Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

RFC: Streamline implementation overrides #392

Draft
wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

RFC: Streamline implementation overrides #392

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
19 changes: 1 addition & 18 deletions silk/NSQ.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -73,24 +73,7 @@ static OPUS_INLINE void silk_noise_shape_quantizer(
);
#endif

void silk_NSQ_c
(
const silk_encoder_state *psEncC, /* I Encoder State */
silk_nsq_state *NSQ, /* I/O NSQ state */
SideInfoIndices *psIndices, /* I/O Quantization Indices */
const opus_int16 x16[], /* I Input */
opus_int8 pulses[], /* O Quantized pulse signal */
const opus_int16 *PredCoef_Q12, /* I Short term prediction coefs */
const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */
const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */
const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */
const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */
const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */
const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */
const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */
const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */
const opus_int LTP_scale_Q14 /* I LTP state scaling */
)
SILK_NSQ_DECL(c)
{
opus_int k, lag, start_idx, LSF_interpolation_flag;
const opus_int16 *A_Q12, *B_Q14, *AR_shp_Q13;
Expand Down
15 changes: 1 addition & 14 deletions silk/VQ_WMat_EC.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,20 +32,7 @@ POSSIBILITY OF SUCH DAMAGE.
#include "main.h"

/* Entropy constrained matrix-weighted VQ, hard-coded to 5-element vectors, for a single input data vector */
void silk_VQ_WMat_EC_c(
opus_int8 *ind, /* O index of best codebook vector */
opus_int32 *res_nrg_Q15, /* O best residual energy */
opus_int32 *rate_dist_Q8, /* O best total bitrate */
opus_int *gain_Q7, /* O sum of absolute LTP coefficients */
const opus_int32 *XX_Q17, /* I correlation matrix */
const opus_int32 *xX_Q17, /* I correlation vector */
const opus_int8 *cb_Q7, /* I codebook */
const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */
const opus_uint8 *cl_Q5, /* I code length for each codebook vector */
const opus_int subfr_len, /* I number of samples per subframe */
const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */
const opus_int L /* I number of vectors in codebook */
)
SILK_VQ_WMAT_EC_DECL(c)
{
opus_int k, gain_tmp_Q7;
const opus_int8 *cb_row_Q7;
Expand Down
9 changes: 4 additions & 5 deletions silk/fixed/encode_frame_FIX.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -206,10 +206,9 @@ opus_int silk_encode_frame_FIX(
sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14,
psEnc->sCmn.arch );
} else {
silk_NSQ( &psEnc->sCmn, &psEnc->sCmn.sNSQ, &psEnc->sCmn.indices, x_frame, psEnc->sCmn.pulses,
silk_NSQ( psEnc->sCmn.arch, &psEnc->sCmn, &psEnc->sCmn.sNSQ, &psEnc->sCmn.indices, x_frame, psEnc->sCmn.pulses,
sEncCtrl.PredCoef_Q12[ 0 ], sEncCtrl.LTPCoef_Q14, sEncCtrl.AR_Q13, sEncCtrl.HarmShapeGain_Q14,
sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14,
psEnc->sCmn.arch);
sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14 );
}

if ( iter == maxIter && !found_lower ) {
Expand Down Expand Up @@ -432,10 +431,10 @@ static OPUS_INLINE void silk_LBRR_encode_FIX(
psEncCtrl->AR_Q13, psEncCtrl->HarmShapeGain_Q14, psEncCtrl->Tilt_Q14, psEncCtrl->LF_shp_Q14,
psEncCtrl->Gains_Q16, psEncCtrl->pitchL, psEncCtrl->Lambda_Q10, psEncCtrl->LTP_scale_Q14, psEnc->sCmn.arch );
} else {
silk_NSQ( &psEnc->sCmn, &sNSQ_LBRR, psIndices_LBRR, x16,
silk_NSQ( psEnc->sCmn.arch, &psEnc->sCmn, &sNSQ_LBRR, psIndices_LBRR, x16,
psEnc->sCmn.pulses_LBRR[ psEnc->sCmn.nFramesEncoded ], psEncCtrl->PredCoef_Q12[ 0 ], psEncCtrl->LTPCoef_Q14,
psEncCtrl->AR_Q13, psEncCtrl->HarmShapeGain_Q14, psEncCtrl->Tilt_Q14, psEncCtrl->LF_shp_Q14,
psEncCtrl->Gains_Q16, psEncCtrl->pitchL, psEncCtrl->Lambda_Q10, psEncCtrl->LTP_scale_Q14, psEnc->sCmn.arch );
psEncCtrl->Gains_Q16, psEncCtrl->pitchL, psEncCtrl->Lambda_Q10, psEncCtrl->LTP_scale_Q14 );
}

