diff --git a/lpcnet b/lpcnet
index 85d12691acd49938f8688479496e669076a78ba7..83e89e7e8b8efeaea9b2c67eb2cb9da8a86e27e6 160000
--- a/lpcnet
+++ b/lpcnet
@@ -1 +1 @@
-Subproject commit 85d12691acd49938f8688479496e669076a78ba7
+Subproject commit 83e89e7e8b8efeaea9b2c67eb2cb9da8a86e27e6
diff --git a/silk/dred_config.h b/silk/dred_config.h
index ff9903bafe9b8177e65b638aa70e4a8f4767600b..ccaf76dbb30de309a5f36cd208b1e0aa833c8420 100644
--- a/silk/dred_config.h
+++ b/silk/dred_config.h
@@ -36,7 +36,7 @@
#define DRED_LATENT_DIM 80
#define DRED_STATE_DIM 24
#define DRED_MAX_FRAMES 100
-#define DRED_SILK_ENCODER_DELAY 79
+#define DRED_SILK_ENCODER_DELAY (79+12)
#define DRED_FRAME_SIZE 160
#define DRED_DFRAME_SIZE (2 * (DRED_FRAME_SIZE))
#define DRED_MAX_DATA_SIZE 1000
diff --git a/silk/dred_encoder.c b/silk/dred_encoder.c
index 83815b1f7905f442d1bb536b8c03d7ca6f8cfd3f..e1db96a2dfdbca2db2d9c103e2986be0f3899673 100644
--- a/silk/dred_encoder.c
+++ b/silk/dred_encoder.c
@@ -41,36 +41,48 @@
#include "celt/entenc.h"
#include "dred_decoder.h"
+#include "float_cast.h"
+#include "os_support.h"
-void init_dred_encoder(DREDEnc* enc)
+void dred_encoder_reset(DREDEnc* enc)
{
- memset(enc, 0, sizeof(*enc));
+ RNN_CLEAR((char*)&enc->DREDENC_RESET_START,
+ sizeof(DREDEnc)-
+ ((char*)&enc->DREDENC_RESET_START - (char*)enc));
+ lpcnet_encoder_init(&enc->lpcnet_enc_state);
+ DRED_rdovae_init_encoder(&enc->rdovae_enc);
+}
+
+void dred_encoder_init(DREDEnc* enc, int Fs, int channels)
+{
+ enc->Fs = Fs;
+ enc->channels = channels;
#ifndef USE_WEIGHTS_FILE
init_rdovaeenc(&enc->model, rdovae_enc_arrays);
#endif
- lpcnet_encoder_init(&enc->lpcnet_enc_state);
- DRED_rdovae_init_encoder(&enc->rdovae_enc);
+ dred_encoder_reset(enc);
}
-void dred_process_silk_frame(DREDEnc *enc, const opus_int16 *silk_frame)
+void dred_process_silk_frame(DREDEnc *enc, const float *silk_frame)
{
+ int i;
float feature_buffer[2 * 36];
float input_buffer[2*DRED_NUM_FEATURES] = {0};
/* delay signal by 79 samples */
- memmove(enc->input_buffer, enc->input_buffer + DRED_DFRAME_SIZE, DRED_SILK_ENCODER_DELAY * sizeof(*enc->input_buffer));
- memcpy(enc->input_buffer + DRED_SILK_ENCODER_DELAY, silk_frame, DRED_DFRAME_SIZE * sizeof(*silk_frame));
+ OPUS_MOVE(enc->input_buffer, enc->input_buffer + DRED_DFRAME_SIZE, DRED_SILK_ENCODER_DELAY);
+ for (i=0;i<DRED_DFRAME_SIZE;i++) enc->input_buffer[DRED_SILK_ENCODER_DELAY+i] = FLOAT2INT16(silk_frame[i]);
/* shift latents buffer */
- memmove(enc->latents_buffer + DRED_LATENT_DIM, enc->latents_buffer, (DRED_MAX_FRAMES - 1) * DRED_LATENT_DIM * sizeof(*enc->latents_buffer));
+ OPUS_MOVE(enc->latents_buffer + DRED_LATENT_DIM, enc->latents_buffer, (DRED_MAX_FRAMES - 1) * DRED_LATENT_DIM);
/* calculate LPCNet features */
- lpcnet_compute_single_frame_features(&enc->lpcnet_enc_state, enc->input_buffer, feature_buffer);
- lpcnet_compute_single_frame_features(&enc->lpcnet_enc_state, enc->input_buffer + DRED_FRAME_SIZE, feature_buffer + 36);
+ lpcnet_compute_single_frame_features_float(&enc->lpcnet_enc_state, enc->input_buffer, feature_buffer);
+ lpcnet_compute_single_frame_features_float(&enc->lpcnet_enc_state, enc->input_buffer + DRED_FRAME_SIZE, feature_buffer + 36);
