Implements hard CBR for SILK

This is achieved by running the encoding process in a loop and
padding when we don't reach the exact rate. It also implements
VBR-with-cap, which means we no longer need to artificially decrease
the SILK bandwidth when it's close to the cap.

6 files changed, 147 insertions, 33 deletions

diff --git a/silk/fixed/LTP_scale_ctrl_FIX.c b/silk/fixed/LTP_scale_ctrl_FIX.c
index 5e85cc74..dd4923f8 100644
--- a/silk/fixed/LTP_scale_ctrl_FIX.c
+++ b/silk/fixed/LTP_scale_ctrl_FIX.c
@@ -35,7 +35,6 @@ void silk_LTP_scale_ctrl_FIX(
     silk_encoder_state_FIX      *psEnc,     /* I/O  encoder state FIX                           */
     silk_encoder_control_FIX    *psEncCtrl, /* I/O  encoder control FIX                         */
     opus_int                     condCoding /* I    The type of conditional coding to use       */
-
 )
 {
     opus_int round_loss;
diff --git a/silk/fixed/encode_frame_FIX.c b/silk/fixed/encode_frame_FIX.c
index e351def9..a8633f5a 100644
--- a/silk/fixed/encode_frame_FIX.c
+++ b/silk/fixed/encode_frame_FIX.c
@@ -79,14 +79,23 @@ opus_int silk_encode_frame_FIX(
     silk_encoder_state_FIX          *psEnc,             /* I/O  Encoder state FIX                       */
     opus_int32                       *pnBytesOut,        /*   O  Number of payload bytes                 */
     ec_enc                          *psRangeEnc,        /* I/O  compressor data structure               */
-    opus_int                         condCoding         /* I    The type of conditional coding to use   */
+    opus_int                         condCoding,        /* I    The type of conditional coding to use   */
+    opus_int                         maxBits,           /* I    If > 0: maximum number of output bits   */
+    opus_int                         useCBR             /* I    Flag to force constant-bitrate operation */
 )
 {
     silk_encoder_control_FIX sEncCtrl;
-    opus_int     ret = 0;
+    opus_int     i, iter, maxIter, found_upper, found_lower, ret = 0;
     opus_int16   *x_frame, *res_pitch_frame;
     opus_int16   xfw[ MAX_FRAME_LENGTH ];
     opus_int16   res_pitch[ 2 * MAX_FRAME_LENGTH + LA_PITCH_MAX ];
+    ec_enc       sRangeEnc_copy, sRangeEnc_copy2;
+    silk_nsq_state sNSQ_copy, sNSQ_copy2;
+    opus_int32   seed_copy, nBits, nBits_lower, nBits_upper, gainMult_Q10, gainMult_lower, gainMult_upper;
+    opus_int16   ec_prevLagIndex_copy;
+    opus_int     ec_prevSignalType_copy;
+    opus_int8    LastGainIndex_copy2;
+    opus_uint8   ec_buf_copy[ 512 ];
 
 TIC(ENCODE_FRAME)
 
@@ -151,21 +160,132 @@ TIC(LBRR)
     silk_LBRR_encode_FIX( psEnc, &sEncCtrl, xfw, condCoding );
 TOC(LBRR)
 