/* Restore original gains */
Expand Down
4 changes: 2 additions & 2 deletions silk/float/wrappers_FLP.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -164,8 +164,8 @@ void silk_NSQ_wrapper_FLP(
silk_NSQ_del_dec( &psEnc->sCmn, psNSQ, psIndices, x16, pulses, PredCoef_Q12[ 0 ], LTPCoef_Q14,
AR_Q13, HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, psEncCtrl->pitchL, Lambda_Q10, LTP_scale_Q14, psEnc->sCmn.arch );
} else {
silk_NSQ( &psEnc->sCmn, psNSQ, psIndices, x16, pulses, PredCoef_Q12[ 0 ], LTPCoef_Q14,
AR_Q13, HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, psEncCtrl->pitchL, Lambda_Q10, LTP_scale_Q14, psEnc->sCmn.arch );
silk_NSQ( psEnc->sCmn.arch, &psEnc->sCmn, psNSQ, psIndices, x16, pulses, PredCoef_Q12[ 0 ], LTPCoef_Q14,
AR_Q13, HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, psEncCtrl->pitchL, Lambda_Q10, LTP_scale_Q14 );
}
}

Expand Down
48 changes: 7 additions & 41 deletions silk/main.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,7 @@ POSSIBILITY OF SUCH DAMAGE.
#include "debug.h"
#include "entenc.h"
#include "entdec.h"
#include "main_overrides.h"

#if defined(OPUS_X86_MAY_HAVE_SSE4_1)
#include "x86/main_sse.h"
Expand Down Expand Up @@ -222,53 +223,18 @@ void silk_quant_LTP_gains(
);

/* Entropy constrained matrix-weighted VQ, for a single input data vector */
void silk_VQ_WMat_EC_c(
opus_int8 *ind, /* O index of best codebook vector */
opus_int32 *res_nrg_Q15, /* O best residual energy */
opus_int32 *rate_dist_Q8, /* O best total bitrate */
opus_int *gain_Q7, /* O sum of absolute LTP coefficients */
const opus_int32 *XX_Q17, /* I correlation matrix */
const opus_int32 *xX_Q17, /* I correlation vector */
const opus_int8 *cb_Q7, /* I codebook */
const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */
const opus_uint8 *cl_Q5, /* I code length for each codebook vector */
const opus_int subfr_len, /* I number of samples per subframe */
const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */
const opus_int L /* I number of vectors in codebook */
);

#if !defined(OVERRIDE_silk_VQ_WMat_EC)
#define silk_VQ_WMat_EC(ind, res_nrg_Q15, rate_dist_Q8, gain_Q7, XX_Q17, xX_Q17, cb_Q7, cb_gain_Q7, cl_Q5, subfr_len, max_gain_Q7, L, arch) \
((void)(arch),silk_VQ_WMat_EC_c(ind, res_nrg_Q15, rate_dist_Q8, gain_Q7, XX_Q17, xX_Q17, cb_Q7, cb_gain_Q7, cl_Q5, subfr_len, max_gain_Q7, L))
SILK_VQ_WMAT_EC_DECL(c);
#ifndef silk_VQ_WMat_EC
#define silk_VQ_WMat_EC(...) OVERRIDE_IMPL_SINGLE(silk_VQ_WMat_EC, c, __VA_ARGS__)
#endif

/************************************/
/* Noise shaping quantization (NSQ) */
/************************************/

void silk_NSQ_c(
const silk_encoder_state *psEncC, /* I Encoder State */
silk_nsq_state *NSQ, /* I/O NSQ state */
SideInfoIndices *psIndices, /* I/O Quantization Indices */
const opus_int16 x16[], /* I Input */
opus_int8 pulses[], /* O Quantized pulse signal */
const opus_int16 *PredCoef_Q12, /* I Short term prediction coefs */
const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */
const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */
const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */
const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */
const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */
const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */
const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */
const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */
const opus_int LTP_scale_Q14 /* I LTP state scaling */
);

#if !defined(OVERRIDE_silk_NSQ)
#define silk_NSQ(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \
HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14, arch) \
((void)(arch),silk_NSQ_c(psEncC, NSQ, psIndices, x16, pulses, PredCoef_Q12, LTPCoef_Q14, AR_Q13, \
HarmShapeGain_Q14, Tilt_Q14, LF_shp_Q14, Gains_Q16, pitchL, Lambda_Q10, LTP_scale_Q14))
SILK_NSQ_DECL(c);
#ifndef silk_NSQ
#define silk_NSQ(...) OVERRIDE_IMPL_SINGLE(silk_NSQ, c, __VA_ARGS__)
#endif

/* Noise shaping using delayed decision */
Expand Down
88 changes: 88 additions & 0 deletions silk/main_overrides.h
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,88 @@
/* Copyright (c) 2025, Samsung R&D Institute Poland
Written by Marek Pikula

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.
*/

#ifndef SILK_MAIN_OVERRIDES_H
#define SILK_MAIN_OVERRIDES_H

/** Override implementation with a single function. */
#define OVERRIDE_IMPL_SINGLE(func, impl, arch, ...) \
((void)(arch), func ## _ ## impl(__VA_ARGS__))

/** Override implementation with a multiple arch-specific functions.
*
* It uses an array declared with OVERRIDE_IMPL_ARRAY_DECL().
*/
#define OVERRIDE_IMPL_ARRAY(func, arch, ...) \
((*OVERRIDE_MAP_ ## func[(arch) & OPUS_ARCHMASK])(__VA_ARGS__))

/** Declare a mapping array for use with OVERRIDE_IMPL_ARRAY(). */
#define OVERRIDE_IMPL_ARRAY_DECL(func) \
const func ## _t OVERRIDE_MAP_ ## func[ OPUS_ARCHMASK + 1 ]


// Function declarations.

