Sync to upstream libopus

Sync to commit bb4b6885a139644cf3ac14e7deda9f633ec2d93c

This brings in a bunch of optimizations to decode speed
and memory usage. Allocations are switched from using
the pseudostack to using the real stack. Enabled hacks
to reduce stack usage.

This should fix crashes on sansa clip, although some
files will not play due to failing allocations in the
codec buffer.

Speeds up decoding of the following test files:

                 H300 (cf)   C200 (arm7tdmi)  ipod classic (arm9e)
16 kbps (silk)   14.28 MHz   4.00 MHz         2.61 MHz
64 kbps (celt)   4.09 MHz    8.08 MHz         6.24 MHz
128 kbps (celt)  1.93 MHz    8.83 MHz         6.53 MHz

Change-Id: I851733a8a5824b61feb363a173091bc7e6629b58

1 files changed, 36 insertions, 17 deletions

diff --git a/lib/rbcodec/codecs/libopus/silk/PLC.c b/lib/rbcodec/codecs/libopus/silk/PLC.c
index 01f40014c4..8b0a8fe57d 100644
--- a/lib/rbcodec/codecs/libopus/silk/PLC.c
+++ b/lib/rbcodec/codecs/libopus/silk/PLC.c
@@ -165,6 +165,30 @@ static OPUS_INLINE void silk_PLC_update(
     psPLC->nb_subfr = psDec->nb_subfr;
 }
 
+static OPUS_INLINE void silk_PLC_energy(opus_int32 *energy1, opus_int *shift1, opus_int32 *energy2, opus_int *shift2,
+      const opus_int32 *exc_Q14, const opus_int32 *prevGain_Q10, int subfr_length, int nb_subfr)
+{
+    int i, k;
+    VARDECL( opus_int16, exc_buf );
+    opus_int16 *exc_buf_ptr;
+    SAVE_STACK;
+    ALLOC( exc_buf, 2*subfr_length, opus_int16 );
+    /* Find random noise component */
+    /* Scale previous excitation signal */
+    exc_buf_ptr = exc_buf;
+    for( k = 0; k < 2; k++ ) {
+        for( i = 0; i < subfr_length; i++ ) {
+            exc_buf_ptr[ i ] = (opus_int16)silk_SAT16( silk_RSHIFT(
+                silk_SMULWW( exc_Q14[ i + ( k + nb_subfr - 2 ) * subfr_length ], prevGain_Q10[ k ] ), 8 ) );
+        }
+        exc_buf_ptr += subfr_length;
+    }
+    /* Find the subframe with lowest energy of the last two and use that as random noise generator */
+    silk_sum_sqr_shift( energy1, shift1, exc_buf,                  subfr_length );
+    silk_sum_sqr_shift( energy2, shift2, &exc_buf[ subfr_length ], subfr_length );
+    RESTORE_STACK;
+}
+
 static OPUS_INLINE void silk_PLC_conceal(
     silk_decoder_state                  *psDec,             /* I/O Decoder state        */
     silk_decoder_control                *psDecCtrl,         /* I/O Decoder control      */
@@ -177,19 +201,26 @@ static OPUS_INLINE void silk_PLC_conceal(
     opus_int32 energy1, energy2, *rand_ptr, *pred_lag_ptr;
     opus_int32 LPC_pred_Q10, LTP_pred_Q12;
     opus_int16 rand_scale_Q14;
-    opus_int16 *B_Q14, *exc_buf_ptr;
+    opus_int16 *B_Q14;
     opus_int32 *sLPC_Q14_ptr;
-    VARDECL( opus_int16, exc_buf );
     opus_int16 A_Q12[ MAX_LPC_ORDER ];
+#ifdef SMALL_FOOTPRINT
+    opus_int16 *sLTP;
+#else
     VARDECL( opus_int16, sLTP );
+#endif
     VARDECL( opus_int32, sLTP_Q14 );
     silk_PLC_struct *psPLC = &psDec->sPLC;
     opus_int32 prevGain_Q10[2];
     SAVE_STACK;
 
-    ALLOC( exc_buf, 2*psPLC->subfr_length, opus_int16 );
-    ALLOC( sLTP, psDec->ltp_mem_length, opus_int16 );
     ALLOC( sLTP_Q14, psDec->ltp_mem_length + psDec->frame_length, opus_int32 );
+#ifdef SMALL_FOOTPRINT
+    /* Ugly hack that breaks aliasing rules to save stack: put sLTP at the very end of sLTP_Q14. */
+    sLTP = ((opus_int16*)&sLTP_Q14[psDec->ltp_mem_length + psDec->frame_length])-psDec->ltp_mem_length;
+#else
+    ALLOC( sLTP, psDec->ltp_mem_length, opus_int16 );
+#endif
 
     prevGain_Q10[0] = silk_RSHIFT( psPLC->prevGain_Q16[ 0 ], 6);
     prevGain_Q10[1] = silk_RSHIFT( psPLC->prevGain_Q16[ 1 ], 6);
@@ -198,19 +229,7 @@ static OPUS_INLINE void silk_PLC_conceal(
        silk_memset( psPLC->prevLPC_Q12, 0, sizeof( psPLC->prevLPC_Q12 ) );
     }
 
-    /* Find random noise component */
-    /* Scale previous excitation signal */
-    exc_buf_ptr = exc_buf;
-    for( k = 0; k < 2; k++ ) {
-        for( i = 0; i < psPLC->subfr_length; i++ ) {
-            exc_buf_ptr[ i ] = (opus_int16)silk_SAT16( silk_RSHIFT(
-                silk_SMULWW( psDec->exc_Q14[ i + ( k + psPLC->nb_subfr - 2 ) * psPLC->subfr_length ], prevGain_Q10[ k ] ), 8 ) );
-        }
-        exc_buf_ptr += psPLC->subfr_length;
-    }
-    /* Find the subframe with lowest energy of the last two and use that as random noise generator */
-    silk_sum_sqr_shift( &energy1, &shift1, exc_buf,                         psPLC->subfr_length );
-    silk_sum_sqr_shift( &energy2, &shift2, &exc_buf[ psPLC->subfr_length ], psPLC->subfr_length );
+    silk_PLC_energy(&energy1, &shift1, &energy2, &shift2, psDec->exc_Q14, prevGain_Q10, psDec->subfr_length, psDec->nb_subfr);
 
     if( silk_RSHIFT( energy1, shift2 ) < silk_RSHIFT( energy2, shift1 ) ) {
         /* First sub-frame has lowest energy */