diff options
author | Vladimir Voroshilov <voroshil@gmail.com> | 2008-05-11 14:07:11 +0400 |
---|---|---|
committer | Vladimir Voroshilov <voroshil@gmail.com> | 2008-05-11 14:07:11 +0400 |
commit | 542c064d1bae296fc55c3058f3eda72df7879d4d (patch) | |
tree | 71191e000ab7f022abb5d55e78e9bae4efddc883 /libavcodec/acelp_filters.c | |
parent | 7c96587c17bd35b70b750f0356996528a3cebe62 (diff) |
Update comment to version, negotiated with Diego, and
fix missing period (not latest revision of patch
was wrongly committed).
Originally committed as revision 13113 to svn://svn.ffmpeg.org/ffmpeg/trunk
Diffstat (limited to 'libavcodec/acelp_filters.c')
-rw-r--r-- | libavcodec/acelp_filters.c | 13 |
1 files changed, 6 insertions, 7 deletions
diff --git a/libavcodec/acelp_filters.c b/libavcodec/acelp_filters.c index b98c6f741e..7e33a674f1 100644 --- a/libavcodec/acelp_filters.c +++ b/libavcodec/acelp_filters.c @@ -37,13 +37,12 @@ void ff_acelp_convolve_circ( memset(fc_out, 0, subframe_size * sizeof(int16_t)); - /* Since there are few pulses over entire subframe (i.e. almost all - fc_in[i] are zero, in case of G.729D the buffer contains two non-zero - samples before the call to ff_acelp_enhance_harmonics, and (due to - pitch_delay bounded to [20; 143]) a maximum four non-zero samples - for a total of 40 after the call to it), it is faster to swap two loops - and process non-zero samples only. This will reduce the number of - multiplications from 40*40 to 4*40 for G.729D */ + /* Since there are few pulses over an entire subframe (i.e. almost + all fc_in[i] are zero) it is faster to swap two loops and process + non-zero samples only. In the case of G.729D the buffer contains + two non-zero samples before the call to ff_acelp_enhance_harmonics + and, due to pitch_delay being bounded by [20; 143], a maximum + of four non-zero samples for a total of 40 after the call. */ for(i=0; i<subframe_size; i++) { if(fc_in[i]) |