[FFmpeg-cvslog] vp9: write uveob as 16-bit value for 16x16/32x32 transforms.

Ronald S. Bultje git at videolan.org
Fri Apr 25 14:17:18 CEST 2014


ffmpeg | branch: master | Ronald S. Bultje <rsbultje at gmail.com> | Fri Apr 25 07:51:39 2014 -0400| [6d69f9f37689c999815a65a2d99999fad3a41705] | committer: Ronald S. Bultje

vp9: write uveob as 16-bit value for 16x16/32x32 transforms.

This fixes make fate-vp9-00-quantizer-01 THREADS=2.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=6d69f9f37689c999815a65a2d99999fad3a41705
---

 libavcodec/vp9.c |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/libavcodec/vp9.c b/libavcodec/vp9.c
index cbc885b..ccdf78f 100644
--- a/libavcodec/vp9.c
+++ b/libavcodec/vp9.c
@@ -2258,7 +2258,11 @@ static void decode_coeffs(AVCodecContext *ctx)
                                   16 * step * step, c, e, p, a[x] + l[y], \
                                   uvscan, uvnb, uv_band_counts, qmul[1]); \
             a[x] = l[y] = !!res; \
-            s->uveob[pl][n] = res; \
+            if (step >= 4) { \
+                AV_WN16A(&s->uveob[pl][n], res); \
+            } else { \
+                s->uveob[pl][n] = res; \
+            } \
         } \
     }
 



More information about the ffmpeg-cvslog mailing list