opus_pvq: port to allow for SIMD functions

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

libavcodec/opus_celt.c

@@ -753,15 +753,15 @@ static void celt_decode_bands(CeltFrame *f, OpusRangeCoder *rc)
         }
 
         if (f->dual_stereo) {
-            cm[0] = ff_celt_decode_band(f, rc, i, X, NULL, band_size, b / 2, f->blocks,
+            cm[0] = f->pvq->decode_band(f->pvq, f, rc, i, X, NULL, band_size, b / 2, f->blocks,
                                         effective_lowband != -1 ? norm + (effective_lowband << f->size) : NULL, f->size,
                                         norm + band_offset, 0, 1.0f, lowband_scratch, cm[0]);
 
-            cm[1] = ff_celt_decode_band(f, rc, i, Y, NULL, band_size, b/2, f->blocks,
+            cm[1] = f->pvq->decode_band(f->pvq, f, rc, i, Y, NULL, band_size, b/2, f->blocks,
                                         effective_lowband != -1 ? norm2 + (effective_lowband << f->size) : NULL, f->size,
                                         norm2 + band_offset, 0, 1.0f, lowband_scratch, cm[1]);
         } else {
-            cm[0] = ff_celt_decode_band(f, rc, i, X, Y, band_size, b, f->blocks,
+            cm[0] = f->pvq->decode_band(f->pvq, f, rc, i, X, Y, band_size, b, f->blocks,
                                         effective_lowband != -1 ? norm + (effective_lowband << f->size) : NULL, f->size,
                                         norm + band_offset, 0, 1.0f, lowband_scratch, cm[0]|cm[1]);
             cm[1] = cm[0];
@@ -984,6 +984,8 @@ void ff_celt_free(CeltFrame **f)
     for (i = 0; i < FF_ARRAY_ELEMS(frm->imdct); i++)
         ff_mdct15_uninit(&frm->imdct[i]);
 
+    ff_celt_pvq_uninit(&frm->pvq);
+
     av_freep(&frm->dsp);
     av_freep(f);
 }
@@ -1006,11 +1008,12 @@ int ff_celt_init(AVCodecContext *avctx, CeltFrame **f, int output_channels)
     frm->avctx           = avctx;
     frm->output_channels = output_channels;
 
-    for (i = 0; i < FF_ARRAY_ELEMS(frm->imdct); i++) {
-        ret = ff_mdct15_init(&frm->imdct[i], 1, i + 3, -1.0f);
-        if (ret < 0)
+    for (i = 0; i < FF_ARRAY_ELEMS(frm->imdct); i++)
+        if ((ret = ff_mdct15_init(&frm->imdct[i], 1, i + 3, -1.0f)) < 0)
             goto fail;
-    }
+
+    if ((ret = ff_celt_pvq_init(&frm->pvq)) < 0)
+        goto fail;
 
     frm->dsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
     if (!frm->dsp) {

libavcodec/opus_celt.h

@@ -27,6 +27,7 @@
 #include <float.h>
 
 #include "opus.h"
+#include "opus_pvq.h"
 
 #include "mdct15.h"
 #include "libavutil/float_dsp.h"
@@ -43,6 +44,8 @@
 #define CELT_POSTFILTER_MINPERIOD    15
 #define CELT_ENERGY_SILENCE          (-28.0f)
 
+typedef struct CeltPVQ CeltPVQ;
+
 enum CeltSpread {
     CELT_SPREAD_NONE,
     CELT_SPREAD_LIGHT,
@@ -92,6 +95,7 @@ struct CeltFrame {
     MDCT15Context       *imdct[4];
     AVFloatDSPContext   *dsp;
     CeltBlock           block[2];
+    CeltPVQ             *pvq;
     int channels;
     int output_channels;
 
@@ -125,8 +129,6 @@ struct CeltFrame {
     int fine_priority[CELT_MAX_BANDS];
     int pulses       [CELT_MAX_BANDS];
     int tf_change    [CELT_MAX_BANDS];
-
-    DECLARE_ALIGNED(32, float, scratch)[22 * 8]; // MAX(ff_celt_freq_range) * 1<<CELT_MAX_LOG_BLOCKS
 };
 
 /* LCG for noise generation */

libavcodec/opus_pvq.h

@@ -23,22 +23,28 @@
 #ifndef AVCODEC_OPUS_PVQ_H
 #define AVCODEC_OPUS_PVQ_H
 
-#include "opus.h"
 #include "opus_celt.h"
 
