[PATCH 0/5] arm/tms570: include hardware initialization and selftests to run RTEMS directly from Flash without loader on TMS570LS3137 HDK.

[Gedare Bloom]

On Wed, Sep 14, 2016 at 4:29 AM, Pavel Pisa <pisa at cmp.felk.cvut.cz> wrote:
> Hello Gedare,
>
> On Wednesday 14 of September 2016 02:13:47 Gedare Bloom wrote:
>> On Tue, Sep 13, 2016 at 5:11 PM, Pavel Pisa <pisa at cmp.felk.cvut.cz> wrote:
>> > Hello Gedare,
>> >
>> > thanks for the review. It is huge piece and not in a state
>> > as nice as I would like it.
>> >
>> > On Tuesday 13 of September 2016 21:42:48 Gedare Bloom wrote:
>> >> I'm slowly reading through these. Is there a doc or is it worth it to
>> >> generate one that maps the HalcoGen functions to their RTEMS renamed
>> >> versions? I'm not sure if it is of much value.
>> >
>> > I am not sure at the end. I have tried to change names
>> > to RTEMS, Linux, POSIX style. The rest of BSP (which
>> > is written mostly from scratch) already follows that.
>> > Other option is to use Ti CammelCase and add tms570_ prefix.
>> > Genrally I am not sure. I have not renamed functions
>> > in the ARM core support tms570_sys_core.S .
>> > But if we agree that all should follow single naming
>> > convention I take a while for that.
>> > But there could be value in preserving Ti names either.
>>
>> I wouldn't mind simple tms570_preserveCamelCase() as an easy rule to
>> follow and to track back clearly to what came from halcogen versus own
>> code that should use bsp style (tms570_xxx_yyy_zzz).
>
> The most HalCoGen generated code is editted considerably anyway.
>
> We have tested other proof of concept/option initially.
> This is RTEMS with unmodified HalCoGen files solution
> and it is archived on "tms570-bsp-with-hwinit" branch.
>
> https://github.com/AoLaD/rtems/tree/tms570-bsp-with-hwinit
>
> The code sits in c/src/lib/libbsp/arm/tms570/halcogen directory
> of the branch
>
> https://github.com/AoLaD/rtems/tree/tms570-bsp-with-hwinit/c/src/lib/libbsp/arm/tms570/halcogen
>
> So you can look at that as the reference.
>
> I think that for this case it does not worth directly include
> generated code in RTEMS repo. When regenerated for other board
> or chip variant or by other HalCoGen version it would differ.
>
I see.

> But for quick start on other chip it could be faster solution.
>
It would make sense if it is as simple as dropping in the halcogen
code in a single spot/library. In the future I'm sure we will adapt to
some hybrid that makes it easy to adopt new halcogen without (much)
modification while preserving the code changes and documentation of
the interface between halcogen and RTEMS that you have been improving.

> May it be, we can use tms570_xxx_yyy_zzz and put original
> name after that in Doxygen first line after new name.
>
Yes, this makes sense. It is most important to easily know how to
translate/find the customized version of these functions whether by
'grepping' for the original function name or applying a 1:1
translation of the name should work well.

> Best wishes,
>
>               Pavel