Reduce SILK encoder state size for mono

Author: jmvalin

silk/API.h

@@ -60,14 +60,16 @@ typedef struct {
 /* Get size in bytes of the Silk encoder state */
 /***********************************************/
 opus_int silk_Get_Encoder_Size(                         /* O    Returns error code                              */
-    opus_int                        *encSizeBytes       /* O    Number of bytes in SILK encoder state           */
+    opus_int                        *encSizeBytes,      /* O    Number of bytes in SILK encoder state           */
+    opus_int                         channels           /* I    Number of channels                              */
 );
 
 /*************************/
 /* Init or reset encoder */
 /*************************/
 opus_int silk_InitEncoder(                              /* O    Returns error code                              */
     void                            *encState,          /* I/O  State                                           */
+    int                              channels,          /* I    Number of channels                              */
     int                              arch,              /* I    Run-time architecture                           */
     silk_EncControlStruct           *encStatus          /* O    Encoder Status                                  */
 );

silk/enc_API.c

@@ -58,12 +58,17 @@ static opus_int silk_QueryEncoder(                      /* O    Returns error co
 /****************************************/
 
 opus_int silk_Get_Encoder_Size(                         /* O    Returns error code                              */
-    opus_int                        *encSizeBytes       /* O    Number of bytes in SILK encoder state           */
+    opus_int                        *encSizeBytes,      /* O    Number of bytes in SILK encoder state           */
+    opus_int                         channels           /* I    Number of channels                              */
 )
 {
     opus_int ret = SILK_NO_ERROR;
 
     *encSizeBytes = sizeof( silk_encoder );
+    /* Skip second encoder state for mono. */
+    if ( channels == 1 ) {
+        *encSizeBytes -= sizeof( silk_encoder_state_Fxx );
+    }
 
     return ret;
 }
@@ -73,6 +78,7 @@ opus_int silk_Get_Encoder_Size(                         /* O    Returns error co
 /*************************/
 opus_int silk_InitEncoder(                              /* O    Returns error code                              */
     void                            *encState,          /* I/O  State                                           */
+    int                              channels,          /* I    Number of channels                              */
     int                              arch,              /* I    Run-time architecture                           */
     silk_EncControlStruct           *encStatus          /* O    Encoder Status                                  */
 )
@@ -82,9 +88,9 @@ opus_int silk_InitEncoder(                              /* O    Returns error co
 
     psEnc = (silk_encoder *)encState;
 
-    /* Reset encoder */
-    silk_memset( psEnc, 0, sizeof( silk_encoder ) );
-    for( n = 0; n < ENCODER_NUM_CHANNELS; n++ ) {
+    /* Reset encoder. Skip second encoder state for mono. */
+    silk_memset( psEnc, 0, sizeof( silk_encoder ) - (channels==1)*sizeof( silk_encoder_state_Fxx ) );
+    for( n = 0; n < channels; n++ ) {
         if( ret += silk_init_encoder( &psEnc->state_Fxx[ n ], arch ) ) {
             celt_assert( 0 );
         }
@@ -162,13 +168,16 @@ opus_int silk_Encode(                                   /* O    Returns error co
     opus_int transition, curr_block, tot_blocks;
     SAVE_STACK;
 
+    celt_assert( encControl->nChannelsAPI >= encControl->nChannelsInternal && encControl->nChannelsAPI >= psEnc->nChannelsInternal );
     if (encControl->reducedDependency)
     {
-       psEnc->state_Fxx[0].sCmn.first_frame_after_reset = 1;
-       psEnc->state_Fxx[1].sCmn.first_frame_after_reset = 1;
+       for( n = 0; n < encControl->nChannelsAPI; n++ ) {
+           psEnc->state_Fxx[ n ].sCmn.first_frame_after_reset = 1;
+       }
+    }
+    for( n = 0; n < encControl->nChannelsAPI; n++ ) {
+        psEnc->state_Fxx[ n ].sCmn.nFramesEncoded = 0;
     }
-    psEnc->state_Fxx[ 0 ].sCmn.nFramesEncoded = psEnc->state_Fxx[ 1 ].sCmn.nFramesEncoded = 0;
-
     /* Check values in encoder control structure */
     if( ( ret = check_control_input( encControl ) ) != 0 ) {
         celt_assert( 0 );

