[prev in list] [next in list] [prev in thread] [next in thread] 

List:       etux
Subject:    RE: The small matter of JFFS2 compression ratios ...
From:       "Dave Hylands" <dhylands () broadcom ! com>
Date:       2003-09-16 13:06:57
Message-ID: 24CDBA67F085904999751B3C4F9E8C0B5C3BD2 () nt-rmna-0740 ! ca ! broadcom ! com
[Download RAW message or body]

Hi Karim,

Compressed (or otherwise random) data doesn't compress very well.
Perhaps you had a vmlinuz image (instead of vmlinux) on your jffs2
image?

Dave Hylands

P.S. Excellent book :)

> -----Original Message-----
> From: Wells, Charles [mailto:Charles.Wells@nielsenmedia.com] 
> Sent: Tuesday, September 16, 2003 5:18 AM
> To: etux@embeddedtux.org
> Cc: 'Agust Karlsson'; karim@opersys.com
> Subject: RE: The small matter of JFFS2 compression ratios ...
> 
> 
> Karim,
> 
> We do a lot of embedded PPC development around here and 
> always compress binary images with a tool I wrote that uses 
> zlib 1.1.3.  This tool prints the compression ratio before 
> exiting.  Other than a few special cases (images containing 
> large tables, etc.), we always get compression ratios of 
> between 59% and 62% (e.g., 1,000K image compresses to 400K).  
> I've marveled in the past at how close the ratio is for 
> entirely different binary images. Our experience seems to 
> agree with Agust's experience as well.
> 
> I'm not sure what mkfs.jffs2 is doing; but if I got a 
> compression ratio of as little as 12.6%, I'd be suspicious.
> 
> Note: Our images are produced by gcc 2.95.
> 
> Regards,
> Charlie


[prev in list] [next in list] [prev in thread] [next in thread]