ra288: utilize DSPContext.vector_fmul()

2 files changed, 23 insertions, 20 deletions

diff --git a/libavcodec/ra288.c b/libavcodec/ra288.c
index d82e52df2f..c58bc31f62 100644
--- a/libavcodec/ra288.c
+++ b/libavcodec/ra288.c
@@ -26,6 +26,7 @@
 #include "lpc.h"
 #include "celp_math.h"
 #include "celp_filters.h"
+#include "dsputil.h"
 
 #define MAX_BACKWARD_FILTER_ORDER  36
 #define MAX_BACKWARD_FILTER_LEN    40
@@ -35,8 +36,9 @@
 #define RA288_BLOCKS_PER_FRAME 32
 
 typedef struct {
-    float sp_lpc[36];      ///< LPC coefficients for speech data (spec: A)
-    float gain_lpc[10];    ///< LPC coefficients for gain        (spec: GB)
+    DSPContext dsp;
+    DECLARE_ALIGNED(16, float,   sp_lpc)[FFALIGN(36, 8)];   ///< LPC coefficients for speech data (spec: A)
+    DECLARE_ALIGNED(16, float, gain_lpc)[FFALIGN(10, 8)];   ///< LPC coefficients for gain        (spec: GB)
 
     /** speech data history                                      (spec: SB).
      *  Its first 70 coefficients are updated only at backward filtering.
@@ -57,16 +59,12 @@ typedef struct {
 
 static av_cold int ra288_decode_init(AVCodecContext *avctx)
 {
+    RA288Context *ractx = avctx->priv_data;
     avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
+    dsputil_init(&ractx->dsp, avctx);
     return 0;
 }
 
-static void apply_window(float *tgt, const float *m1, const float *m2, int n)
-{
-    while (n--)
-        *tgt++ = *m1++ * *m2++;
-}
-
 static void convolve(float *tgt, const float *src, int len, int n)
 {
     for (; n >= 0; n--)
@@ -123,15 +121,18 @@ static void decode(RA288Context *ractx, float gain, int cb_coef)
  * @param out2    pointer to the recursive part of the output
  * @param window  pointer to the windowing function table
  */
-static void do_hybrid_window(int order, int n, int non_rec, float *out,
+static void do_hybrid_window(RA288Context *ractx,
+                             int order, int n, int non_rec, float *out,
                              float *hist, float *out2, const float *window)
 {
     int i;
     float buffer1[MAX_BACKWARD_FILTER_ORDER + 1];
     float buffer2[MAX_BACKWARD_FILTER_ORDER + 1];
-    float work[MAX_BACKWARD_FILTER_ORDER + MAX_BACKWARD_FILTER_LEN + MAX_BACKWARD_FILTER_NONREC];
+    LOCAL_ALIGNED_16(float, work)[FFALIGN(MAX_BACKWARD_FILTER_ORDER +
+                                          MAX_BACKWARD_FILTER_LEN   +
+                                          MAX_BACKWARD_FILTER_NONREC, 8)];
 
-    apply_window(work, window, hist, order + n + non_rec);
+    ractx->dsp.vector_fmul(work, window, hist, FFALIGN(order + n + non_rec, 8));
 
     convolve(buffer1, work + order    , n      , order);
     convolve(buffer2, work + order + n, non_rec, order);
@@ -148,16 +149,17 @@ static void do_hybrid_window(int order, int n, int non_rec, float *out,
 /**
  * Backward synthesis filter, find the LPC coefficients from past speech data.
  */
-static void backward_filter(float *hist, float *rec, const float *window,
+static void backward_filter(RA288Context *ractx,
+                            float *hist, float *rec, const float *window,
                             float *lpc, const float *tab,
                             int order, int n, int non_rec, int move_size)
 {
     float temp[MAX_BACKWARD_FILTER_ORDER+1];
 
