In one of Chris' patches to loop.c, he noted that subblock_gain is
always 0, and took out some code that uses it. This prompted me to
look up what subblock_gain is. It effectively allows the three blocks
within a shortblock granule to have different global gain settings.
This is a pretty good i
lame3.05pre3 is out. It includes some patches to compile the frame
analyzer under the WIN32. You need to get the GTK/GLIB win32
developer libraries. Albert had included these and a win32/lame/mp3x
binary, but they make the file over 1Mb so I left them out. (and in
the US we cant distribute bi
Here's what the standard says about quantizerStepSize:
global_gain = 8 bit, unsigned int (0..255)
global_gain = quantizerStepSize + 210.
So it looks like quantizerStepSize can range from -210 to 45.
I tested a few samples, and didn't see anything > 0, but we
probably shouldn't assume
Sorry - I wasn't thinking too clearly for my last message about the
fft's. Looks like I forgot that fft is just the interface for fht.
The total fft+fht cost is more like 10%. So maybe there is still
a little room for improvment?
> Func Func+Child Hit
>
On Mon, 24 May 1999, Gabriel Bouvigne wrote:
> quantizerstepsize is double, isn't it? So how will you build a table for
> those powers?
Uhm.. the type is double, but the values are always integer. I have built
a table for quantizerstepsizes between 255 to -255 that seems to work, but
I'm not s
> How can Xing be as fast? It takes 10s when we take 37s. Even without the
> psychoaccoustic model, lame takes about 22s for the same track. I don't
> understand how can it be as fast.
Lots of MMX instructions - at least it has an awful lot in the binary.
Scott Manley (aka Szyzyg)
Hello Gabriel,
U just did a profile on one of my test files (will do profile on
castanets.wav later).
My system is a Pentium 166 Mhz MMX, over clocked at 200 Mhz, using 10 ns
SD-RAM (96 MB)
I've used the Intel Performance compiler on Windows 32 (about 15% faster
than regular MSVC). I still have so
How can Xing be as fast? It takes 10s when we take 37s. Even without the
psychoaccoustic model, lame takes about 22s for the same track. I don't
understand how can it be as fast.
Gabriel Bouvigne - France
[EMAIL PROTECTED]
icq: 12138873
MP3' Tech: www.mp3tech.org
--
MP3 ENCODER mailing list
I remember seeing a 768 points FFT somewhere just before an adaptation of
KLT for mp3 encoding.
The problem is to remember where I saw it.
Regards,
Gabriel Bouvigne - France
[EMAIL PROTECTED]
icq: 12138873
MP3' Tech: www.mp3tech.org
- Message d'origine -
De : Mark Taylor <[EMAIL PROTE
Anyone knows how much time is spent in each function?
Gabriel Bouvigne - France
[EMAIL PROTECTED]
icq: 12138873
MP3' Tech: www.mp3tech.org
--
MP3 ENCODER mailing list
It's interesting to see what can be done using dist10 and a lot of time
Gabriel Bouvigne - France
[EMAIL PROTECTED]
icq: 12138873
MP3' Tech: www.mp3tech.org
- Message d'origine -
De : Mark Taylor <[EMAIL PROTECTED]>
À : <[EMAIL PROTECTED]>
Envoyé : lundi 24 mai 1999 08:01
Objet : [MP3
--
X-Sun-Data-Type: text
X-Sun-Data-Description: text
X-Sun-Data-Name: text
X-Sun-Charset: us-ascii
X-Sun-Content-Lines: 13
>
> I think Chris' FFT is down to about 3% of the total cost. So even if
> we found an FFT that was twice as fast (very unlikely), it would only
> save 1.5%. Thus
Why not removing the -h switch and always using it?
After all, it won't be slower than lame 3.03.
What about using the FFTW package instead of our fft. Perhaps it can be
faster? Perhaps Chris got an indication about this? (if he subscribed to the
list)
And what about the win32 release of the fra
> What about using the FFTW package instead of our fft. Perhaps it can be
> faster? Perhaps Chris got an indication about this?
The fft we have is quite fast and simple. There are faster ones (FFTW is one of
the fastest as far as I know), but I the one I used was one of the fastest and
simplest
I remember seeing somewhere that a big change in the needed bits ammount was
sometimes an indicator for window switching.
Perhaps anyone got an idea to use it?
Gabriel Bouvigne - France
[EMAIL PROTECTED]
icq: 12138873
MP3' Tech: www.mp3tech.org
--
MP3 ENCODER mailing list
How large can quantizerstepsize be at a maximum?
Would it be possible to replace it by a table for all the
pow(2, quantizerstepsize * 0.25) calls?
By doing that it might be possible to change the type to int, that might
speed up a bit.
I tried this but didn't notice much of a speedup, using a tabl
16 matches
Mail list logo
