qemu-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] implementing ARM926EJ-S support


From: Антон Кочков
Subject: Re: [Qemu-devel] implementing ARM926EJ-S support
Date: Tue, 3 May 2011 02:04:13 +0400

Alessandro, I think you can try add support for one of the Qualcomm
MSM6100, MSM6125, MSM6225, MSM6245, MSM6250, MSM6255A, MSM6260,
MSM6275, MSM6280, MSM6300, MSM6500, MSM6800;
Of course, if you dont know yet which you want.

Best regards,
Anton Kochkov.




On Tue, May 3, 2011 at 01:04, Peter Maydell <address@hidden> wrote:
> On 2 May 2011 13:46, Alessandro <address@hidden> wrote:
>> ARM core-related code appears to be located (mainly)under ARM directory;
>> peripherals-related files should be located under "machine" dir.
>
> I'm not sure what source tree you're looking at. Code for ARM core
> as a target is in target-arm/. Device models are in hw/. (tcg/arm
> is support for emulating other CPUs on an ARM host, so not relevant
> for you.)
>
>> I still have some questions about QEMU ARM926 support:
>>
>> 1- What standard extensions are supported? e.g Jazelle.
>
> We don't support Jazelle. We don't implement the TCMs (in the same
> way we don't implement caches). We probably don't get all the
> device-specific cp15 registers right. (None of that is likely to be
> of any practical importance.) I can't see anything else missing from
> a quick scan.
>
> The rough rule of thumb for support is "if Linux uses it it's almost
> certainly supported; otherwise it might be missing or broken" :-)
>
>> 2- Sometimes, SoC includes "inusual"(and _poorly_ documented) hardware: ISP,
>> video Coproc, and so on.
>> This increase significantly the final complexity of project. What are the
>> "guidelines" to follow? What must be implemented, and what could be safely
>> ignored?
>
> Well, if you're just doing things for your own amusement you can
> implement or leave out what you like. For things to be included
> in QEMU my personal opinion (I don't have a veto or anything) would
> be that the question is whether there's any benefit to the new
> SoC model that the existing ones don't provide -- is it some
> common piece of hardware that a lot of people own and might want
> to use a model of, for example? If you pass that hurdle then there
> is presumably an OS and set of applications that run on the real
> hardware, and the minimum level of peripheral support would be
> "enough to run that".
>
> What SoC are you planning to model? I assume you have one in
> mind since you were specific about wanting the ARM926 rather
> than a more recent ARM core...
>
> -- PMM
>
>



reply via email to

[Prev in Thread] Current Thread [Next in Thread]