Re: Continued existance of armeb?
Wouter Verhelst wrote:
> It's obviously true that big endian architectures don't need to
> byteswap, since they already have the data in network byte order in
> their memory. Not having to do mangle your data before throwing it on
> the network or after reading it from a socket clearly implies that you
> have less work to do; and gcc will optimize away any calls to ntohl and
> friends if they don't have to do anything anyway, so it's not just
> something that would theoretically work.
> What the real-life impact of this is, though, I don't know.
Especially if the data is going across the slooooowwwww USB interface to
the disk on an NSLU2.
Now for Lennert with his big hunking iron IOP boxes which route packets
at 10Gb/s, it might be a different story ...