-    /*****************************************/
-    /* Noise shaping quantization            */
-    /*****************************************/
+    /* Loop over quantizer and entroy coding to control bitrate */
+    maxIter = 5;
+    gainMult_Q10 = SILK_FIX_CONST( 1, 10 );
+    found_lower = 0;
+    found_upper = 0;
+    for( iter = 0; ; iter++ ) {
+        if( maxBits > 0 && !psEnc->sCmn.prefillFlag ) {
+            /* Copy part of the input state */
+            silk_memcpy( &sRangeEnc_copy, psRangeEnc, sizeof( ec_enc ) );
+            silk_memcpy( &sNSQ_copy, &psEnc->sCmn.sNSQ, sizeof( silk_nsq_state ) );
+            seed_copy = psEnc->sCmn.indices.Seed;
+            ec_prevLagIndex_copy = psEnc->sCmn.ec_prevLagIndex;
+            ec_prevSignalType_copy = psEnc->sCmn.ec_prevSignalType;
+        }
+
+        /*****************************************/
+        /* Noise shaping quantization            */
+        /*****************************************/
 TIC(NSQ)
-    if( psEnc->sCmn.nStatesDelayedDecision > 1 || psEnc->sCmn.warping_Q16 > 0 ) {
-        silk_NSQ_del_dec( &psEnc->sCmn, &psEnc->sCmn.sNSQ, &psEnc->sCmn.indices, xfw, psEnc->sCmn.pulses,
-            sEncCtrl.PredCoef_Q12[ 0 ], sEncCtrl.LTPCoef_Q14, sEncCtrl.AR2_Q13, sEncCtrl.HarmShapeGain_Q14,
-            sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14 );
-    } else {
-        silk_NSQ( &psEnc->sCmn, &psEnc->sCmn.sNSQ, &psEnc->sCmn.indices, xfw, psEnc->sCmn.pulses,
-            sEncCtrl.PredCoef_Q12[ 0 ], sEncCtrl.LTPCoef_Q14, sEncCtrl.AR2_Q13, sEncCtrl.HarmShapeGain_Q14,
-            sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14 );
-    }
+        if( psEnc->sCmn.nStatesDelayedDecision > 1 || psEnc->sCmn.warping_Q16 > 0 ) {
+            silk_NSQ_del_dec( &psEnc->sCmn, &psEnc->sCmn.sNSQ, &psEnc->sCmn.indices, xfw, psEnc->sCmn.pulses,
+                sEncCtrl.PredCoef_Q12[ 0 ], sEncCtrl.LTPCoef_Q14, sEncCtrl.AR2_Q13, sEncCtrl.HarmShapeGain_Q14,
+                sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14 );
+        } else {
+            silk_NSQ( &psEnc->sCmn, &psEnc->sCmn.sNSQ, &psEnc->sCmn.indices, xfw, psEnc->sCmn.pulses,
+                sEncCtrl.PredCoef_Q12[ 0 ], sEncCtrl.LTPCoef_Q14, sEncCtrl.AR2_Q13, sEncCtrl.HarmShapeGain_Q14,
+                sEncCtrl.Tilt_Q14, sEncCtrl.LF_shp_Q14, sEncCtrl.Gains_Q16, sEncCtrl.pitchL, sEncCtrl.Lambda_Q10, sEncCtrl.LTP_scale_Q14 );
+        }
 TOC(NSQ)
 
+        if( psEnc->sCmn.prefillFlag ) {
+            break;
+        } else {
+            /****************************************/
+            /* Encode Parameters                    */
+            /****************************************/
+TIC(ENCODE_PARAMS)
+            silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0, condCoding );
+TOC(ENCODE_PARAMS)
+
+            /****************************************/
+            /* Encode Excitation Signal             */
+            /****************************************/
+TIC(ENCODE_PULSES)
+            silk_encode_pulses( psRangeEnc, psEnc->sCmn.indices.signalType, psEnc->sCmn.indices.quantOffsetType,
+                psEnc->sCmn.pulses, psEnc->sCmn.frame_length );
+TOC(ENCODE_PULSES)
+
+        }
+
+        nBits = ec_tell( psRangeEnc );
+
+        if( maxBits == 0 || ( useCBR == 0 && iter == 0 && nBits <= maxBits ) ) {
+            break;
+        }
+
+        if( iter == maxIter ) {
+            if( nBits > maxBits && found_lower ) {
+                /* Restore output state from earlier iteration that did meet the bitrate budget */
+                silk_memcpy( psRangeEnc, &sRangeEnc_copy2, sizeof( ec_enc ) );
+                silk_memcpy( psRangeEnc->buf, ec_buf_copy, sRangeEnc_copy2.offs );
+                silk_memcpy( &psEnc->sCmn.sNSQ, &sNSQ_copy2, sizeof( silk_nsq_state ) );
+                psEnc->sShape.LastGainIndex = LastGainIndex_copy2;
+            }
+            break;
+        }
+
+        if( nBits > maxBits ) {
+            found_upper = 1;
+            nBits_upper = nBits;
+            gainMult_upper = gainMult_Q10;
+            if( found_lower == 0 && iter >= 2 ) {
+                /* Adjust the quantizer's rate/distortion tradeoff */
+                sEncCtrl.Lambda_Q10 = silk_ADD_RSHIFT32( sEncCtrl.Lambda_Q10, sEncCtrl.Lambda_Q10, 1 );
+            }
+        } else if( nBits < maxBits - 5 ) {
+            found_lower = 1;
+            nBits_lower = nBits;
+            gainMult_lower = gainMult_Q10;
+            /* Copy part of the output state */
+            silk_memcpy( &sRangeEnc_copy2, psRangeEnc, sizeof( ec_enc ) );
+            silk_memcpy( ec_buf_copy, psRangeEnc->buf, psRangeEnc->offs );
+            silk_memcpy( &sNSQ_copy2, &psEnc->sCmn.sNSQ, sizeof( silk_nsq_state ) );
+            LastGainIndex_copy2 = psEnc->sShape.LastGainIndex;
+        } else {
+            /* Within 5 bits of budget: close enough */
+            break;
+        }
+
+        if( ( found_lower & found_upper ) == 0 ) {
+            /* Adjust gain according to high-rate rate/distortion curve */
+            opus_int32 gain_factor_Q16;

