Re: Amiga delay implementation

["Michael L. Hitch" <mhitch%NetBSD.org@localhost>]

On Fri, 27 Nov 2009, Frank Wille wrote:

> After reviewing the old delay() from 4.0 I have to agree with Ignatios
> that there is not much reason to change the old calibration loop method,
> except Michael has some arguments for it.

  I do?

> BTW, what happens when the old delay-loop is interrupted? It will become
> longer? That would be an argument for Michael's new delay() implementation.

  It's not my implementation - it came with the timecounter diff, and
since I didn't experience any problems with it, I didn't look closely at
it other than noting in passing that the old delay()/DELAY() had been
replaced.

  As to going back to the cpu delay loop, interrupts would change the
amount of the delay, so the delay could be variable.  The current
implementaion should be less affected by interrupts, although could also
delay longer depending upon how long the interrupt processing takes.

  I don't know how accurate the clock frequency on the Draco and don't
know if the currently delay() implementation would be accurate enough
to work on the Draco [once the current delay() implementation is correctly
fixed].

  The other question would be how many Dracos are still running NetBSD?