[FFmpeg-devel] [PATCH 05/11] dca dsp C: factorize scaling in lfe_fir

[Christophe Gisquet]

Scaling on output leads to respectively 32 or 64 multiplications, while doing
it on the input is respectively 8 or 4 multiplications. As a side note,
the trick can be used in other implementations.

Win64: 728 to 562 cycles.
---
 libavcodec/dcadsp.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/libavcodec/dcadsp.c b/libavcodec/dcadsp.c
index b14c435..6600fef 100644
--- a/libavcodec/dcadsp.c
+++ b/libavcodec/dcadsp.c
@@ -39,19 +39,23 @@ dca_lfe_fir(float *out, const float *in, const float *coefs,
     float *out2 = out + decifactor;
     const float *cf0 = coefs;
     const float *cf1 = coefs + 256;
+    float scaled_in[8];
     int j, k;
 
+    for (j = 0; j < 256 / decifactor; j++)
+        scaled_in[j] = in[-j] * scale;
+
     /* One decimated sample generates 2*decifactor interpolated ones */
     for (k = 0; k < decifactor; k++) {
         float v0 = 0.0;
         float v1 = 0.0;
         for (j = 0; j < 256 / decifactor; j++) {
-            float s = in[-j];
+            float s = scaled_in[j];
             v0 += s * *cf0++;
             v1 += s * *--cf1;
         }
-        *out++  = v0 * scale;
-        *out2++ = v1 * scale;
+        *out++  = v0;
+        *out2++ = v1;
     }
 }
 
-- 
1.8.0.msysgit.0