[FFmpeg-devel] [PATCH] G.729A (now fixed-point) decoder
Sat Mar 15 19:24:58 CET 2008
I'm back again and glad to show you reworked G.729A decoder.
Now all code uses fixed-point arithmetic.
Here is results of comparing with reference fixed-point code results:
ALGTHM.BIT stddev: 44.15 PSNR:63.42 bytes:4096
ERASURE.BIT stddev: 12.79 PSNR:74.18 bytes:47104
FIXED.BIT stddev: 3.82 PSNR:84.67 bytes:18432
LSP.BIT stddev: 11.96 PSNR:74.76 bytes:356352
OVERFLOW.BIT stddev:681.71 PSNR:39.65 bytes:61440
PARITY.BIT stddev: 5.57 PSNR:81.40 bytes:47104
PITCH.BIT stddev: 16.43 PSNR:72.01 bytes:292864
SPEECH.BIT stddev: 9.63 PSNR:76.64 bytes:598016
TAME.BIT stddev: 24.67 PSNR:68.47 bytes:20480
TEST.BIT stddev: 15.82 PSNR:72.33 bytes:26624
What was also done:
0. variables/parameters renaming is in progress as long as adding spaces.
1. Input buffers (as much as possible) were marked as "const"
2. Pitch delay code decoding code inlined into main loop. Several
Code becomes more readable. imho.
3. decode_frame now works only with one packet.
4. Doxygen comments were fixed in many places.
7. duplicated code in decode_frame routine replaced with a loop
8. a lot of errors/typos fixed in code
Hopefully this version will be better.
Vladimir Voroshilov mailto:voroshil at gmail.com
JID: voroshil at gmail.com, voroshil at jabber.ru
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
More information about the ffmpeg-devel