[haiku-gsoc] Re: First timer.diff (from private e-mails)

My main concern with implementing something like gX86SwapFPUFunc() is
that there are multiple timer methods that need to be switched out,
which is why I opted instead for a structure; i instinctively chose
the module structure because it provided what I needed.

I'm not sure how exactly to go about this then. Should I just use a
standard structure full of function pointers, with a timer name, or
should I really set each function individually.

Everything else (coding style, copyrights, paths, etc.) has been resolved.

I would not be opposed to somehow moving ACPI into the kernel to help
with HPET (need to talk to Korli, i imagine).. otherwise, I'm not
quite sure how to implement the timers as true modules. Moving the
module init code earlier in the kernel (before timers) may not be the
best way to go (speaking from an uneducated-on-this-topic standpoint,
of course) only because scheduling is needed for modules (I believe
Stefano mentioned) and it would be a good deal of work.

Currently trying to find a good way to do it and overcome the booting
issues having them as "modules" brought.

On Tue, Jun 17, 2008 at 11:40 AM, Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> wrote:
> Hi Dustin,
>
> "Dustin Howett" <alaricx@xxxxxxxxx> wrote:
>> * Will adapt to the coding style
>> * As for relative_timeout, i was reusing the variable names from
>> arch_timer.c
>
> I guessed so, but I wanted to mentioning it anyway, as there were other
> example of new code, too.
>
>> * I was intending to copy over the copyrights
>> * Since timer_module_info is missing, i note that you have an older
>> version of the patch.
>
> I only got what Stefano sent me.
> BTW would you mind to move any further discussion to the gsoc mailing
> list? Coding style problems and the like are probably are problem of
> everyone :-)
>
>> * I had something in mind that would make module.cpp build; each
>> arch's arch_timer would have a list of the timers it supports.
>
> That would work. Anyway, you made a specific comment about another
> architecture that would need such a timer architecture - which one were
> you referring to?
>
>> * Will look into making them non-modular. I realize this is a better
>> decision and will rectify the rest of the things you've listed.
>
> At least it's a better decision for now. If not possible otherwise, I
> also wouldn't mind of moving the ACPI module (or some core
> functionality of it) into the kernel.
>
>> I also need to fix the booting issues they cause. I sent the patch
>> only for review anyway- actually, for exactly this kind of thing.
>> Thank you very much for explaining what was wrong :)
>
> Glad to help :-)
> If you have any questions, don't hesitate, I'm just away over the
> weekend, but I'll be back on Monday; I don't remember exactly when
> Stefano will return.
>
> Bye,
>   Axel.
>
>



-- 

- DH

Other related posts: