Subject: Re: Heads up: Thumb code working on NetBSD
To: Christos Zoulas <christos@zoulas.com>
From: Richard Earnshaw <rearnsha@buzzard.freeserve.co.uk>
List: port-hpcarm
Date: 08/23/2004 21:28:43
> | > Have you tried to mount a 2TB ext2 filesystem on Linux?
> | 
> | Nope, are you really suggesting that this is slow because the mount_extfs 
> | (or whatever) program is slow? or because the kernel is slow?  If the 
> | latter, then it's irrelevant to this discussion.
> | 
> Probably because the kernel is slow. What I am saying is that there is
> a limit to the amount of time a user will find tolerable even for one-off
> commands.

Undoubtedly.  However, most (though obviously not all) applications in the 
base distribution are probably more i/o (or kernel) bound than they are 
are compute bound.  In those cases it's probably worth concentrating more 
on reducing the size of the application than increasing its speed.  Even 
the shell probably falls into this category.

And even those applications that are compute bound probably only have 
fairly small parts of the code that really benefit from time-based 
optimizations.

R.