-    do_hybrid_window(order, n, non_rec, temp, hist, rec, window);
+    do_hybrid_window(ractx, order, n, non_rec, temp, hist, rec, window);
 
     if (!compute_lpc_coefs(temp, order, lpc, 0, 1, 1))
-        apply_window(lpc, lpc, tab, order);
+        ractx->dsp.vector_fmul(lpc, lpc, tab, FFALIGN(order, 8));
 
     memmove(hist, hist + n, move_size*sizeof(*hist));
 }
@@ -198,10 +200,10 @@ static int ra288_decode_frame(AVCodecContext * avctx, void *data,
         out += RA288_BLOCK_SIZE;
 
         if ((i & 7) == 3) {
-            backward_filter(ractx->sp_hist, ractx->sp_rec, syn_window,
+            backward_filter(ractx, ractx->sp_hist, ractx->sp_rec, syn_window,
                             ractx->sp_lpc, syn_bw_tab, 36, 40, 35, 70);
 
-            backward_filter(ractx->gain_hist, ractx->gain_rec, gain_window,
+            backward_filter(ractx, ractx->gain_hist, ractx->gain_rec, gain_window,
                             ractx->gain_lpc, gain_bw_tab, 10, 8, 20, 28);
         }
     }
diff --git a/libavcodec/ra288.h b/libavcodec/ra288.h
index 8857f40ac8..1c98c16ea4 100644
--- a/libavcodec/ra288.h
+++ b/libavcodec/ra288.h
@@ -23,6 +23,7 @@
 #define AVCODEC_RA288_H
 
 #include <stdint.h>
+#include "dsputil.h"
 
 static const float amptable[8]={
      0.515625,  0.90234375,  1.57910156,  2.76342773,
@@ -96,7 +97,7 @@ static const int16_t codetable[128][5]={
     {  3746,  -606,    53,  -269, -3301}, {   606,  2018, -1316,  4064,   398}
 };
 
-static const float syn_window[111]={
+DECLARE_ALIGNED(16, static const float, syn_window)[FFALIGN(111, 8)]={
   0.576690972, 0.580838025, 0.585013986, 0.589219987, 0.59345597,  0.597723007,
   0.602020264, 0.606384277, 0.610748291, 0.615142822, 0.619598389, 0.624084473,
   0.628570557, 0.633117676, 0.637695313, 0.642272949, 0.646911621, 0.651580811,
@@ -118,7 +119,7 @@ static const float syn_window[111]={
   0.142852783, 0.0954284668,0.0477600098
 };
 
-static const float gain_window[38]={
+DECLARE_ALIGNED(16, static const float, gain_window)[FFALIGN(38, 8)]={
   0.505699992, 0.524200022, 0.54339999,  0.563300014, 0.583953857, 0.60534668,
   0.627502441, 0.650482178, 0.674316406, 0.699005127, 0.724578857, 0.75112915,
   0.778625488, 0.807128906, 0.836669922, 0.86730957,  0.899078369, 0.932006836,
@@ -129,7 +130,7 @@ static const float gain_window[38]={
 };
 
 /** synthesis bandwidth broadening table */
-static const float syn_bw_tab[36]={
+DECLARE_ALIGNED(16, static const float, syn_bw_tab)[FFALIGN(36, 8)] = {
   0.98828125,  0.976699829, 0.965254128, 0.953942537, 0.942763507, 0.931715488,
   0.920796931, 0.910006344, 0.899342179, 0.888803005, 0.878387332, 0.868093729,
   0.857920766, 0.847867012, 0.837931097, 0.828111589, 0.818407178, 0.808816493,
@@ -139,7 +140,7 @@ static const float syn_bw_tab[36]={
 };
 
 /** gain bandwidth broadening table */
-static const float gain_bw_tab[10]={
+DECLARE_ALIGNED(16, static const float, gain_bw_tab)[FFALIGN(10, 8)] = {
   0.90625,     0.821289063, 0.74432373,  0.674499512, 0.61126709,
   0.553955078, 0.50201416,  0.454956055, 0.41229248,  0.373657227
 };