+            gain_factor_Q16 = silk_log2lin( silk_LSHIFT( nBits - maxBits, 7 ) / psEnc->sCmn.frame_length + SILK_FIX_CONST( 16, 7 ) );

+            if( nBits > maxBits ) {
+                gain_factor_Q16 = silk_max_32( gain_factor_Q16, SILK_FIX_CONST( 1.3, 16 ) );
+            }
+            gainMult_Q10 = silk_SMULWW( gainMult_Q10, gain_factor_Q16 );

+        } else {
+            /* Adjust gain by interpolating */
+            gainMult_Q10 = gainMult_lower + silk_DIV32_16( silk_MUL( gainMult_upper - gainMult_lower, maxBits - nBits_lower ), nBits_upper - nBits_lower );
+            /* New gain multplier must be between 25% and 75% of old range (note that gainMult_upper < gainMult_lower) */
+            if( gainMult_Q10 > gainMult_lower + silk_RSHIFT32( gainMult_upper - gainMult_lower, 2 ) ) {
+                gainMult_Q10 = gainMult_lower + silk_RSHIFT32( gainMult_upper - gainMult_lower, 2 );
+            } else 
+            if( gainMult_Q10 < gainMult_upper - silk_RSHIFT32( gainMult_upper - gainMult_lower, 2 ) ) {
+                gainMult_Q10 = gainMult_upper - silk_RSHIFT32( gainMult_upper - gainMult_lower, 2 );
+            } 
+        }
+
+        for( i = 0; i < psEnc->sCmn.nb_subfr; i++ ) {
+            sEncCtrl.Gains_Q16[ i ] = silk_LSHIFT_SAT32( silk_SMULWW( sEncCtrl.GainsUnq_Q16[ i ], gainMult_Q10 ), 6 );
+        }
+        psEnc->sShape.LastGainIndex = sEncCtrl.lastGainIndexPrev;
+
+        /* Noise shaping quantization */
+        silk_gains_quant( psEnc->sCmn.indices.GainsIndices, sEncCtrl.Gains_Q16,
+                &psEnc->sShape.LastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr );
+
+        /* Restore part of the input state */
+        silk_memcpy( psRangeEnc, &sRangeEnc_copy, sizeof( ec_enc ) );
+        silk_memcpy( &psEnc->sCmn.sNSQ, &sNSQ_copy, sizeof( silk_nsq_state ) );
+        psEnc->sCmn.indices.Seed = seed_copy;
+        psEnc->sCmn.ec_prevLagIndex = ec_prevLagIndex_copy;
+        psEnc->sCmn.ec_prevSignalType = ec_prevSignalType_copy;
+    }
+
     /* Update input buffer */
     silk_memmove( psEnc->x_buf, &psEnc->x_buf[ psEnc->sCmn.frame_length ],
         ( psEnc->sCmn.ltp_mem_length + LA_SHAPE_MS * psEnc->sCmn.fs_kHz ) * sizeof( opus_int16 ) );
@@ -182,27 +302,13 @@ TOC(NSQ)
     }
 
     /****************************************/
-    /* Encode Parameters                    */
-    /****************************************/
-TIC(ENCODE_PARAMS)
-    silk_encode_indices( &psEnc->sCmn, psRangeEnc, psEnc->sCmn.nFramesEncoded, 0, condCoding );
-TOC(ENCODE_PARAMS)
-
-    /****************************************/
-    /* Encode Excitation Signal             */
-    /****************************************/
-TIC(ENCODE_PULSES)
-    silk_encode_pulses( psRangeEnc, psEnc->sCmn.indices.signalType, psEnc->sCmn.indices.quantOffsetType,
-        psEnc->sCmn.pulses, psEnc->sCmn.frame_length );
-TOC(ENCODE_PULSES)
-
-    /****************************************/
     /* Finalize payload                     */
     /****************************************/
     psEnc->sCmn.first_frame_after_reset = 0;
     /* Payload size */
     *pnBytesOut = silk_RSHIFT( ec_tell( psRangeEnc ) + 7, 3 );
-TOC(ENCODE_FRAME)
+
+    TOC(ENCODE_FRAME)
 
