[Top] [All Lists]

Re: XFS architecture ?

To: Ken McDonell <kenmcd@xxxxxxxxxxxxxxxxx>
Subject: Re: XFS architecture ?
From: "Andi Kleen" <ak@xxxxxxx>
Date: Mon, 3 Apr 2000 23:58:42 +0200
Cc: Stefan Smietanowski <stefan@xxxxxxxxx>, Linux XFS mailinglist <linux-xfs@xxxxxxxxxxx>
In-reply-to: <Pine.SGI.4.10.10004040735500.3397-100000@rattle.melbourne.sgi.com>; from kenmcd@melbourne.sgi.com on Tue, Apr 04, 2000 at 07:45:23AM +1000
References: <38E8F90B.44E33FEE@hanse.com> <Pine.SGI.4.10.10004040735500.3397-100000@rattle.melbourne.sgi.com>
Sender: owner-linux-xfs@xxxxxxxxxxx
On Tue, Apr 04, 2000 at 07:45:23AM +1000, Ken McDonell wrote:
> (b) will the Linux version of XFS support both a little endian on-disk
>     format (what you see today with the released code on Intel
>     platforms), _and_ the IRIX MIPS on-disk format

Switching endians is very cheap on modern CPUs (ppro+ have a special
instruction, ultrasparc/ia64/et.al. have special bits in the store
instructions to store both endians). What is expensive is checking
whether you need to switch (or rather, it is not that expensive at runtime,
but bloats the code/binary a lot because a lot of routines will be duplicated)

Big-Endian XFS everywhere probably makes sense.  TCP/IP has proven that it 
is possible :-) 


<Prev in Thread] Current Thread [Next in Thread>