silk/fixed/structs_FIX.h

@@ -96,7 +96,6 @@ typedef struct {
 /* Encoder Super Struct */
 /************************/
 typedef struct {
-    silk_encoder_state_FIX      state_Fxx[ ENCODER_NUM_CHANNELS ];
     stereo_enc_state            sStereo;
     opus_int32                  nBitsUsedLBRR;
     opus_int32                  nBitsExceeded;
@@ -106,6 +105,8 @@ typedef struct {
     opus_int                    timeSinceSwitchAllowed_ms;
     opus_int                    allowBandwidthSwitch;
     opus_int                    prev_decode_only_middle;
+    /* This needs to be last so we can skip the second state for mono. */
+    silk_encoder_state_FIX      state_Fxx[ ENCODER_NUM_CHANNELS ];
 } silk_encoder;
 
 

silk/float/structs_FLP.h

@@ -93,7 +93,6 @@ typedef struct {
 /* Encoder Super Struct */
 /************************/
 typedef struct {
-    silk_encoder_state_FLP      state_Fxx[ ENCODER_NUM_CHANNELS ];
     stereo_enc_state            sStereo;
     opus_int32                  nBitsUsedLBRR;
     opus_int32                  nBitsExceeded;
@@ -103,6 +102,8 @@ typedef struct {
     opus_int                    timeSinceSwitchAllowed_ms;
     opus_int                    allowBandwidthSwitch;
     opus_int                    prev_decode_only_middle;
+    /* This needs to be last so we can skip the second state for mono. */
+    silk_encoder_state_FLP      state_Fxx[ ENCODER_NUM_CHANNELS ];
 } silk_encoder;
 
 #ifdef __cplusplus

src/opus_encoder.c

@@ -198,7 +198,7 @@ int opus_encoder_get_size(int channels)
     int ret;
     if (channels<1 || channels > 2)
         return 0;
-    ret = silk_Get_Encoder_Size( &silkEncSizeBytes );
+    ret = silk_Get_Encoder_Size( &silkEncSizeBytes, channels );
     if (ret)
         return 0;
     silkEncSizeBytes = align(silkEncSizeBytes);
@@ -224,7 +224,7 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat
 
     OPUS_CLEAR((char*)st, opus_encoder_get_size(channels));
     /* Create SILK encoder */
-    ret = silk_Get_Encoder_Size( &silkEncSizeBytes );
+    ret = silk_Get_Encoder_Size( &silkEncSizeBytes, channels );
     if (ret)
         return OPUS_BAD_ARG;
     silkEncSizeBytes = align(silkEncSizeBytes);
@@ -239,7 +239,7 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat
 
     st->arch = opus_select_arch();
 
-    ret = silk_InitEncoder( silk_enc, st->arch, &st->silk_mode );
+    ret = silk_InitEncoder( silk_enc, st->channels, st->arch, &st->silk_mode );
     if(ret)return OPUS_INTERNAL_ERROR;
 
     /* default SILK parameters */
@@ -1520,7 +1520,7 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_res *pcm, int frame_si
     if (st->mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY)
     {
         silk_EncControlStruct dummy;
-        silk_InitEncoder( silk_enc, st->arch, &dummy);
+        silk_InitEncoder( silk_enc, st->channels, st->arch, &dummy);
         prefill=1;
     }
 
@@ -3151,7 +3151,7 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
            OPUS_CLEAR(start, sizeof(OpusEncoder) - (start - (char*)st));
 
            celt_encoder_ctl(celt_enc, OPUS_RESET_STATE);
-           silk_InitEncoder( silk_enc, st->arch, &dummy );
+           silk_InitEncoder( silk_enc, st->channels, st->arch, &dummy );
 #ifdef ENABLE_DRED
            /* Initialize DRED Encoder */
            dred_encoder_reset( &st->dred_encoder );