/* prepare input buffer (discard LPC coefficients) */
- memcpy(input_buffer, feature_buffer, DRED_NUM_FEATURES * sizeof(input_buffer[0]));
- memcpy(input_buffer + DRED_NUM_FEATURES, feature_buffer + 36, DRED_NUM_FEATURES * sizeof(input_buffer[0]));
+ OPUS_COPY(input_buffer, feature_buffer, DRED_NUM_FEATURES);
+ OPUS_COPY(input_buffer + DRED_NUM_FEATURES, feature_buffer + 36, DRED_NUM_FEATURES);
/* run RDOVAE encoder */
DRED_rdovae_encode_dframe(&enc->rdovae_enc, &enc->model, enc->latents_buffer, enc->state_buffer, input_buffer);
diff --git a/silk/dred_encoder.h b/silk/dred_encoder.h
index 8fcb07d1f31e7aefeb0a7e369a662af3bccd0e3b..414f73df5088c4807342fdf7c2d860aa226d3d32 100644
--- a/silk/dred_encoder.h
+++ b/silk/dred_encoder.h
@@ -38,7 +38,11 @@
typedef struct {
RDOVAEEnc model;
- opus_int16 input_buffer[DRED_DFRAME_SIZE + DRED_SILK_ENCODER_DELAY];
+ int Fs;
+ int channels;
+
+#define DREDENC_RESET_START input_buffer
+ float input_buffer[DRED_DFRAME_SIZE + DRED_SILK_ENCODER_DELAY];
float latents_buffer[DRED_MAX_FRAMES * DRED_LATENT_DIM];
int latents_buffer_fill;
float state_buffer[24];
@@ -47,11 +51,12 @@ typedef struct {
} DREDEnc;
-void init_dred_encoder(DREDEnc* enc);
+void dred_encoder_init(DREDEnc* enc, int Fs, int channels);
+void dred_encoder_reset(DREDEnc* enc);
void dred_deinit_encoder(DREDEnc *enc);
-void dred_process_silk_frame(DREDEnc *enc, const opus_int16 *silk_frame);
+void dred_process_silk_frame(DREDEnc *enc, const float *silk_frame);
int dred_encode_silk_frame(DREDEnc *enc, unsigned char *buf, int max_chunks, int max_bytes);
diff --git a/silk/enc_API.c b/silk/enc_API.c
index bdb858afc1ee5302702087edf1db7c161c501fc5..ff3070a1d7dfa6d4d48f344e3a2d29ad584e8484 100644
--- a/silk/enc_API.c
+++ b/silk/enc_API.c
@@ -468,15 +468,6 @@ opus_int silk_Encode( /* O Returns error co
}
silk_encode_do_VAD_Fxx( &psEnc->state_Fxx[ 0 ], activity );
-#ifdef ENABLE_NEURAL_FEC
- if ( encControl->useDRED ) {
- /* DRED Encoder */
- dred_process_silk_frame( &psEnc->state_Fxx[ 0 ].sCmn.dred_encoder, &psEnc->state_Fxx[ 0 ].sCmn.inputBuf[0] );
- } else {
- psEnc->state_Fxx[ 0 ].sCmn.dred_encoder.latents_buffer_fill = 0;
- }
-#endif
-
/* Encode */
for( n = 0; n < encControl->nChannelsInternal; n++ ) {
opus_int maxBits, useCBR;
diff --git a/silk/init_encoder.c b/silk/init_encoder.c
index 50ec22b34f8cb1f0f27fba255a132e86ca1a4f40..78b5ee88b4ae80f14d9779ced3c3f53ff0a00f2e 100644
--- a/silk/init_encoder.c
+++ b/silk/init_encoder.c
@@ -64,9 +64,5 @@ opus_int silk_init_encoder(
/* Initialize Silk VAD */
ret += silk_VAD_Init( &psEnc->sCmn.sVAD );
-#ifdef ENABLE_NEURAL_FEC
- /* Initialize DRED Encoder */
- init_dred_encoder( &psEnc->sCmn.dred_encoder );
-#endif
return ret;
}
diff --git a/silk/structs.h b/silk/structs.h
index 6a0889eeeaee87bf40b85d503bb55086a51e42ba..2058d4e86d23864f4c597a65689b173e206f8193 100644
--- a/silk/structs.h
+++ b/silk/structs.h
@@ -235,9 +235,6 @@ typedef struct {
opus_int LBRR_GainIncreases; /* Gains increment for coding LBRR frames */
SideInfoIndices indices_LBRR[ MAX_FRAMES_PER_PACKET ];
opus_int8 pulses_LBRR[ MAX_FRAMES_PER_PACKET ][ MAX_FRAME_LENGTH ];
-#ifdef ENABLE_NEURAL_FEC
