shithub: opus

Info  •  Files  •  Log  •  Branches

Fixes a SILK bandwidth switching regression

The bug was triggered because f982b84d started using prefill for
SILK bandwidth changes, which reinitialized the encoder state and
prevented the variable lowpass from working properly. To fix the problem,
we preserve the sampling rate and variable low-pass when prefilling.

--- a/silk/control_audio_bandwidth.c

+++ b/silk/control_audio_bandwidth.c

@@ -39,9 +39,15 @@

 )

 {

     opus_int   fs_kHz;

+    opus_int   orig_kHz;

     opus_int32 fs_Hz;

-    fs_kHz = psEncC->fs_kHz;

+    orig_kHz = psEncC->fs_kHz;

+    /* Handle a bandwidth-switching reset where we need to be aware what the last sampling rate was. */

+    if( orig_kHz == 0 ) {

+        orig_kHz = psEncC->sLP.saved_fs_kHz;

+    }

+    fs_kHz = orig_kHz;

     fs_Hz = silk_SMULBB( fs_kHz, 1000 );

     if( fs_Hz == 0 ) {

         /* Encoder has just been initialized */

@@ -61,7 +67,7 @@

         }

         if( psEncC->allow_bandwidth_switch || encControl->opusCanSwitch ) {

             /* Check if we should switch down */

-            if( silk_SMULBB( psEncC->fs_kHz, 1000 ) > psEncC->desiredInternal_fs_Hz )

+            if( silk_SMULBB( orig_kHz, 1000 ) > psEncC->desiredInternal_fs_Hz )

             {

                 /* Switch down */

                 if( psEncC->sLP.mode == 0 ) {

@@ -76,7 +82,7 @@

                     psEncC->sLP.mode = 0;

                     /* Switch to a lower sample frequency */

-                    fs_kHz = psEncC->fs_kHz == 16 ? 12 : 8;

+                    fs_kHz = orig_kHz == 16 ? 12 : 8;

                 } else {

                    if( psEncC->sLP.transition_frame_no <= 0 ) {

                        encControl->switchReady = 1;

@@ -90,12 +96,12 @@

             }

             else

             /* Check if we should switch up */

-            if( silk_SMULBB( psEncC->fs_kHz, 1000 ) < psEncC->desiredInternal_fs_Hz )

+            if( silk_SMULBB( orig_kHz, 1000 ) < psEncC->desiredInternal_fs_Hz )

             {

                 /* Switch up */

                 if( encControl->opusCanSwitch ) {

                     /* Switch to a higher sample frequency */

-                    fs_kHz = psEncC->fs_kHz == 8 ? 12 : 16;

+                    fs_kHz = orig_kHz == 8 ? 12 : 16;

                     /* New transition */

                     psEncC->sLP.transition_frame_no = 0;

--- a/silk/enc_API.c

+++ b/silk/enc_API.c

@@ -200,6 +200,7 @@

     tot_blocks = ( nBlocksOf10ms > 1 ) ? nBlocksOf10ms >> 1 : 1;

     curr_block = 0;

     if( prefillFlag ) {

+        silk_LP_state save_LP;

         /* Only accept input length of 10 ms */

         if( nBlocksOf10ms != 1 ) {

             celt_assert( 0 );

@@ -206,9 +207,18 @@

             RESTORE_STACK;

             return SILK_ENC_INPUT_INVALID_NO_OF_SAMPLES;

         }

+        if ( prefillFlag == 2 ) {

+            save_LP = psEnc->state_Fxx[ 0 ].sCmn.sLP;

+            /* Save the sampling rate so the bandwidth switching code can keep handling transitions. */

+            save_LP.saved_fs_kHz = psEnc->state_Fxx[ 0 ].sCmn.fs_kHz;

+        }

         /* Reset Encoder */

         for( n = 0; n < encControl->nChannelsInternal; n++ ) {

             ret = silk_init_encoder( &psEnc->state_Fxx[ n ], psEnc->state_Fxx[ n ].sCmn.arch );

+            /* Restore the variable LP state. */

+            if ( prefillFlag == 2 ) {

+                psEnc->state_Fxx[ n ].sCmn.sLP = save_LP;

+            }

             celt_assert( !ret );

         }

         tmp_payloadSize_ms = encControl->payloadSize_ms;

--- a/silk/structs.h

+++ b/silk/structs.h

@@ -78,6 +78,7 @@

     opus_int32                   In_LP_State[ 2 ];           /* Low pass filter state */

     opus_int32                   transition_frame_no;        /* Counter which is mapped to a cut-off frequency */

     opus_int                     mode;                       /* Operating mode, <0: switch down, >0: switch up; 0: do nothing           */

+    opus_int32                   saved_fs_kHz;               /* If non-zero, holds the last sampling rate before a bandwidth switching reset. */

 } silk_LP_state;

 /* Structure containing NLSF codebook */

--- a/src/opus_encoder.c

+++ b/src/opus_encoder.c

@@ -1599,7 +1599,8 @@

        redundancy = 1;

        celt_to_silk = 1;

        st->silk_bw_switch = 0;

-       prefill=1;

+       /* Do a prefill without reseting the sampling rate control. */

+       prefill=2;

     }

     /* If we decided to go with CELT, make sure redundancy is off, no matter what

@@ -1825,7 +1826,9 @@

             for (i=0;i<st->encoder_buffer*st->channels;i++)

                 pcm_silk[i] = FLOAT2INT16(st->delay_buffer[i]);

 #endif

-            silk_Encode( silk_enc, &st->silk_mode, pcm_silk, st->encoder_buffer, NULL, &zero, 1, activity );

+            silk_Encode( silk_enc, &st->silk_mode, pcm_silk, st->encoder_buffer, NULL, &zero, prefill, activity );

+            /* Prevent a second switch in the real encode call. */

+            st->silk_mode.opusCanSwitch = 0;

         }

 #ifdef FIXED_POINT