc_icdf(enc, alloc_trim, trim_icdf, 7);
tell = ec_tell_frac(enc); tell = ec_tell_frac(enc);
} }
@ -1916,17 +2063,36 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
/* Don't attempt to use more than 510 kb/s, even for frames smaller than 20 ms. /* Don't attempt to use more than 510 kb/s, even for frames smaller than 20 ms.
The CELT allocator will just not be able to use more than that anyway. */ The CELT allocator will just not be able to use more than that anyway. */
nbCompressedBytes = IMIN(nbCompressedBytes,1275>>(3-LM)); nbCompressedBytes = IMIN(nbCompressedBytes,1275>>(3-LM));
if (!hybrid)
{
base_target = vbr_rate - ((40*C+20)<<BITRES); base_target = vbr_rate - ((40*C+20)<<BITRES);
} else {
base_target = IMAX(0, vbr_rate - ((9*C+4)<<BITRES));
}
if (st->constrained_vbr) if (st->constrained_vbr)
base_target += (st->vbr_offset>>lm_diff); base_target += (st->vbr_offset>>lm_diff);
if (!hybrid)
{
target = compute_vbr(mode, &st->analysis, base_target, LM, equiv_rate, target = compute_vbr(mode, &st->analysis, base_target, LM, equiv_rate,
st->lastCodedBands, C, st->intensity, st->constrained_vbr, st->lastCodedBands, C, st->intensity, st->constrained_vbr,
st->stereo_saving, tot_boost, tf_estimate, pitch_change, maxDepth, st->stereo_saving, tot_boost, tf_estimate, pitch_change, maxDepth,
st->variable_duration, st->lfe, st->energy_mask!=NULL, surround_masking, st->lfe, st->energy_mask!=NULL, surround_masking,
temporal_vbr); temporal_vbr);
} else {
target = base_target;
/* Tonal frames (offset<100) need more bits than noisy (offset>100) ones. */
if (st->silk_info.offset < 100) target += 12 << BITRES >> (3-LM);
if (st->silk_info.offset > 100) target -= 18 << BITRES >> (3-LM);
/* Boosting bitrate on transients and vowels with significant temporal
spikes. */
target += (opus_int32)MULT16_16_Q14(tf_estimate-QCONST16(.25f,14), (50<<BITRES));
/* If we have a strong transient, let's make sure it has enough bits to code
the first two bands, so that it can use folding rather than noise. */
if (tf_estimate > QCONST16(.7f,14))
target = IMAX(target, 50<<BITRES);
}
/* The current offset is removed from the target and the space used /* The current offset is removed from the target and the space used
so far is added*/ so far is added*/
target=target+tell; target=target+tell;
@ -1934,11 +2100,16 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
result in the encoder running out of bits. result in the encoder running out of bits.
The margin of 2 bytes ensures that none of the bust-prevention logic The margin of 2 bytes ensures that none of the bust-prevention logic
in the decoder will have triggered so far. */ in the decoder will have triggered so far. */
min_allowed = ((tell+total_boost+(1<<(BITRES+3))-1)>>(BITRES+3)) + 2 - nbFilledBytes; min_allowed = ((tell+total_boost+(1<<(BITRES+3))-1)>>(BITRES+3)) + 2;
/* Take into account the 37 bits we need to have left in the packet to
signal a redundant frame in hybrid mode. Creating a shorter packet would
create an entropy coder desync. */
if (hybrid)
min_allowed = IMAX(min_allowed, (tell0_frac+(37<<BITRES)+total_boost+(1<<(BITRES+3))-1)>>(BITRES+3));
nbAvailableBytes = (target+(1<<(BITRES+2)))>>(BITRES+3); nbAvailableBytes = (target+(1<<(BITRES+2)))>>(BITRES+3);
nbAvailableBytes = IMAX(min_allowed,nbAvailableBytes); nbAvailableBytes = IMAX(min_allowed,nbAvailableBytes);
nbAvailableBytes = IMIN(nbCompressedBytes,nbAvailableBytes+nbFilledBytes) - nbFilledBytes; nbAvailableBytes = IMIN(nbCompressedBytes,nbAvailableBytes);
/* By how much did we "miss" the target on that frame */ /* By how much did we "miss" the target on that frame */
delta = target - vbr_rate; delta = target - vbr_rate;
@ -1985,7 +2156,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
st->vbr_reservoir = 0; st->vbr_reservoir = 0;
/*printf ("+%d\n", adjust);*/ /*printf ("+%d\n", adjust);*/
} }
nbCompressedBytes = IMIN(nbCompressedBytes,nbAvailableBytes+nbFilledBytes); nbCompressedBytes = IMIN(nbCompressedBytes,nbAvailableBytes);
/*printf("%d\n", nbCompressedBytes*50*8);*/ /*printf("%d\n", nbCompressedBytes*50*8);*/
/* This moves the raw bits to take into account the new compressed size */ /* This moves the raw bits to take into account the new compressed size */
ec_enc_shrink(enc, nbCompressedBytes); ec_enc_shrink(enc, nbCompressedBytes);
@ -2020,7 +2191,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
#endif #endif
if (st->lfe) if (st->lfe)
signalBandwidth = 1; signalBandwidth = 1;
codedBands = compute_allocation(mode, start, end, offsets, cap, codedBands = clt_compute_allocation(mode, start, end, offsets, cap,
alloc_trim, &st->intensity, &dual_stereo, bits, &balance, pulses, alloc_trim, &st->intensity, &dual_stereo, bits, &balance, pulses,
fine_quant, fine_priority, C, LM, enc, 1, st->lastCodedBands, signalBandwidth); fine_quant, fine_priority, C, LM, enc, 1, st->lastCodedBands, signalBandwidth);
if (st->lastCodedBands) if (st->lastCodedBands)
@ -2035,7 +2206,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
quant_all_bands(1, mode, start, end, X, C==2 ? X+N : NULL, collapse_masks, quant_all_bands(1, mode, start, end, X, C==2 ? X+N : NULL, collapse_masks,
bandE, pulses, shortBlocks, st->spread_decision, bandE, pulses, shortBlocks, st->spread_decision,
dual_stereo, st->intensity, tf_res, nbCompressedBytes*(8<<BITRES)-anti_collapse_rsv, dual_stereo, st->intensity, tf_res, nbCompressedBytes*(8<<BITRES)-anti_collapse_rsv,
balance, enc, LM, codedBands, &st->rng, st->arch); balance, enc, LM, codedBands, &st->rng, st->complexity, st->arch, st->disable_inv);
if (anti_collapse_rsv > 0) if (anti_collapse_rsv > 0)
{ {
@ -2046,6 +2217,14 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
ec_enc_bits(enc, anti_collapse_on, 1); ec_enc_bits(enc, anti_collapse_on, 1);
} }
quant_energy_finalise(mode, start, end, oldBandE, error, fine_quant, fine_priority, nbCompressedBytes*8-ec_tell(enc), enc, C); quant_energy_finalise(mode, start, end, oldBandE, error, fine_quant, fine_priority, nbCompressedBytes*8-ec_tell(enc), enc, C);
OPUS_CLEAR(energyError, nbEBands*CC);
c=0;
do {
for (i=start;i<end;i++)
{
energyError[i+c*nbEBands] = MAX16(-QCONST16(0.5f, 15), MIN16(QCONST16(0.5f, 15), error[i+c*nbEBands]));
}
} while (++c < C);
if (silence) if (silence)
{ {
@ -2318,10 +2497,24 @@ int opus_custom_encoder_ctl(CELTEncoder * OPUS_RESTRICT st, int request, ...)
*value=st->lsb_depth; *value=st->lsb_depth;
} }
break; break;
case OPUS_SET_EXPERT_FRAME_DURATION_REQUEST: case OPUS_SET_PHASE_INVERSION_DISABLED_REQUEST:
{ {
opus_int32 value = va_arg(ap, opus_int32); opus_int32 value = va_arg(ap, opus_int32);
st->variable_duration = value; if(value<0 || value>1)
{
goto bad_arg;
}
st->disable_inv = value;
}
break;
case OPUS_GET_PHASE_INVERSION_DISABLED_REQUEST:
{
opus_int32 *value = va_arg(ap, opus_int32*);
if (!value)
{
goto bad_arg;
}
*value = st->disable_inv;
} }
break; break;
case OPUS_RESET_STATE: case OPUS_RESET_STATE:
@ -2365,6 +2558,13 @@ int opus_custom_encoder_ctl(CELTEncoder * OPUS_RESTRICT st, int request, ...)
OPUS_COPY(&st->analysis, info, 1); OPUS_COPY(&st->analysis, info, 1);
} }
break; break;
case CELT_SET_SILK_INFO_REQUEST:
{
SILKInfo *info = va_arg(ap, SILKInfo *);
if (info)
OPUS_COPY(&st->silk_info, info, 1);
}
break;
case CELT_GET_MODE_REQUEST: case CELT_GET_MODE_REQUEST:
{ {
const CELTMode ** value = va_arg(ap, const CELTMode**); const CELTMode ** value = va_arg(ap, const CELTMode**);

