Re: [PATCH 3/5] Deforest the function pointer jungle in the time code.

[Sergei Shtylyov <sshtylyov@ru.mvista.com>]

Franck Bui-Huu wrote:

>> On Tue, 19 Jun 2007 09:33:33 +0200, "Franck Bui-Huu" 
>> <vagabon.xyz@gmail.com> wrote:
>> > What do you mean by "pnx8550 can have customized copy of cp0_hpt
>> > routines" ? Do you mean that it should copy the whole clock event
>> > driver ?

>> > It seems to me that using cp0 hpt as a clock event only is a valid 
>> usage...

>> Well, I thought the customized cp0 clockevent codes (custom
>> .set_next_event routine is needed anyway, isn't it?)

> I don't think so.

> hpt-cp0.c clock event part doesn't care if the counter is cleared when
> an interrupt is triggered.

    Well, in the generic case it must read back the Count reg. before writing 
to the Compare reg. and for PNX8550 this is unnecessary -- but indeed, should 
not harm...

> It matters only for clock source (well I
> think) that's why I suggested to rewrite a clock source driver only
> for this platform...

    Yeah, this platform certainly *needs* another clocksource than the counter 
used for the clock events -- currently it's count/compare 2.
    And this platform also *needs* a separate clocksource driver as well since 
the PNX8550 counters *do* support auto-reaload mode here -- in fact, this 
seems to be the only supported mode from the manual excerpt cited here:

http://www.linux-mips.org/archives/linux-mips/2006-12/msg00194.html

WBR, Sergei