[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: JFFS2 endianness.



On Thu, 17 Jan 2002, Tim Riker wrote:

> well thanx for the counter proposal. ;-)
> 
> I'm serious about the BE request though. Any device with a TCP stack
> needs to deal with this quickly. Devices do not NEED to deal with LE
> stuff.
> 
> Though i386 and arm kernels are both LE, I think that filesystems and
> network code should all be BE.

Seriously there were benchmarks performed on Linux with both network and 
filesystem code before, using both endianness.  It was demonstrated that the 
endianness conversion penalty was lost in the noise.  That's what PPC people 
said at the time they worked on support for LE ext2fs.  The installed user 
base is much more of a concern in these cases.

JFFS2 should of course keep its in-memory data in the native endian format, 
but converting it from/to LE while accessing the flash is nothing compared 
to the flash latency.

"Big Endian is much faster for networking" might be true for things like 
silicon routers that already process your address as the nibles are coming 
from the wire, but that claim doesn't stand when multi purpose CPUs are 
involved.


Nicolas


To unsubscribe from this list: send the line "unsubscribe jffs-dev" in
the body of a message to majordomo@xxxxxxx.com