View File

@ -49,8 +49,7 @@ int p
float *lpc = _lpc; float *lpc = _lpc;
#endif #endif
for (i = 0; i < p; i++) OPUS_CLEAR(lpc, p);
lpc[i] = 0;
if (ac[0] != 0) if (ac[0] != 0)
{ {
for (i = 0; i < p; i++) { for (i = 0; i < p; i++) {
@ -90,58 +89,40 @@ int p
void celt_fir_c( void celt_fir_c(
const opus_val16 *_x, const opus_val16 *x,
const opus_val16 *num, const opus_val16 *num,
opus_val16 *_y, opus_val16 *y,
int N, int N,
int ord, int ord,
opus_val16 *mem,
int arch) int arch)
{ {
int i,j; int i,j;
VARDECL(opus_val16, rnum); VARDECL(opus_val16, rnum);
VARDECL(opus_val16, x);
SAVE_STACK; SAVE_STACK;
celt_assert(x != y);
ALLOC(rnum, ord, opus_val16); ALLOC(rnum, ord, opus_val16);
ALLOC(x, N+ord, opus_val16);
for(i=0;i<ord;i++) for(i=0;i<ord;i++)
rnum[i] = num[ord-i-1]; rnum[i] = num[ord-i-1];
for(i=0;i<ord;i++)
x[i] = mem[ord-i-1];
for (i=0;i<N;i++)
x[i+ord]=_x[i];
for(i=0;i<ord;i++)
mem[i] = _x[N-i-1];
#ifdef SMALL_FOOTPRINT
(void)arch;
for (i=0;i<N;i++)
{
opus_val32 sum = SHL32(EXTEND32(_x[i]), SIG_SHIFT);
for (j=0;j<ord;j++)
{
sum = MAC16_16(sum,rnum[j],x[i+j]);
}
_y[i] = SATURATE16(PSHR32(sum, SIG_SHIFT));
}
#else
for (i=0;i<N-3;i+=4) for (i=0;i<N-3;i+=4)
{ {
opus_val32 sum[4]={0,0,0,0}; opus_val32 sum[4];
xcorr_kernel(rnum, x+i, sum, ord, arch); sum[0] = SHL32(EXTEND32(x[i ]), SIG_SHIFT);
_y[i ] = SATURATE16(ADD32(EXTEND32(_x[i ]), PSHR32(sum[0], SIG_SHIFT))); sum[1] = SHL32(EXTEND32(x[i+1]), SIG_SHIFT);
_y[i+1] = SATURATE16(ADD32(EXTEND32(_x[i+1]), PSHR32(sum[1], SIG_SHIFT))); sum[2] = SHL32(EXTEND32(x[i+2]), SIG_SHIFT);
_y[i+2] = SATURATE16(ADD32(EXTEND32(_x[i+2]), PSHR32(sum[2], SIG_SHIFT))); sum[3] = SHL32(EXTEND32(x[i+3]), SIG_SHIFT);
_y[i+3] = SATURATE16(ADD32(EXTEND32(_x[i+3]), PSHR32(sum[3], SIG_SHIFT))); xcorr_kernel(rnum, x+i-ord, sum, ord, arch);
y[i ] = ROUND16(sum[0], SIG_SHIFT);
y[i+1] = ROUND16(sum[1], SIG_SHIFT);
y[i+2] = ROUND16(sum[2], SIG_SHIFT);
y[i+3] = ROUND16(sum[3], SIG_SHIFT);
} }
for (;i<N;i++) for (;i<N;i++)
{ {
opus_val32 sum = 0; opus_val32 sum = SHL32(EXTEND32(x[i]), SIG_SHIFT);
for (j=0;j<ord;j++) for (j=0;j<ord;j++)
sum = MAC16_16(sum,rnum[j],x[i+j]); sum = MAC16_16(sum,rnum[j],x[i+j-ord]);
_y[i] = SATURATE16(ADD32(EXTEND32(_x[i]), PSHR32(sum, SIG_SHIFT))); y[i] = ROUND16(sum, SIG_SHIFT);
} }
#endif
RESTORE_STACK; RESTORE_STACK;
} }
@ -167,7 +148,7 @@ void celt_iir(const opus_val32 *_x,
{ {
mem[j]=mem[j-1]; mem[j]=mem[j-1];
} }
mem[0] = ROUND16(sum,SIG_SHIFT); mem[0] = SROUND16(sum, SIG_SHIFT);
_y[i] = sum; _y[i] = sum;
} }
#else #else
@ -196,20 +177,20 @@ void celt_iir(const opus_val32 *_x,
xcorr_kernel(rden, y+i, sum, ord, arch); xcorr_kernel(rden, y+i, sum, ord, arch);
/* Patch up the result to compensate for the fact that this is an IIR */ /* Patch up the result to compensate for the fact that this is an IIR */
y[i+ord ] = -ROUND16(sum[0],SIG_SHIFT); y[i+ord ] = -SROUND16(sum[0],SIG_SHIFT);
_y[i ] = sum[0]; _y[i ] = sum[0];
