1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-11-26 19:01:44 +02:00
Commit Graph

112 Commits

Author SHA1 Message Date
Vitor Sessak
08c433972e Use ff_dot_productf() in ra288.c
Originally committed as revision 15757 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-30 22:24:40 +00:00
Vitor Sessak
d552d04dc9 Simplify: avoid duplication backward_filter()
Originally committed as revision 15428 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-26 18:42:06 +00:00
Vitor Sessak
64e4af2ac1 Do instead of divising a float by a constant, multiply by its inverse
Originally committed as revision 15417 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-25 20:03:57 +00:00
Vitor Sessak
b6c7758168 Move constant multiplication out of the loop
Originally committed as revision 15404 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-24 20:31:24 +00:00
Vitor Sessak
e60daa340b Cosmetics: move convolve() together with the other DSP functions
Originally committed as revision 15403 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-24 20:05:22 +00:00
Vitor Sessak
a987a126fd Simplify: move division by constant off the loop
Originally committed as revision 15402 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-24 20:03:58 +00:00
Michael Niedermayer
645a6be1e7 Merge the 2 lpc loops.
This changes the output by (stddev:    0.21 PSNR:109.51, file:ddinterview28.ra),
the changes are due to float rounding inaccuracies and do not happen if doubles
are used.

Originally committed as revision 15394 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-23 23:16:48 +00:00
Michael Niedermayer
571bbaf814 Factorize lpc code slightly.
Originally committed as revision 15393 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-23 22:49:22 +00:00
Vitor Sessak
faf6d6964b Use SAMPLE_FMT_FLT instead of doing the float->int conversion in the
decoder. This changes the output very little and any difference should
be inaudible.

Originally committed as revision 15343 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-17 19:26:37 +00:00
Vitor Sessak
9547cadb8d Cosmetics: line breaks
Originally committed as revision 15308 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-13 16:56:58 +00:00
Vitor Sessak
111734de09 Simplify: use a single history buffer for gain and a single one for
speech instead of having two for each in the context.

Originally committed as revision 15307 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-13 16:49:05 +00:00
Vitor Sessak
9c10ab6cd1 Cosmetics: align spec references
Originally committed as revision 15300 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-11 19:10:26 +00:00
Vitor Sessak
5381a00a2b Rename function: s/colmult/apply_window/
Originally committed as revision 15299 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-11 19:06:54 +00:00
Diego Biurrun
36b3e36e00 misc spelling/wording/grammar fixes
Originally committed as revision 15296 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-11 08:13:23 +00:00
Vitor Sessak
fd8897e803 Convert table to integer.
Originally committed as revision 15292 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-10 19:00:02 +00:00
Vitor Sessak
1be0fc2909 Avoid duplicating compute_lpc_coefs() function in both the RA288 and AAC decoders.
Originally committed as revision 15193 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-09-04 11:03:14 +00:00
Vitor Sessak
8089c652a7 Check *data_size in decode_frame()
Originally committed as revision 14636 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-06 03:00:15 +00:00
Vitor Sessak
5991704634 Cosmetics: alignment
Originally committed as revision 14635 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-06 02:49:22 +00:00
Vitor Sessak
e3751aa6ec Flip (by making buf[i] -> buf[size-i-1]) two buffers: {sp,gain}_block. This
needs duplicating a few loops, but now the code is clearer.

Originally committed as revision 14634 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-06 02:44:59 +00:00
Vitor Sessak
6888b4fcec Rename prodsum() function to convolve()
Originally committed as revision 14615 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-08-05 04:52:33 +00:00
Peter Ross
fd76c37fd9 Modify all codecs to report their supported input and output sample format(s).
Originally committed as revision 14482 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-31 10:47:31 +00:00
Vitor Sessak
4ca7e74c1f Do not declare a counter as unsigned when it is not needed
Originally committed as revision 14469 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-29 18:22:31 +00:00
Vitor Sessak
4e33ed368a Cosmetics: rename loop counter vars to i,j
Originally committed as revision 14468 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-29 18:20:49 +00:00
Vitor Sessak
045e21cc19 Cosmetics: new line
Originally committed as revision 14457 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 10:49:44 +00:00
Vitor Sessak
e7e4f476c0 Remove the now useless phase variable from context
Originally committed as revision 14456 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 10:49:14 +00:00
Vitor Sessak
c749f3db98 Cosmetics: make comment match var name
Originally committed as revision 14454 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 05:18:39 +00:00
Vitor Sessak
6509507fad Cosmetics: align comments
Originally committed as revision 14453 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 05:16:51 +00:00
Vitor Sessak
f285e6dd31 Rename two context vars: s/sb/sp_block/, s/lhist/gain_block/
Originally committed as revision 14452 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 04:53:05 +00:00
Vitor Sessak
36826ee0df Declare temporary buffers to be only of the necessary size
Originally committed as revision 14451 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 04:37:58 +00:00
Vitor Sessak
5e707c5905 Remove redundant comments
Originally committed as revision 14450 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 04:25:51 +00:00
Vitor Sessak
c6ad13044a Cosmetics: do not align with now removed next instruction
Originally committed as revision 14449 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 04:24:08 +00:00
Vitor Sessak
2c46367163 Cosmetics: remove braces and useless newline
Originally committed as revision 14448 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 04:07:18 +00:00
Vitor Sessak
2b356efec2 Change the way the input is passed to do_hybrid_filter(). Before, in[0] was
the oldest input sample passed and in[n-1] was the latest. Now it is the
contrary. This allows making backward_filter() somewhat simpler.

