[Flac-dev] question regarding compression %

[gidluck at alltel.net]

Hi Josh,
I made some choices for the encoder based upon how quickly the pocket pc could convert files. I think that I chose flac -0 because of this and also I think that I may have increased the buffer already to 4608. 

However, I will do some additional testing.

The cpu's used are in the ARM family. Xscale (PXA255) as found on the hp 5100/5500 series and 2200 series. 

The best performer right now is the Dell X50v 624mHz(PXA270), again same processor family.

Because this is programmed I/O, media writes contribute largely to cpu utilization. FLAC helps, but it is a trade-off between lowering the amount of data written and the increased cpu requirements for conversion.

Gordon

> 
> From: Josh Coalson <xflac at yahoo.com>
> Date: 2005/09/27 Tue AM 11:09:22 CDT
> To: flac-dev at xiph.org
> Subject: Re: [Flac-dev] question regarding compression %
> 
> --- Gordon Gidluck <gidluck at alltel.net> wrote:
> > Hi,
> > For those of you who don't know, I now have an application on pocket
> > pc using flac. Live2496 does realtime encoding while recording.
> > Live2496 typically is used to record a digital source using the Core
> > Sound PDAudio device. Links are now on the main flac page if you are
> > interested in learning more. 
> >  
> > I have a question. We have run across a 24-bit file that when we
> > record it, flac level zero compression only results in about a 2%
> > reduction in the file size. I use flac level zero because we don't
> > have the computing power required for (higher level flac) realtime
> > encoding on pocket pc's and windows CE. 
> > 
> > What is it about a file that would cause such a poor encoding rate?
> 
> hard to tell without a sample.  or you could decode and reencode
> the sample on another machine with -1 .. -8 switches and post
> the compresses sizes, that could also give a clue.
> 
> if there's no way to do -1 on the pocketpc (which is equivalent
> to -0 -M and will help with stereo), there are a couple of
> tweaks you can try that shouldn't affect the encoding time but
> might help for hi-res recording.  you could try increasing the
> blocksize with -b.  -0 defaults to -b 1152, try 2304 and 4608.
> 
> also -0 defaults to -r 2,2; you could try other values like
> 1,1 3,3 4,4 ...  increasing -r tends to work better with larger
> blocksizes.
> 
> what kind of CPU is in it?
> 
> Josh
> 
> 
> 
> 		
> __________________________________ 
> Yahoo! Mail - PC Magazine Editors' Choice 2005 
> http://mail.yahoo.com
> _______________________________________________
> Flac-dev mailing list
> Flac-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/flac-dev
>