sum[1] = MAC16_16(sum[1], y[i+ord ], den[0]); sum[1] = MAC16_16(sum[1], y[i+ord ], den[0]);
y[i+ord+1] = -ROUND16(sum[1],SIG_SHIFT); y[i+ord+1] = -SROUND16(sum[1],SIG_SHIFT);
_y[i+1] = sum[1]; _y[i+1] = sum[1];
sum[2] = MAC16_16(sum[2], y[i+ord+1], den[0]); sum[2] = MAC16_16(sum[2], y[i+ord+1], den[0]);
sum[2] = MAC16_16(sum[2], y[i+ord ], den[1]); sum[2] = MAC16_16(sum[2], y[i+ord ], den[1]);
y[i+ord+2] = -ROUND16(sum[2],SIG_SHIFT); y[i+ord+2] = -SROUND16(sum[2],SIG_SHIFT);
_y[i+2] = sum[2]; _y[i+2] = sum[2];
sum[3] = MAC16_16(sum[3], y[i+ord+2], den[0]); sum[3] = MAC16_16(sum[3], y[i+ord+2], den[0]);
sum[3] = MAC16_16(sum[3], y[i+ord+1], den[1]); sum[3] = MAC16_16(sum[3], y[i+ord+1], den[1]);
sum[3] = MAC16_16(sum[3], y[i+ord ], den[2]); sum[3] = MAC16_16(sum[3], y[i+ord ], den[2]);
y[i+ord+3] = -ROUND16(sum[3],SIG_SHIFT); y[i+ord+3] = -SROUND16(sum[3],SIG_SHIFT);
_y[i+3] = sum[3]; _y[i+3] = sum[3];
} }
for (;i<N;i++) for (;i<N;i++)
@ -217,7 +198,7 @@ void celt_iir(const opus_val32 *_x,
opus_val32 sum = _x[i]; opus_val32 sum = _x[i];
for (j=0;j<ord;j++) for (j=0;j<ord;j++)
sum -= MULT16_16(rden[j],y[i+j]); sum -= MULT16_16(rden[j],y[i+j]);
y[i+ord] = ROUND16(sum,SIG_SHIFT); y[i+ord] = SROUND16(sum,SIG_SHIFT);
_y[i] = sum; _y[i] = sum;
} }
for(i=0;i<ord;i++) for(i=0;i<ord;i++)

View File

@ -45,12 +45,11 @@ void celt_fir_c(
opus_val16 *y, opus_val16 *y,
int N, int N,
int ord, int ord,
opus_val16 *mem,
int arch); int arch);
#if !defined(OVERRIDE_CELT_FIR) #if !defined(OVERRIDE_CELT_FIR)
#define celt_fir(x, num, y, N, ord, mem, arch) \ #define celt_fir(x, num, y, N, ord, arch) \
(celt_fir_c(x, num, y, N, ord, mem, arch)) (celt_fir_c(x, num, y, N, ord, arch))
#endif #endif
void celt_iir(const opus_val32 *x, void celt_iir(const opus_val32 *x,

View File

@ -74,7 +74,7 @@ int log2_frac(opus_uint32 val, int frac)
/*Although derived separately, the pulse vector coding scheme is equivalent to /*Although derived separately, the pulse vector coding scheme is equivalent to
a Pyramid Vector Quantizer \cite{Fis86}. a Pyramid Vector Quantizer \cite{Fis86}.
Some additional notes about an early version appear at Some additional notes about an early version appear at
http://people.xiph.org/~tterribe/notes/cwrs.html, but the codebook ordering https://people.xiph.org/~tterribe/notes/cwrs.html, but the codebook ordering
and the definitions of some terms have evolved since that was written. and the definitions of some terms have evolved since that was written.
The conversion from a pulse vector to an integer index (encoding) and back The conversion from a pulse vector to an integer index (encoding) and back
@ -482,7 +482,7 @@ static opus_val32 cwrsi(int _n,int _k,opus_uint32 _i,int *_y){
k0=_k; k0=_k;
q=row[_n]; q=row[_n];
if(q>_i){ if(q>_i){
celt_assert(p>q); celt_sig_assert(p>q);
_k=_n; _k=_n;
do p=CELT_PVQ_U_ROW[--_k][_n]; do p=CELT_PVQ_U_ROW[--_k][_n];
while(p>_i); while(p>_i);

