home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Amiga Plus 1999 #1
/
Amiga Plus 1999 #1.iso
/
System-Boost
/
Sound
/
8hz-mp3
/
8hz-mp3.readme
< prev
next >
Wrap
Text File
|
1998-06-28
|
4KB
|
96 lines
Short: mp3 encoder from 8hz.v0.2b.m8 Cstarware port
Author: 8hz group, mikecheng@cryogen.com
Uploader: mikecheng@cryogen.com
Version: 0.2b.m8
Type:mus/misc
Requires: ixemul.library v47+
Replaces: mus/misc/8hz-mp3*
* Now with a "proper" m68060 version using GeekGadgets egcs-1.0.2 *
Please send me feedback on speed/performance
- does the 060 version cause too many exceptions?
- is the 040 version faster than the 060 version?
- what speed was the 060 encoding at? (ie how many times
faster than realtime?). Please use "timer" from "Executive"
to get proper CPU timings.
- Please please please please send me detailed feedback of 060 versions.
I can't improve the 060 compiles if you just say "it's crap".
- If you *really* want to help, go to the web page
http://www.cryogen.com/mikecheng
and try the "060.lha" archive. This has 4 different versions with slightly
different compile time options. Try them and get back to me.
Encoder based on demonstration C source code released by the ISO/MPEG
Audio Subgroup Software Simulations Group.
A port of the co-ordinated 8hz effort: www.8hz.com
This is pretty much a straight port of the code with a couple of my
modifications from mp3enc included. My modifications have been sent back
to the 8hz group and will hopefully be in the next official source release.
latest Cstarware version/port available from
http://www.cryogen.com/mikecheng
--------------------------------------------------------
CHANGES - as made by Mike Cheng (MFC)
mikecheng@cryogen.com
http://linux.tc3net.com/cstar/
http://www.cryogen.com/mikecheng
(m8) 22nd June 1998
Now encodes from stdin. Thanks to Brad Threatt. Simply use '-'
as the input filename.
fft.c Played with fft.c again. Worked out that from a 1024 point
FFT, we only ever use the first *6* phi values, and the first
465 energy values. Saves heaps of calculations.
QUAL1 Included a quality switch. You can improve encoding speed by
slightly decreasing the quality with the new '-q' switch. I
have terrible ears and can't even hear the difference! :)
Quality range is from 0 to 100, but after 30, you don't get
any speed improvements. Thanks to Jan Peman for this tip.
fft.c (m7) 21st June 1998
Another drop-in replacement for fft.c about *twice* as fast as the
previous fastest fft. Exploits fact that we only require a
real->complex fft, and symmetry conditions on energy and phi
calculations.
Switch MFCFFT turns off the initialization of the wsamp_i arrays
which isn't needed, as the new fft.c just assumes correctly
that wsamp_i is all zeros.
fft.c (m6)
Made a drop in replacement for fft.c 10-15% faster than the old one
afaict
MFCPCM (m5)
had problems with using different pcm samples on different endian
architectures. -> put in a pcm byte-ordering switch.
MFCPCM (m4)
hacked back in raw pcm support (changes to main.c, wave.c and types.h)
Must be a stereo 16bit PCM file.
New command line switches:
-r specify that this is a raw PCM file
-f <Hz> specify PCM file sample rate. (default: 44100)
MFCASMFFT (m3) *AMIGA ONLY*
added in the asm FFT from mp3enc for m68k
based on sources from Henryk Richter and Stephane Tavenard
PSYD (m2)
cleaned up a few bits in l3psy.c
replaced a sparse matrix multiply with a hand configured unrolling
WINDST2 (m2)
minor, minor cleanup in window_subband
FILTST (m2)
filter_subband in l3subband.c
speed ups borrowed from Stephane Tavenard
LOOPD (m1)
sped up the quantize() function
MDCTD & MDCTD2 (m1)
sped up l3mdct.c