Re: Hidden differences between hp620/660 and hp680/690

[Kristoffer E. <kri...@gm...>]

The things is that I've never heard anyone state any difference (aside from pure cpu mhz) between the hp620/660 and hp680/690 platform. Even NetBSD seems to treat them as same platform. So why are we having these issues, seems quite strange.

I might rip a 620 in half just to see what kind of hardware actually is in there.

I suspect that either people just assumed the hardware was same or that they never actually tried it. I got pictures of hp680 ripped apart, but that wont help much since that is working well.

Well, I'll think of it some more. Will see if I cannot get hold of a 620 and rip it in half.

/Kristoffer


On Fri, 3 Aug 2007 01:30:39 +0900
"Magnus Damm" <mag...@gm...> wrote:

> Hi Kristoffer,
> 
> On 8/2/07, Kristoffer Ericson <kri...@gm...> wrote:
> > Greetings,
> >
> > Wanted some expert input on this. We are having situations where things are working perfectly fine on hp680/690 while working badly on 620/660lx.
> >
> > Examples:
> > New alsa sounddriver works perfectly fine on hp680/hp690 but not at all on 620/660
> > Pata platform driver works perfectly fine on hp680/hp690 while not at all on 620/660, we have solved this by doing polling instead.
> > HP680/HP690 reports 7729 CPU, while 620/660 reports 7708. ?
> >
> > Any ideas anyone? Assume 7709 vs 7708 would this explain the issues?
> 
> Not sure, but 7708 is missing a few things compared to the rest of the
> 770x. For interrupt related stuff please have a look at the newly
> added setup-sh770x.c file. I guess the most important part of it from
> my interrupt point of view is that 7708 doesn't support pint. And I
> remember something about CF on pint, but I'm not sure it that's true
> for all models. So there you have something related to your
> pata_platform polling. The DMA controller is also missing from 7708,
> maybe that one is used by the sound driver?
> 
> It would be useful with more hardware information. Are there any
> documents out there where people have marked out which pins that go
> where? If you can find that then it should be pretty easy hack up
> kernel support for all models. sh3-based models at least that is. =)
> 
> / magnus


-- 
Kristoffer Ericson <Kri...@Gm...>