View File

@ -122,7 +122,7 @@ opus_uint32 ec_tell_frac(ec_ctx *_this);
/* Tested exhaustively for all n and for 1<=d<=256 */ /* Tested exhaustively for all n and for 1<=d<=256 */
static OPUS_INLINE opus_uint32 celt_udiv(opus_uint32 n, opus_uint32 d) { static OPUS_INLINE opus_uint32 celt_udiv(opus_uint32 n, opus_uint32 d) {
celt_assert(d>0); celt_sig_assert(d>0);
#ifdef USE_SMALL_DIV_TABLE #ifdef USE_SMALL_DIV_TABLE
if (d>256) if (d>256)
return n/d; return n/d;
@ -138,7 +138,7 @@ static OPUS_INLINE opus_uint32 celt_udiv(opus_uint32 n, opus_uint32 d) {
} }
static OPUS_INLINE opus_int32 celt_sudiv(opus_int32 n, opus_int32 d) { static OPUS_INLINE opus_int32 celt_sudiv(opus_int32 n, opus_int32 d) {
celt_assert(d>0); celt_sig_assert(d>0);
#ifdef USE_SMALL_DIV_TABLE #ifdef USE_SMALL_DIV_TABLE
if (n<0) if (n<0)
return -(opus_int32)celt_udiv(-n, d); return -(opus_int32)celt_udiv(-n, d);

View File

@ -85,7 +85,7 @@ int ec_dec_icdf(ec_dec *_this,const unsigned char *_icdf,unsigned _ftb);
The bits must have been encoded with ec_enc_uint(). The bits must have been encoded with ec_enc_uint().
No call to ec_dec_update() is necessary after this call. No call to ec_dec_update() is necessary after this call.
_ft: The number of integers that can be decoded (one more than the max). _ft: The number of integers that can be decoded (one more than the max).
This must be at least one, and no more than 2**32-1. This must be at least 2, and no more than 2**32-1.
Return: The decoded bits.*/ Return: The decoded bits.*/
opus_uint32 ec_dec_uint(ec_dec *_this,opus_uint32 _ft); opus_uint32 ec_dec_uint(ec_dec *_this,opus_uint32 _ft);

View File

@ -67,7 +67,7 @@ void ec_enc_icdf(ec_enc *_this,int _s,const unsigned char *_icdf,unsigned _ftb);
/*Encodes a raw unsigned integer in the stream. /*Encodes a raw unsigned integer in the stream.
_fl: The integer to encode. _fl: The integer to encode.
_ft: The number of integers that can be encoded (one more than the max). _ft: The number of integers that can be encoded (one more than the max).
This must be at least one, and no more than 2**32-1.*/ This must be at least 2, and no more than 2**32-1.*/
void ec_enc_uint(ec_enc *_this,opus_uint32 _fl,opus_uint32 _ft); void ec_enc_uint(ec_enc *_this,opus_uint32 _fl,opus_uint32 _ft);
/*Encodes a sequence of raw bits in the stream. /*Encodes a sequence of raw bits in the stream.

View File

@ -59,6 +59,14 @@ extern opus_int64 celt_mips;
#define SHR(a,b) SHR32(a,b) #define SHR(a,b) SHR32(a,b)
#define PSHR(a,b) PSHR32(a,b) #define PSHR(a,b) PSHR32(a,b)
/** Add two 32-bit values, ignore any overflows */
#define ADD32_ovflw(a,b) (celt_mips+=2,(opus_val32)((opus_uint32)(a)+(opus_uint32)(b)))
/** Subtract two 32-bit values, ignore any overflows */
#define SUB32_ovflw(a,b) (celt_mips+=2,(opus_val32)((opus_uint32)(a)-(opus_uint32)(b)))
/* Avoid MSVC warning C4146: unary minus operator applied to unsigned type */
/** Negate 32-bit value, ignore any overflows */
#define NEG32_ovflw(a) (celt_mips+=2,(opus_val32)(0-(opus_uint32)(a)))
static OPUS_INLINE short NEG16(int x) static OPUS_INLINE short NEG16(int x)
{ {
int res; int res;
@ -227,12 +235,11 @@ static OPUS_INLINE int SHL32_(opus_int64 a, int shift, char *file, int line)
#define VSHR32(a, shift) (((shift)>0) ? SHR32(a, shift) : SHL32(a, -(shift))) #define VSHR32(a, shift) (((shift)>0) ? SHR32(a, shift) : SHL32(a, -(shift)))
#define ROUND16(x,a) (celt_mips--,EXTRACT16(PSHR32((x),(a)))) #define ROUND16(x,a) (celt_mips--,EXTRACT16(PSHR32((x),(a))))
#define SROUND16(x,a) (celt_mips--,EXTRACT16(SATURATE(PSHR32(x,a), 32767)));
#define HALF16(x) (SHR16(x,1)) #define HALF16(x) (SHR16(x,1))
#define HALF32(x) (SHR32(x,1)) #define HALF32(x) (SHR32(x,1))
//#define SHR(a,shift) ((a) >> (shift))
//#define SHL(a,shift) ((a) << (shift))
#define ADD16(a, b) ADD16_(a, b, __FILE__, __LINE__) #define ADD16(a, b) ADD16_(a, b, __FILE__, __LINE__)
static OPUS_INLINE short ADD16_(int a, int b, char *file, int line) static OPUS_INLINE short ADD16_(int a, int b, char *file, int line)
{ {

View File

@ -37,16 +37,32 @@
#define MULT16_16SU(a,b) ((opus_val32)(opus_val16)(a)*(opus_val32)(opus_uint16)(b)) #define MULT16_16SU(a,b) ((opus_val32)(opus_val16)(a)*(opus_val32)(opus_uint16)(b))
/** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */ /** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */
#if OPUS_FAST_INT64
#define MULT16_32_Q16(a,b) ((opus_val32)SHR((opus_int64)((opus_val16)(a))*(b),16))
#else
#define MULT16_32_Q16(a,b) ADD32(MULT16_16((a),SHR((b),16)), SHR(MULT16_16SU((a),((b)&0x0000ffff)),16)) #define MULT16_32_Q16(a,b) ADD32(MULT16_16((a),SHR((b),16)), SHR(MULT16_16SU((a),((b)&0x0000ffff)),16))
#endif
/** 16x32 multiplication, followed by a 16-bit shift right (round-to-nearest). Results fits in 32 bits */ /** 16x32 multiplication, followed by a 16-bit shift right (round-to-nearest). Results fits in 32 bits */
#if OPUS_FAST_INT64
#define MULT16_32_P16(a,b) ((opus_val32)PSHR((opus_int64)((opus_val16)(a))*(b),16))
#else
#define MULT16_32_P16(a,b) ADD32(MULT16_16((a),SHR((b),16)), PSHR(MULT16_16SU((a),((b)&0x0000ffff)),16)) #define MULT16_32_P16(a,b) ADD32(MULT16_16((a),SHR((b),16)), PSHR(MULT16_16SU((a),((b)&0x0000ffff)),16))
#endif
/** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */ /** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */
#if OPUS_FAST_INT64
#define MULT16_32_Q15(a,b) ((opus_val32)SHR((opus_int64)((opus_val16)(a))*(b),15))
#else
#define MULT16_32_Q15(a,b) ADD32(SHL(MULT16_16((a),SHR((b),16)),1), SHR(MULT16_16SU((a),((b)&0x0000ffff)),15)) #define MULT16_32_Q15(a,b) ADD32(SHL(MULT16_16((a),SHR((b),16)),1), SHR(MULT16_16SU((a),((b)&0x0000ffff)),15))
#endif
/** 32x32 multiplication, followed by a 31-bit shift right. Results fits in 32 bits */ /** 32x32 multiplication, followed by a 31-bit shift right. Results fits in 32 bits */
#if OPUS_FAST_INT64
#define MULT32_32_Q31(a,b) ((opus_val32)SHR((opus_int64)(a)*(opus_int64)(b),31))
#else
#define MULT32_32_Q31(a,b) ADD32(ADD32(SHL(MULT16_16(SHR((a),16),SHR((b),16)),1), SHR(MULT16_16SU(SHR((a),16),((b)&0x0000ffff)),15)), SHR(MULT16_16SU(SHR((b),16),((a)&0x0000ffff)),15)) #define MULT32_32_Q31(a,b) ADD32(ADD32(SHL(MULT16_16(SHR((a),16),SHR((b),16)),1), SHR(MULT16_16SU(SHR((a),16),((b)&0x0000ffff)),15)), SHR(MULT16_16SU(SHR((b),16),((a)&0x0000ffff)),15))
#endif
/** Compile-time conversion of float constant to 16-bit value */ /** Compile-time conversion of float constant to 16-bit value */
#define QCONST16(x,bits) ((opus_val16)(.5+(x)*(((opus_val32)1)<<(bits)))) #define QCONST16(x,bits) ((opus_val16)(.5+(x)*(((opus_val32)1)<<(bits))))
@ -88,6 +104,9 @@
/** Shift by a and round-to-neareast 32-bit value. Result is a 16-bit value */ /** Shift by a and round-to-neareast 32-bit value. Result is a 16-bit value */
#define ROUND16(x,a) (EXTRACT16(PSHR32((x),(a)))) #define ROUND16(x,a) (EXTRACT16(PSHR32((x),(a))))
/** Shift by a and round-to-neareast 32-bit value. Result is a saturated 16-bit value */
#define SROUND16(x,a) EXTRACT16(SATURATE(PSHR32(x,a), 32767));
/** Divide by two */ /** Divide by two */
#define HALF16(x) (SHR16(x,1)) #define HALF16(x) (SHR16(x,1))
#define HALF32(x) (SHR32(x,1)) #define HALF32(x) (SHR32(x,1))
@ -101,6 +120,14 @@
/** Subtract two 32-bit values */ /** Subtract two 32-bit values */
#define SUB32(a,b) ((opus_val32)(a)-(opus_val32)(b)) #define SUB32(a,b) ((opus_val32)(a)-(opus_val32)(b))
/** Add two 32-bit values, ignore any overflows */
#define ADD32_ovflw(a,b) ((opus_val32)((opus_uint32)(a)+(opus_uint32)(b)))
/** Subtract two 32-bit values, ignore any overflows */
#define SUB32_ovflw(a,b) ((opus_val32)((opus_uint32)(a)-(opus_uint32)(b)))
/* Avoid MSVC warning C4146: unary minus operator applied to unsigned type */
/** Negate 32-bit value, ignore any overflows */
#define NEG32_ovflw(a) ((opus_val32)(0-(opus_uint32)(a)))
/** 16x16 multiplication where the result fits in 16 bits */ /** 16x16 multiplication where the result fits in 16 bits */
#define MULT16_16_16(a,b) ((((opus_val16)(a))*((opus_val16)(b)))) #define MULT16_16_16(a,b) ((((opus_val16)(a))*((opus_val16)(b))))

View File

@ -61,7 +61,13 @@
** the config.h file. ** the config.h file.
*/ */
#if (HAVE_LRINTF) /* With GCC, when SSE is available, the fastest conversion is cvtss2si. */
#if defined(__GNUC__) && defined(__SSE__)
#include <xmmintrin.h>
static OPUS_INLINE opus_int32 float2int(float x) {return _mm_cvt_ss2si(_mm_set_ss(x));}
#elif defined(HAVE_LRINTF)
/* These defines enable functionality introduced with the 1999 ISO C /* These defines enable functionality introduced with the 1999 ISO C
** standard. They must be defined before the inclusion of math.h to ** standard. They must be defined before the inclusion of math.h to
@ -90,10 +96,10 @@
#include <math.h> #include <math.h>
#define float2int(x) lrint(x) #define float2int(x) lrint(x)
#elif (defined(_MSC_VER) && _MSC_VER >= 1400) && defined (_M_X64) #elif (defined(_MSC_VER) && _MSC_VER >= 1400) && (defined(_M_X64) || (defined(_M_IX86_FP) && _M_IX86_FP >= 1))
#include <xmmintrin.h> #include <xmmintrin.h>
__inline long int float2int(float value) static __inline long int float2int(float value)
{ {
return _mm_cvtss_si32(_mm_load_ss(&value)); return _mm_cvtss_si32(_mm_load_ss(&value));
} }
@ -104,7 +110,7 @@
** Therefore implement OPUS_INLINE versions of these functions here. ** Therefore implement OPUS_INLINE versions of these functions here.
*/ */
__inline long int static __inline long int
float2int (float flt) float2int (float flt)
{ int intgr; { int intgr;