- DREDEnc dred_encoder;
-#endif
} silk_encoder_state;
diff --git a/src/opus_encoder.c b/src/opus_encoder.c
index 3764b373caaadbc5be82a725f4cf9232fee2e4d6..c6759c8c733ddf06762c4f5fb57151c3f403f9b4 100644
--- a/src/opus_encoder.c
+++ b/src/opus_encoder.c
@@ -67,6 +67,9 @@ struct OpusEncoder {
int celt_enc_offset;
int silk_enc_offset;
silk_EncControlStruct silk_mode;
+#ifdef ENABLE_NEURAL_FEC
+ DREDEnc dred_encoder;
+#endif
int application;
int channels;
int delay_compensation;
@@ -240,6 +243,11 @@ int opus_encoder_init(OpusEncoder* st, opus_int32 Fs, int channels, int applicat
celt_encoder_ctl(celt_enc, CELT_SET_SIGNALLING(0));
celt_encoder_ctl(celt_enc, OPUS_SET_COMPLEXITY(st->silk_mode.complexity));
+#ifdef ENABLE_NEURAL_FEC
+ /* Initialize DRED Encoder */
+ dred_encoder_init( &st->dred_encoder, Fs, channels );
+#endif
+
st->use_vbr = 1;
/* Makes constrained VBR the default (safer for real-time use) */
st->vbr_constraint = 1;
@@ -1461,6 +1469,10 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_
{
silk_EncControlStruct dummy;
silk_InitEncoder( silk_enc, st->arch, &dummy);
+#ifdef ENABLE_NEURAL_FEC
+ /* Initialize DRED Encoder */
+ dred_encoder_reset( &st->dred_encoder );
+#endif
prefill=1;
}
@@ -1814,6 +1826,15 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_
}
}
+#ifdef ENABLE_NEURAL_FEC
+ if ( st->dred_duration > 0 ) {
+ /* DRED Encoder */
+ dred_process_silk_frame( &st->dred_encoder, &pcm_buf[total_buffer*st->channels] );
+ } else {
+ st->dred_encoder.latents_buffer_fill = 0;
+ }
+#endif
+
if (prefill)
{
opus_int32 zero=0;
@@ -1888,7 +1909,7 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_
#ifdef ENABLE_NEURAL_FEC
/* If we're not in SILK mode, delete all the processed DRED.
TODO: Remove this if/when DRED gets encoded for CELT. */
- DREDEnc *dred = &((silk_encoder*)silk_enc)->state_Fxx[0].sCmn.dred_encoder;
+ DREDEnc *dred = &st->dred_encoder;
dred->latents_buffer_fill = 0;
#endif
}
@@ -2196,7 +2217,6 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_
unsigned char buf[DRED_MAX_DATA_SIZE];
int dred_chunks;
int dred_bytes_left;
- DREDEnc *dred = &((silk_encoder*)silk_enc)->state_Fxx[0].sCmn.dred_encoder;
dred_chunks = IMIN(st->dred_duration/4, DRED_NUM_REDUNDANCY_FRAMES/2);
dred_bytes_left = IMIN(DRED_MAX_DATA_SIZE, max_data_bytes-ret-2);
/* Check whether we actually have something to encode. */
@@ -2206,7 +2226,7 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_
These bytes will be removed once extension is finalized. */
buf[0] = 'D';
buf[1] = DRED_VERSION;
- dred_bytes = dred_encode_silk_frame(dred, buf+2, dred_chunks, dred_bytes_left-2);
+ dred_bytes = dred_encode_silk_frame(&st->dred_encoder, buf+2, dred_chunks, dred_bytes_left-2);
dred_bytes += 2;
celt_assert(dred_bytes <= dred_bytes_left);
extension.id = 127;
@@ -2763,6 +2783,10 @@ int opus_encoder_ctl(OpusEncoder *st, int request, ...)
celt_encoder_ctl(celt_enc, OPUS_RESET_STATE);
silk_InitEncoder( silk_enc, st->arch, &dummy );
+#ifdef ENABLE_NEURAL_FEC
+ /* Initialize DRED Encoder */
+ dred_encoder_reset( &st->dred_encoder );
+#endif
st->stream_channels = st->channels;
st->hybrid_stereo_width_Q14 = 1 << 14;
st->prev_HB_gain = Q15ONE;