Originally committed as revision 14447 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 02:55:47 +00:00
Vitor Sessak
aed39f6c02 Remove RA288Context.output buffer. This buffer is just RA288Context.sb
backwards (output[i] == sb[N-i], where N is the buffer length).

This makes the code slower, this will be fixed in my next commit.

Originally committed as revision 14446 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-28 02:53:07 +00:00
Vitor Sessak
b26d320583 Add a few comments pointing to the G.728 specification
Originally committed as revision 14444 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-27 22:22:51 +00:00
Vitor Sessak
854ce441c3 Remove the history buffer from the context. It can easily be evaluated
from the lhist buffer.

Originally committed as revision 14443 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-27 22:21:41 +00:00
Vitor Sessak
0143ec2a7e Rename two context buffers: pr{1,2} -> {sp,gain}_lpc. Also add a few doxy
comments.

Originally committed as revision 14437 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-27 20:58:29 +00:00
Vitor Sessak
59da0390dc Remove st1 and st2 temporary buffers in backward_filter() and use instead
RA288Context.pr{1,2}. Note that the pr{1,2} buffers are one unity smaller
than the st{1,2} buffers. My guess is that the original coder decided to
add one to the array sizes "just to be sure".

Originally committed as revision 14435 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-27 20:55:27 +00:00
Vitor Sessak
3df136d438 Change slightly the meaning of RA288Context.phase. This
allows to remove two intermediary buffers and avoid a few
memcpy's.

Originally committed as revision 14432 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-27 20:35:32 +00:00
Vitor Sessak
f42eb5715d Remove st1 and st2 variables from context, since they could be just local vars
Originally committed as revision 14421 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-26 15:14:50 +00:00
Vitor Sessak
272d258ab9 Rename st{1,2}{a,b} context variables to more meaningful names: {sp,gain}_{rec,hist}
Originally committed as revision 14420 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-26 15:10:17 +00:00
Vitor Sessak
efa86ebe4c Cosmetics: remove useless parenthesis
Originally committed as revision 14418 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-26 14:38:29 +00:00
Vitor Sessak
e07c5ade5e Move colmult() function to the beginning of file to group DSP-related functions.
Originally committed as revision 14417 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-26 14:35:50 +00:00
Vitor Sessak
5af20c207e Remove an useless comment describing colmult()
Originally committed as revision 14416 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-26 14:33:53 +00:00
Vitor Sessak
d899014ee2 Remove an useless comment describing prodsum()
Originally committed as revision 14415 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-26 14:33:25 +00:00
Vitor Sessak
952a980d75 Give the context variables of ra288 a better name: s/glob/ractx/
Originally committed as revision 14414 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-26 14:31:47 +00:00
Vitor Sessak
029e1c01b5 Give the context structure of ra288 a better name: s/Real288_internal/RA288Context/
Originally committed as revision 14413 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-26 14:29:39 +00:00
Vitor Sessak
a53b545068 Cosmetics: grammar nits
Originally committed as revision 14368 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-24 11:55:01 +00:00
Vitor Sessak
2477d60944 Rename update() function to backward_filter() and add a doxy comment
Originally committed as revision 14364 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-24 04:29:23 +00:00
Vitor Sessak
a6755f8d53 Rename and comment bandwidth broadening tables
(s/table1a/syn_bw_tab/, s/table2a/gain_bw_tab/).

Originally committed as revision 14363 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-07-24 04:15:50 +00:00