View File

@ -82,8 +82,8 @@ static void kf_bfly2(
C_SUB( Fout2[0] , Fout[0] , t ); C_SUB( Fout2[0] , Fout[0] , t );
C_ADDTO( Fout[0] , t ); C_ADDTO( Fout[0] , t );
t.r = S_MUL(Fout2[1].r+Fout2[1].i, tw); t.r = S_MUL(ADD32_ovflw(Fout2[1].r, Fout2[1].i), tw);
t.i = S_MUL(Fout2[1].i-Fout2[1].r, tw); t.i = S_MUL(SUB32_ovflw(Fout2[1].i, Fout2[1].r), tw);
C_SUB( Fout2[1] , Fout[1] , t ); C_SUB( Fout2[1] , Fout[1] , t );
C_ADDTO( Fout[1] , t ); C_ADDTO( Fout[1] , t );
@ -92,8 +92,8 @@ static void kf_bfly2(
C_SUB( Fout2[2] , Fout[2] , t ); C_SUB( Fout2[2] , Fout[2] , t );
C_ADDTO( Fout[2] , t ); C_ADDTO( Fout[2] , t );
t.r = S_MUL(Fout2[3].i-Fout2[3].r, tw); t.r = S_MUL(SUB32_ovflw(Fout2[3].i, Fout2[3].r), tw);
t.i = S_MUL(-Fout2[3].i-Fout2[3].r, tw); t.i = S_MUL(NEG32_ovflw(ADD32_ovflw(Fout2[3].i, Fout2[3].r)), tw);
C_SUB( Fout2[3] , Fout[3] , t ); C_SUB( Fout2[3] , Fout[3] , t );
C_ADDTO( Fout[3] , t ); C_ADDTO( Fout[3] , t );
Fout += 8; Fout += 8;
@ -126,10 +126,10 @@ static void kf_bfly4(
C_ADDTO( *Fout , scratch1 ); C_ADDTO( *Fout , scratch1 );
C_SUB( scratch1 , Fout[1] , Fout[3] ); C_SUB( scratch1 , Fout[1] , Fout[3] );
Fout[1].r = scratch0.r + scratch1.i; Fout[1].r = ADD32_ovflw(scratch0.r, scratch1.i);
Fout[1].i = scratch0.i - scratch1.r; Fout[1].i = SUB32_ovflw(scratch0.i, scratch1.r);
Fout[3].r = scratch0.r - scratch1.i; Fout[3].r = SUB32_ovflw(scratch0.r, scratch1.i);
Fout[3].i = scratch0.i + scratch1.r; Fout[3].i = ADD32_ovflw(scratch0.i, scratch1.r);
Fout+=4; Fout+=4;
} }
} else { } else {
@ -160,10 +160,10 @@ static void kf_bfly4(
tw3 += fstride*3; tw3 += fstride*3;
C_ADDTO( *Fout , scratch[3] ); C_ADDTO( *Fout , scratch[3] );
Fout[m].r = scratch[5].r + scratch[4].i; Fout[m].r = ADD32_ovflw(scratch[5].r, scratch[4].i);
Fout[m].i = scratch[5].i - scratch[4].r; Fout[m].i = SUB32_ovflw(scratch[5].i, scratch[4].r);
Fout[m3].r = scratch[5].r - scratch[4].i; Fout[m3].r = SUB32_ovflw(scratch[5].r, scratch[4].i);
Fout[m3].i = scratch[5].i + scratch[4].r; Fout[m3].i = ADD32_ovflw(scratch[5].i, scratch[4].r);
++Fout; ++Fout;
} }
} }
@ -191,7 +191,7 @@ static void kf_bfly3(
kiss_fft_cpx * Fout_beg = Fout; kiss_fft_cpx * Fout_beg = Fout;
#ifdef FIXED_POINT #ifdef FIXED_POINT
epi3.r = -16384; /*epi3.r = -16384;*/ /* Unused */
epi3.i = -28378; epi3.i = -28378;
#else #else
epi3 = st->twiddles[fstride*m]; epi3 = st->twiddles[fstride*m];
@ -212,18 +212,18 @@ static void kf_bfly3(
tw1 += fstride; tw1 += fstride;
tw2 += fstride*2; tw2 += fstride*2;
Fout[m].r = Fout->r - HALF_OF(scratch[3].r); Fout[m].r = SUB32_ovflw(Fout->r, HALF_OF(scratch[3].r));
Fout[m].i = Fout->i - HALF_OF(scratch[3].i); Fout[m].i = SUB32_ovflw(Fout->i, HALF_OF(scratch[3].i));
C_MULBYSCALAR( scratch[0] , epi3.i ); C_MULBYSCALAR( scratch[0] , epi3.i );
C_ADDTO(*Fout,scratch[3]); C_ADDTO(*Fout,scratch[3]);
Fout[m2].r = Fout[m].r + scratch[0].i; Fout[m2].r = ADD32_ovflw(Fout[m].r, scratch[0].i);
Fout[m2].i = Fout[m].i - scratch[0].r; Fout[m2].i = SUB32_ovflw(Fout[m].i, scratch[0].r);
Fout[m].r -= scratch[0].i; Fout[m].r = SUB32_ovflw(Fout[m].r, scratch[0].i);
Fout[m].i += scratch[0].r; Fout[m].i = ADD32_ovflw(Fout[m].i, scratch[0].r);
++Fout; ++Fout;
} while(--k); } while(--k);
@ -282,22 +282,22 @@ static void kf_bfly5(
C_ADD( scratch[8],scratch[2],scratch[3]); C_ADD( scratch[8],scratch[2],scratch[3]);
C_SUB( scratch[9],scratch[2],scratch[3]); C_SUB( scratch[9],scratch[2],scratch[3]);
Fout0->r += scratch[7].r + scratch[8].r; Fout0->r = ADD32_ovflw(Fout0->r, ADD32_ovflw(scratch[7].r, scratch[8].r));
Fout0->i += scratch[7].i + scratch[8].i; Fout0->i = ADD32_ovflw(Fout0->i, ADD32_ovflw(scratch[7].i, scratch[8].i));
scratch[5].r = scratch[0].r + S_MUL(scratch[7].r,ya.r) + S_MUL(scratch[8].r,yb.r); scratch[5].r = ADD32_ovflw(scratch[0].r, ADD32_ovflw(S_MUL(scratch[7].r,ya.r), S_MUL(scratch[8].r,yb.r)));
scratch[5].i = scratch[0].i + S_MUL(scratch[7].i,ya.r) + S_MUL(scratch[8].i,yb.r); scratch[5].i = ADD32_ovflw(scratch[0].i, ADD32_ovflw(S_MUL(scratch[7].i,ya.r), S_MUL(scratch[8].i,yb.r)));
scratch[6].r = S_MUL(scratch[10].i,ya.i) + S_MUL(scratch[9].i,yb.i); scratch[6].r = ADD32_ovflw(S_MUL(scratch[10].i,ya.i), S_MUL(scratch[9].i,yb.i));
scratch[6].i = -S_MUL(scratch[10].r,ya.i) - S_MUL(scratch[9].r,yb.i); scratch[6].i = NEG32_ovflw(ADD32_ovflw(S_MUL(scratch[10].r,ya.i), S_MUL(scratch[9].r,yb.i)));
C_SUB(*Fout1,scratch[5],scratch[6]); C_SUB(*Fout1,scratch[5],scratch[6]);
C_ADD(*Fout4,scratch[5],scratch[6]); C_ADD(*Fout4,scratch[5],scratch[6]);
scratch[11].r = scratch[0].r + S_MUL(scratch[7].r,yb.r) + S_MUL(scratch[8].r,ya.r); scratch[11].r = ADD32_ovflw(scratch[0].r, ADD32_ovflw(S_MUL(scratch[7].r,yb.r), S_MUL(scratch[8].r,ya.r)));
scratch[11].i = scratch[0].i + S_MUL(scratch[7].i,yb.r) + S_MUL(scratch[8].i,ya.r); scratch[11].i = ADD32_ovflw(scratch[0].i, ADD32_ovflw(S_MUL(scratch[7].i,yb.r), S_MUL(scratch[8].i,ya.r)));
scratch[12].r = - S_MUL(scratch[10].i,yb.i) + S_MUL(scratch[9].i,ya.i); scratch[12].r = SUB32_ovflw(S_MUL(scratch[9].i,ya.i), S_MUL(scratch[10].i,yb.i));
scratch[12].i = S_MUL(scratch[10].r,yb.i) - S_MUL(scratch[9].r,ya.i); scratch[12].i = SUB32_ovflw(S_MUL(scratch[10].r,yb.i), S_MUL(scratch[9].r,ya.i));
C_ADD(*Fout2,scratch[11],scratch[12]); C_ADD(*Fout2,scratch[11],scratch[12]);
C_SUB(*Fout3,scratch[11],scratch[12]); C_SUB(*Fout3,scratch[11],scratch[12]);

