Masao Uebayashi wrote:

> Hi!  We have a similar patch for MPC5200 in a local tree, where
> registers are big-endian.
> 
> > 1) add logic to ohci.c to support run-time conditionalized byte 
> > swapping.  Probably via a function pointer in the softc, this would 
> > work, but might be a little bit slower.  (I'd conditionalize this 
> > support via a new OHCI_NATIVE_ORDER or somesuch.)
> 
> I'm in favor of this approach.

I agree.

The current ochi.o for the PB1000 kernel is 28k of text - you don't
really want to have two copies of that in your kernel.

Perhaps you could instrument the number of calls to where the byte
swapping happens to get an idea of the overhead?  It should be pretty
easy to throw in some event counters that you can then monitor with
"vmstat -e".  We may not then even need an OHCI_NATIVE_ORDER or
similarly named option.

Simon.
--
Simon Burge                            <simonb@wasabisystems.com>
NetBSD Support and Service:         http://www.wasabisystems.com/