diff options
author | Ramiro Polla <ramiro.polla@gmail.com> | 2009-12-07 03:52:55 +0000 |
---|---|---|
committer | Ramiro Polla <ramiro.polla@gmail.com> | 2009-12-07 03:52:55 +0000 |
commit | a8bcc7b42b275b18bbc2f65f48fa637e20847d36 (patch) | |
tree | 6be73e34e554096453cd0402e06262d0795d4976 /libswscale | |
parent | 445a9deea684528e0c27c875e19ac153a1b9efec (diff) | |
download | ffmpeg-streaming-a8bcc7b42b275b18bbc2f65f48fa637e20847d36.zip ffmpeg-streaming-a8bcc7b42b275b18bbc2f65f48fa637e20847d36.tar.gz |
Document the differences between the two asm fragments in initMMX2HScaler().
Originally committed as revision 29978 to svn://svn.mplayerhq.hu/mplayer/trunk/libswscale
Diffstat (limited to 'libswscale')
-rw-r--r-- | libswscale/swscale.c | 6 |
1 files changed, 6 insertions, 0 deletions
diff --git a/libswscale/swscale.c b/libswscale/swscale.c index 222c0c0..079ed31 100644 --- a/libswscale/swscale.c +++ b/libswscale/swscale.c @@ -1697,6 +1697,12 @@ static int initMMX2HScaler(int dstW, int xInc, uint8_t *filterCode, int16_t *fil int xpos, i; // create an optimized horizontal scaling routine + /* This scaler is made of runtime-generated MMX2 code using specially + * tuned pshufw instructions. For every four output pixels, if four + * input pixels are enough for the fast bilinear scaling, then a chunk + * of fragmentB is used. If five input pixels are needed, then a chunk + * of fragmentA is used. + */ //code fragment |