View File

@ -38,7 +38,8 @@
#include "mathops.h" #include "mathops.h"
/*Compute floor(sqrt(_val)) with exact arithmetic. /*Compute floor(sqrt(_val)) with exact arithmetic.
This has been tested on all possible 32-bit inputs.*/ _val must be greater than 0.
This has been tested on all possible 32-bit inputs greater than 0.*/
unsigned isqrt32(opus_uint32 _val){ unsigned isqrt32(opus_uint32 _val){
unsigned b; unsigned b;
unsigned g; unsigned g;
@ -164,7 +165,7 @@ opus_val16 celt_cos_norm(opus_val32 x)
{ {
return _celt_cos_pi_2(EXTRACT16(x)); return _celt_cos_pi_2(EXTRACT16(x));
} else { } else {
return NEG32(_celt_cos_pi_2(EXTRACT16(65536-x))); return NEG16(_celt_cos_pi_2(EXTRACT16(65536-x)));
} }
} else { } else {
if (x&0x0000ffff) if (x&0x0000ffff)
@ -182,7 +183,7 @@ opus_val32 celt_rcp(opus_val32 x)
int i; int i;
opus_val16 n; opus_val16 n;
opus_val16 r; opus_val16 r;
celt_assert2(x>0, "celt_rcp() only defined for positive values"); celt_sig_assert(x>0);
i = celt_ilog2(x); i = celt_ilog2(x);
/* n is Q15 with range [0,1). */ /* n is Q15 with range [0,1). */
n = VSHR32(x,i-15)-32768; n = VSHR32(x,i-15)-32768;

