On Saturday 21 April 2001 10:10 am, IWAMOTO Toshihiro wrote:
> At Fri, 20 Apr 2001 21:55:38 +0100,
>
> Chris Gilbert wrote:
> > On Friday 20 April 2001  2:56 pm, Ben Harris wrote:
> > > In article <01042010362000.00321@pinky.paradox.demon.co.uk> you write:
> > > >I've made the patch to improve the delay handling more generic.  What
> > > > I've done is move the delay routine (and related check and
> > > > calibration routines) into arch/arm/mainbus/cpu.c.  I've then removed
> > > > all the delay routines from the different clock files. (iomd,
> > > > footbridge, ofw and isa) to use the common delay routine that calls
> > > > microtime.
> > > >
> > > >However not having all the bits of hardware to check this works I'd be
> > > >grateful if people can check that the code still works as expected on
> > > > all platforms.
> > >
> > > You'll need some way to exclude it since (e.g.) hpcarm has its own
> > > implementation of delay().  Either that or rip out the hpcarm
> > > implementation as well.
> >
> > It's probably best to take it out of hpcarm as well then, as it is common
> > code.  common enough to possibly be shared by arm26?
>
> delay() should use hardware timer if possible, IMO.  This is more
> accurate.
> How about putting your delay code in, say, delay_generic.c and linking
> it conditionally using a config option?

Do you mean it should make use of it's own timer?  the delay code in the 
patch uses microtime to track the clock timer.  The calibrated loop is only 
used for small quantities of time.

Chris