-/* Decodes a band using PVQ */
-uint32_t ff_celt_decode_band(CeltFrame *f, OpusRangeCoder *rc, const int band,
-                             float *X, float *Y, int N, int b, uint32_t blocks,
-                             float *lowband, int duration, float *lowband_out, int level,
-                             float gain, float *lowband_scratch, int fill);
+#define QUANT_FN(name) uint32_t (name)(struct CeltPVQ *pvq, CeltFrame *f,            \
+                                       OpusRangeCoder *rc, const int band, float *X, \
+                                       float *Y, int N, int b, uint32_t blocks,      \
+                                       float *lowband, int duration,                 \
+                                       float *lowband_out, int level, float gain,    \
+                                       float *lowband_scratch, int fill)
 
-/* Encodes a band using PVQ */
-uint32_t ff_celt_encode_band(CeltFrame *f, OpusRangeCoder *rc, const int band,
-                             float *X, float *Y, int N, int b, uint32_t blocks,
-                             float *lowband, int duration, float *lowband_out, int level,
-                             float gain, float *lowband_scratch, int fill);
+struct CeltPVQ {
+    DECLARE_ALIGNED(32, int,   qcoeff      )[176];
+    DECLARE_ALIGNED(32, float, hadamard_tmp)[176];
 
-float ff_celt_quant_band_cost(CeltFrame *f, OpusRangeCoder *rc, int band,
-                              float *bits, float lambda);
+    float (*pvq_search)(float *X, int *y, int K, int N);
+
+    QUANT_FN(*decode_band);
+    QUANT_FN(*encode_band);
+    float (*band_cost)(struct CeltPVQ *pvq, CeltFrame *f, OpusRangeCoder *rc,
+                       int band, float *bits, float lambda);
+};
+
+int  ff_celt_pvq_init  (struct CeltPVQ **pvq);
+void ff_celt_pvq_uninit(struct CeltPVQ **pvq);
 
 #endif /* AVCODEC_OPUS_PVQ_H */

libavcodec/opusenc.c

@@ -55,6 +55,7 @@ typedef struct OpusEncContext {
     AudioFrameQueue afq;
     AVFloatDSPContext *dsp;
     MDCT15Context *mdct[CELT_BLOCK_NB];
+    CeltPVQ *pvq;
     struct FFBufQueue bufqueue;
 
     enum OpusMode mode;
@@ -797,15 +798,15 @@ static void celt_quant_bands(OpusRangeCoder *rc, CeltFrame *f)
         }
 
         if (f->dual_stereo) {
-            cm[0] = ff_celt_encode_band(f, rc, i, X, NULL, band_size, b / 2, f->blocks,
+            cm[0] = f->pvq->encode_band(f->pvq, f, rc, i, X, NULL, band_size, b / 2, f->blocks,
                                         effective_lowband != -1 ? norm + (effective_lowband << f->size) : NULL, f->size,
                                         norm + band_offset, 0, 1.0f, lowband_scratch, cm[0]);
 
-            cm[1] = ff_celt_encode_band(f, rc, i, Y, NULL, band_size, b / 2, f->blocks,
+            cm[1] = f->pvq->encode_band(f->pvq, f, rc, i, Y, NULL, band_size, b / 2, f->blocks,
                                         effective_lowband != -1 ? norm2 + (effective_lowband << f->size) : NULL, f->size,
                                         norm2 + band_offset, 0, 1.0f, lowband_scratch, cm[1]);
         } else {
-            cm[0] = ff_celt_encode_band(f, rc, i, X, Y, band_size, b, f->blocks,
+            cm[0] = f->pvq->encode_band(f->pvq, f, rc, i, X, Y, band_size, b, f->blocks,
                                         effective_lowband != -1 ? norm + (effective_lowband << f->size) : NULL, f->size,
                                         norm + band_offset, 0, 1.0f, lowband_scratch, cm[0] | cm[1]);
             cm[1] = cm[0];
@@ -883,6 +884,7 @@ static void ff_opus_psy_celt_frame_setup(OpusEncContext *s, CeltFrame *f, int in
 
     f->avctx = s->avctx;
     f->dsp = s->dsp;
+    f->pvq = s->pvq;
     f->start_band = (s->mode == OPUS_MODE_HYBRID) ? 17 : 0;
     f->end_band = ff_celt_band_end[s->bandwidth];
     f->channels = s->channels;
@@ -1019,6 +1021,7 @@ static av_cold int opus_encode_end(AVCodecContext *avctx)
     for (i = 0; i < CELT_BLOCK_NB; i++)
         ff_mdct15_uninit(&s->mdct[i]);
 
+    ff_celt_pvq_uninit(&s->pvq);
     av_freep(&s->dsp);
     av_freep(&s->frame);
     av_freep(&s->rc);
@@ -1075,6 +1078,9 @@ static av_cold int opus_encode_init(AVCodecContext *avctx)
 
     ff_af_queue_init(avctx, &s->afq);
 
+    if ((ret = ff_celt_pvq_init(&s->pvq)) < 0)
+        return ret;
+
     if (!(s->dsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT)))
         return AVERROR(ENOMEM);