[FFmpeg-cvslog] dca: factorize scaling in inverse ADPCM

Janne Grunau git at videolan.org
Fri Feb 28 21:38:00 CET 2014


ffmpeg | branch: master | Janne Grunau <janne-libav at jannau.net> | Fri Feb 21 13:13:36 2014 +0100| [7686afd049be98d18663682b92d983340fa2c305] | committer: Janne Grunau

dca: factorize scaling in inverse ADPCM

Based on a patch from Christophe Gisquet.

Unrolling of the m == 0 case avoids a possible use of the uninitilized
value sum when s->predictor_history is not set. I failed to find a
sample for it. It also reduced the cycle count from 220 to 150 on
sandy bridge, x86_64 linux, gcc 4.8.2 compared to his patch.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=7686afd049be98d18663682b92d983340fa2c305
---

 libavcodec/dcadec.c |   27 +++++++++++++++++++--------
 1 file changed, 19 insertions(+), 8 deletions(-)

diff --git a/libavcodec/dcadec.c b/libavcodec/dcadec.c
index c251db8..5ede61d 100644
--- a/libavcodec/dcadec.c
+++ b/libavcodec/dcadec.c
@@ -1192,16 +1192,27 @@ static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
              */
             if (s->prediction_mode[k][l]) {
                 int n;
-                for (m = 0; m < 8; m++) {
-                    for (n = 1; n <= 4; n++)
+                if (s->predictor_history)
+                    subband_samples[k][l][0] += (adpcm_vb[s->prediction_vq[k][l]][0] *
+                                                 s->subband_samples_hist[k][l][3] +
+                                                 adpcm_vb[s->prediction_vq[k][l]][1] *
+                                                 s->subband_samples_hist[k][l][2] +
+                                                 adpcm_vb[s->prediction_vq[k][l]][2] *
+                                                 s->subband_samples_hist[k][l][1] +
+                                                 adpcm_vb[s->prediction_vq[k][l]][3] *
+                                                 s->subband_samples_hist[k][l][0]) *
+                                                (1.0f / 8192);
+                for (m = 1; m < 8; m++) {
+                    float sum = adpcm_vb[s->prediction_vq[k][l]][0] *
+                                subband_samples[k][l][m - 1];
+                    for (n = 2; n <= 4; n++)
                         if (m >= n)
-                            subband_samples[k][l][m] +=
-                                (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
-                                 subband_samples[k][l][m - n] / 8192);
+                            sum += adpcm_vb[s->prediction_vq[k][l]][n - 1] *
+                                   subband_samples[k][l][m - n];
                         else if (s->predictor_history)
-                            subband_samples[k][l][m] +=
-                                (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
-                                 s->subband_samples_hist[k][l][m - n + 4] / 8192);
+                            sum += adpcm_vb[s->prediction_vq[k][l]][n - 1] *
+                                   s->subband_samples_hist[k][l][m - n + 4];
+                    subband_samples[k][l][m] += sum * 1.0f / 8192;
                 }
             }
         }



More information about the ffmpeg-cvslog mailing list