/* Entropy constrained matrix-weighted VQ, for a single input data vector */
#define SILK_VQ_WMAT_EC_DECL(impl, ...) \
void (__VA_ARGS__ silk_VQ_WMat_EC_ ## impl)( \
opus_int8 *ind, /* O index of best codebook vector */ \
opus_int32 *res_nrg_Q15, /* O best residual energy */ \
opus_int32 *rate_dist_Q8, /* O best total bitrate */ \
opus_int *gain_Q7, /* O sum of absolute LTP coefficients */ \
const opus_int32 *XX_Q17, /* I correlation matrix */ \
const opus_int32 *xX_Q17, /* I correlation vector */ \
const opus_int8 *cb_Q7, /* I codebook */ \
const opus_uint8 *cb_gain_Q7, /* I codebook effective gain */ \
const opus_uint8 *cl_Q5, /* I code length for each codebook vector */ \
const opus_int subfr_len, /* I number of samples per subframe */ \
const opus_int32 max_gain_Q7, /* I maximum sum of absolute LTP coefficients */ \
const opus_int L /* I number of vectors in codebook */ \
)
typedef SILK_VQ_WMAT_EC_DECL(t, *const);

/* Noise shaping quantization (NSQ) */
#define SILK_NSQ_DECL(impl, ...) \
void (__VA_ARGS__ silk_NSQ_ ## impl)( \
const silk_encoder_state *psEncC, /* I Encoder State */ \
silk_nsq_state *NSQ, /* I/O NSQ state */ \
SideInfoIndices *psIndices, /* I/O Quantization Indices */ \
const opus_int16 x16[], /* I Input */ \
opus_int8 pulses[], /* O Quantized pulse signal */ \
const opus_int16 *PredCoef_Q12, /* I Short term prediction coefs */ \
const opus_int16 LTPCoef_Q14[ LTP_ORDER * MAX_NB_SUBFR ], /* I Long term prediction coefs */ \
const opus_int16 AR_Q13[ MAX_NB_SUBFR * MAX_SHAPE_LPC_ORDER ], /* I Noise shaping coefs */ \
const opus_int HarmShapeGain_Q14[ MAX_NB_SUBFR ], /* I Long term shaping coefs */ \
const opus_int Tilt_Q14[ MAX_NB_SUBFR ], /* I Spectral tilt */ \
const opus_int32 LF_shp_Q14[ MAX_NB_SUBFR ], /* I Low frequency shaping coefs */ \
const opus_int32 Gains_Q16[ MAX_NB_SUBFR ], /* I Quantization step sizes */ \
const opus_int pitchL[ MAX_NB_SUBFR ], /* I Pitch lags */ \
const opus_int Lambda_Q10, /* I Rate/distortion tradeoff */ \
const opus_int LTP_scale_Q14 /* I LTP state scaling */ \
)
typedef SILK_NSQ_DECL(t, *const);

#endif // SILK_MAIN_OVERRIDES_H
4 changes: 2 additions & 2 deletions silk/quant_LTP_gains.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -82,6 +82,7 @@ void silk_quant_LTP_gains(
max_gain_Q7 = silk_log2lin( ( SILK_FIX_CONST( MAX_SUM_LOG_GAIN_DB / 6.0, 7 ) - sum_log_gain_tmp_Q7 )
+ SILK_FIX_CONST( 7, 7 ) ) - gain_safety;
silk_VQ_WMat_EC(
arch, /* I Run-time architecture */
&temp_idx[ j ], /* O index of best codebook vector */
&res_nrg_Q15_subfr, /* O residual energy */
&rate_dist_Q7_subfr, /* O best weighted quantization error + mu * rate */
Expand All @@ -93,8 +94,7 @@ void silk_quant_LTP_gains(
cl_ptr_Q5, /* I code length for each codebook vector */
subfr_len, /* I number of samples per subframe */
max_gain_Q7, /* I maximum sum of absolute LTP coefficients */
cbk_size, /* I number of vectors in codebook */
arch /* I Run-time architecture */
cbk_size /* I number of vectors in codebook */
);

res_nrg_Q15 = silk_ADD_POS_SAT32( res_nrg_Q15, res_nrg_Q15_subfr );
Expand Down
Loading