bsps/xilinx-zynqmp : Add BSP for RPU

[Chris Johns]

On 17/6/2023 5:14 am, Gedare Bloom wrote:
> On Fri, Jun 16, 2023 at 2:17 AM Philip Kirkpatrick
> <p.kirkpatrick at reflexaerospace.com> wrote:
>> On Fri, Jun 16, 2023 at 7:14 AM Chris Johns <chrisj at rtems.org> wrote:
>>>
>>> On 15/6/2023 6:16 pm, Philip Kirkpatrick wrote:
>>>> Thanks for all the good feedback.
>>>>
>>>> RE Joel:
>>>> I'll fix my sloppy formatting that you caught and submit a revised patch.  If
>>>> I'm realistic about my schedule, I probably won't be able to get to it until
>>>> next week.
>>>> For xttcps_hw.h, there already is one #ifndef __rtems__ around the #includes,
>>>> but on review there is another spot where I got lazy and used a #if 0.  I'll
>>>> correct that too.  Other than that, the file is unmodified.
>>>>
>>>> On the discussion about a shared space, I'll leave that decision up to you.
>>>> Tell me what you want and I can adjust as needed, or it could be done in a
>>>> follow-on patch.
>>>
>>> Should the RPU BSP be located under bsps/arm/xilinx-rpu?
>>
>>
>> I went back and forth on that decision and decided to keep them combined since the APU and RPU share a moderate amount of code.  However, I can definitely see an argument that they are different enough to split.  If you want it the other way, I can make that change when I address the other items.

Thanks, I think this is worth while.

> I think we should split it out. Shared code should likely be
> refactored to arm/shared depending what that is.

Agreed. I am fine with the code moving, leaving the arm/xilinx-zynqmp BSP as is
because it may be removed (see below).

> I'm not sure that carrying forward a 32-bit arm/xilinx-zynqmp makes
> sense now that we have a functional aarch64 port. Splitting the RPU
> out will make it easier to make that decision.

I agree. I think it is confusing. I also think it should happen before 6.

Chris