 #ifdef SAVE_ALL_INTERNAL_DATA
     {
diff --git a/silk/fixed/find_pred_coefs_FIX.c b/silk/fixed/find_pred_coefs_FIX.c
index e9891cd4..2cefd905 100644
--- a/silk/fixed/find_pred_coefs_FIX.c
+++ b/silk/fixed/find_pred_coefs_FIX.c
@@ -127,6 +127,6 @@ void silk_find_pred_coefs_FIX(
     silk_residual_energy_FIX( psEncCtrl->ResNrg, psEncCtrl->ResNrgQ, LPC_in_pre, psEncCtrl->PredCoef_Q12, local_gains,
         psEnc->sCmn.subfr_length, psEnc->sCmn.nb_subfr, psEnc->sCmn.predictLPCOrder );
 
-    /* Copy to prediction struct for use in next frame for fluctuation reduction */
+    /* Copy to prediction struct for use in next frame for interpolation */
     silk_memcpy( psEnc->sCmn.prev_NLSFq_Q15, NLSF_Q15, sizeof( psEnc->sCmn.prev_NLSFq_Q15 ) );
 }
diff --git a/silk/fixed/main_FIX.h b/silk/fixed/main_FIX.h
index 57b648eb..f4b0270c 100644
--- a/silk/fixed/main_FIX.h
+++ b/silk/fixed/main_FIX.h
@@ -66,7 +66,9 @@ opus_int silk_encode_frame_FIX(
     silk_encoder_state_FIX          *psEnc,             /* I/O  Pointer to Silk FIX encoder state       */
     opus_int32                       *pnBytesOut,        /*   O  Pointer to number of payload bytes;     */
     ec_enc                          *psRangeEnc,        /* I/O  compressor data structure               */
-    opus_int                         condCoding         /* I    The type of conditional coding to use   */
+    opus_int                         condCoding,        /* I    The type of conditional coding to use   */
+    opus_int                         maxBits,           /* I    If > 0: maximum number of output bits   */
+    opus_int                         useCBR             /* I    Flag to force constant-bitrate operation */
 );
 
 /* Initializes the Silk encoder state */
diff --git a/silk/fixed/process_gains_FIX.c b/silk/fixed/process_gains_FIX.c
index 25a13a57..c3d5664d 100644
--- a/silk/fixed/process_gains_FIX.c
+++ b/silk/fixed/process_gains_FIX.c
@@ -86,6 +86,10 @@ void silk_process_gains_FIX(
         }
     }
 
+    /* Save unquantized gains and gain Index */
+    silk_memcpy( psEncCtrl->GainsUnq_Q16, psEncCtrl->Gains_Q16, psEnc->sCmn.nb_subfr * sizeof( opus_int32 ) );
+    psEncCtrl->lastGainIndexPrev = psShapeSt->LastGainIndex;
+
     /* Noise shaping quantization */
     silk_gains_quant( psEnc->sCmn.indices.GainsIndices, psEncCtrl->Gains_Q16,
         &psShapeSt->LastGainIndex, condCoding == CODE_CONDITIONALLY, psEnc->sCmn.nb_subfr );
diff --git a/silk/fixed/structs_FIX.h b/silk/fixed/structs_FIX.h
index 47e845d1..8803c7f2 100644
--- a/silk/fixed/structs_FIX.h
+++ b/silk/fixed/structs_FIX.h
@@ -109,6 +109,9 @@ typedef struct {
     opus_int32   ResNrg[ MAX_NB_SUBFR ];             /* Residual energy per subframe                             */
     opus_int     ResNrgQ[ MAX_NB_SUBFR ];            /* Q domain for the residual energy > 0                     */
 
+    /* Parameters for CBR mode */
+    opus_int32                    GainsUnq_Q16[ MAX_NB_SUBFR ];
+    opus_int8                     lastGainIndexPrev;
 } silk_encoder_control_FIX;
 
 /************************/