CVS tree

[george a. <ge...@mv...>]

Well I feel that the tick less thing is put to bed.  Unless some one
comes up with a good argument with numbers, that is.  In any case, I am
going about the work of putting the timers together.  

To this end, I have updated the CVS tree to 2.4.7 (each kernel version
is there in case any one cares).  I intend to push my daily work to the
CVS tree so you can all have a look.  I will tag only the stuff that
actually works (or that I think works).

To this end, Robert, did you really test all the different platforms? 
Seems like we need a cpu farm somewhere to test each of these.

A plan:

The issue of how to do this work on one platform and have it extend in
some easy way to the others has caused me to think about just how to
structure the code.  The main issue is that we will need intra tick
interrupts which can only be generated by platform specific code with
platform specific resolutions.  What I am thinking about is a "clock"
registration function.  The function would fill out a table of info on
the various "clocks" available on the system.  The generic code (i.e.
the platform independent code) would fill in the table for 10 ms (or
1/HZ) resolution stuff.  The platform code would register higher
resolution clocks and provide the required function addresses so the
generic code would know what to call for each clock and request
combination.  The interesting thing is that I think this also means that
the platform code could be a module, as long as we provide a unregister
function.

More on this and other things as the work progresses.

George