View File

@ -38,11 +38,44 @@
#include "entcode.h" #include "entcode.h"
#include "os_support.h" #include "os_support.h"
#define PI 3.141592653f
/* Multiplies two 16-bit fractional values. Bit-exactness of this macro is important */ /* Multiplies two 16-bit fractional values. Bit-exactness of this macro is important */
#define FRAC_MUL16(a,b) ((16384+((opus_int32)(opus_int16)(a)*(opus_int16)(b)))>>15) #define FRAC_MUL16(a,b) ((16384+((opus_int32)(opus_int16)(a)*(opus_int16)(b)))>>15)
unsigned isqrt32(opus_uint32 _val); unsigned isqrt32(opus_uint32 _val);
/* CELT doesn't need it for fixed-point, by analysis.c does. */
#if !defined(FIXED_POINT) || defined(ANALYSIS_C)
#define cA 0.43157974f
#define cB 0.67848403f
#define cC 0.08595542f
#define cE ((float)PI/2)
static OPUS_INLINE float fast_atan2f(float y, float x) {
float x2, y2;
x2 = x*x;
y2 = y*y;
/* For very small values, we don't care about the answer, so
we can just return 0. */
if (x2 + y2 < 1e-18f)
{
return 0;
}
if(x2<y2){
float den = (y2 + cB*x2) * (y2 + cC*x2);
return -x*y*(y2 + cA*x2) / den + (y<0 ? -cE : cE);
}else{
float den = (x2 + cB*y2) * (x2 + cC*y2);
return x*y*(x2 + cA*y2) / den + (y<0 ? -cE : cE) - (x*y<0 ? -cE : cE);
}
}
#undef cA
#undef cB
#undef cC
#undef cE
#endif
#ifndef OVERRIDE_CELT_MAXABS16 #ifndef OVERRIDE_CELT_MAXABS16
static OPUS_INLINE opus_val32 celt_maxabs16(const opus_val16 *x, int len) static OPUS_INLINE opus_val32 celt_maxabs16(const opus_val16 *x, int len)
{ {
@ -80,7 +113,6 @@ static OPUS_INLINE opus_val32 celt_maxabs32(const opus_val32 *x, int len)
#ifndef FIXED_POINT #ifndef FIXED_POINT
#define PI 3.141592653f
#define celt_sqrt(x) ((float)sqrt(x)) #define celt_sqrt(x) ((float)sqrt(x))
#define celt_rsqrt(x) (1.f/celt_sqrt(x)) #define celt_rsqrt(x) (1.f/celt_sqrt(x))
#define celt_rsqrt_norm(x) (celt_rsqrt(x)) #define celt_rsqrt_norm(x) (celt_rsqrt(x))
@ -147,7 +179,7 @@ static OPUS_INLINE float celt_exp2(float x)
/** Integer log in base2. Undefined for zero and negative numbers */ /** Integer log in base2. Undefined for zero and negative numbers */
static OPUS_INLINE opus_int16 celt_ilog2(opus_int32 x) static OPUS_INLINE opus_int16 celt_ilog2(opus_int32 x)
{ {
celt_assert2(x>0, "celt_ilog2() only defined for strictly positive numbers"); celt_sig_assert(x>0);
return EC_ILOG(x)-1; return EC_ILOG(x)-1;
} }
#endif #endif

View File

@ -270,8 +270,8 @@ void clt_mdct_backward_c(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_sca
int rev; int rev;
kiss_fft_scalar yr, yi; kiss_fft_scalar yr, yi;
rev = *bitrev++; rev = *bitrev++;
yr = S_MUL(*xp2, t[i]) + S_MUL(*xp1, t[N4+i]); yr = ADD32_ovflw(S_MUL(*xp2, t[i]), S_MUL(*xp1, t[N4+i]));
yi = S_MUL(*xp1, t[i]) - S_MUL(*xp2, t[N4+i]); yi = SUB32_ovflw(S_MUL(*xp1, t[i]), S_MUL(*xp2, t[N4+i]));
/* We swap real and imag because we use an FFT instead of an IFFT. */ /* We swap real and imag because we use an FFT instead of an IFFT. */
yp[2*rev+1] = yr; yp[2*rev+1] = yr;
yp[2*rev] = yi; yp[2*rev] = yi;
@ -301,8 +301,8 @@ void clt_mdct_backward_c(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_sca
t0 = t[i]; t0 = t[i];
t1 = t[N4+i]; t1 = t[N4+i];
/* We'd scale up by 2 here, but instead it's done when mixing the windows */ /* We'd scale up by 2 here, but instead it's done when mixing the windows */
yr = S_MUL(re,t0) + S_MUL(im,t1); yr = ADD32_ovflw(S_MUL(re,t0), S_MUL(im,t1));
yi = S_MUL(re,t1) - S_MUL(im,t0); yi = SUB32_ovflw(S_MUL(re,t1), S_MUL(im,t0));
/* We swap real and imag because we're using an FFT instead of an IFFT. */ /* We swap real and imag because we're using an FFT instead of an IFFT. */
re = yp1[1]; re = yp1[1];
im = yp1[0]; im = yp1[0];
@ -312,8 +312,8 @@ void clt_mdct_backward_c(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_sca
t0 = t[(N4-i-1)]; t0 = t[(N4-i-1)];
t1 = t[(N2-i-1)]; t1 = t[(N2-i-1)];
/* We'd scale up by 2 here, but instead it's done when mixing the windows */ /* We'd scale up by 2 here, but instead it's done when mixing the windows */
yr = S_MUL(re,t0) + S_MUL(im,t1); yr = ADD32_ovflw(S_MUL(re,t0), S_MUL(im,t1));
yi = S_MUL(re,t1) - S_MUL(im,t0); yi = SUB32_ovflw(S_MUL(re,t1), S_MUL(im,t0));
yp1[0] = yr; yp1[0] = yr;
yp0[1] = yi; yp0[1] = yi;
yp0 += 2; yp0 += 2;
@ -333,8 +333,8 @@ void clt_mdct_backward_c(const mdct_lookup *l, kiss_fft_scalar *in, kiss_fft_sca
kiss_fft_scalar x1, x2; kiss_fft_scalar x1, x2;
x1 = *xp1; x1 = *xp1;
x2 = *yp1; x2 = *yp1;
*yp1++ = MULT16_32_Q15(*wp2, x2) - MULT16_32_Q15(*wp1, x1); *yp1++ = SUB32_ovflw(MULT16_32_Q15(*wp2, x2), MULT16_32_Q15(*wp1, x1));
*xp1-- = MULT16_32_Q15(*wp1, x2) + MULT16_32_Q15(*wp2, x1); *xp1-- = ADD32_ovflw(MULT16_32_Q15(*wp1, x2), MULT16_32_Q15(*wp2, x1));
wp1++; wp1++;
wp2--; wp2--;
} }

View File

@ -53,6 +53,7 @@
#include "celt_lpc.h" #include "celt_lpc.h"
#include "vq.h" #include "vq.h"
#define OVERRIDE_COMB_FILTER_CONST
#define OVERRIDE_comb_filter #define OVERRIDE_comb_filter
void comb_filter(opus_val32 *y, opus_val32 *x, int T0, int T1, int N, void comb_filter(opus_val32 *y, opus_val32 *x, int T0, int T1, int N,
opus_val16 g0, opus_val16 g1, int tapset0, int tapset1, opus_val16 g0, opus_val16 g1, int tapset0, int tapset1,

Some files were not shown because too many files have changed in this diff Show More