TMS570 headers ready for review

Pavel Pisa pisa at cmp.felk.cvut.cz
Fri Jun 5 19:01:49 UTC 2015


Hello Gedare,

thanks for response.

On Friday 05 of June 2015 18:18:15 Gedare Bloom wrote:
> On Tue, Jun 2, 2015 at 9:39 AM, Pavel Pisa <pisa at cmp.felk.cvut.cz> wrote:
> > Please, look on the files and express your opinion.
> > The new temporary top level header files is tms5702.h.
> > It should replace tms570.h.
>
> I took a glance at some, I didn't see anything overtly bad, although
> the generated license is (1) interleaved with excess newlines

That is already corrected in the script.

> and (2) possibly spurious to apply on automatically generated content.
> The script he wrote should specify itself what the license is on generated
> files. I'd recommend something that is as permissive as possible when
> dealing with generated files.

The license is not part of the script, it is provided as template
so everybody who contributes can add his/her name there.

I support to provide some license in the headers because
if there is no license then users have no guarantee that
would not be sued. At least in our country code without
writtent premission or copyright can be considered
as fully owned by author. We have switched to two clause BSD,
which is really permissive license after last discussion.
I would even prefer that over RTEMS specific one because it
allows to use header files even in other projects.
I.e. U-boot for TMS570 would be great even that I do not
expect to find resources for that.

Last but not least, Premysl Houdek has spent considerable time
on that work so he deserves some attribution in my opinion.

> I'll be curious to see what tms570 users think of the generated headers.
>
> > As for the header files organization, we propose to
> > move peripherals registers header files to the
> > subdirectory of BSP include path. But because Texas Instruments
> > IP blocks can appear even on slightly modified chips we
> > do not propose tms570 name. Our idea is to use "tmsreg" name
> > and use peripherals headers file names in lowercase form.
> > The files would be generally included from code through
> > "tms570.h".
>
> I'm fine with the more generic use of names, although I might prefer
> separation of tms_reg. We also have some generic BSPs that span
> multiple boards, and I think we prefer something like generic_tms_...
> now, e.g. generic_or1k was recently added. Older generic code used
> something more like gen5200. If the IP blocks may be unrelated to the
> TMS line however, it might be sensible to have something more like
> generic_ti_...
>
> However, until the code is actually shared by multiple BSPs, I might
> prefer to see it kept local to the tms570.

Yes I agree to keep them local for now but moving to include subdirs
ensures that they do not mix with the real programmers written BSP
files. As for the name we could use something like ti_hercules.
In the fact we should use the gen_ti_hercules for BSP if we
expect to extend it in future for RM48 and RM57. Ti tools seems
to put generic stuff of CCS under "Hercules" name so there is some
hope that they keep IPs from name clashing in that family.
We have now RM48 boards because paying industry partner wants
or Matlab/Simulink for both TMS570 and RM48 now. But that project
is based on Ti tools unfortunately.

> > Do you agree with proposal?
> > Have you some more ideas?
> >
> > The adjustment of actual BSP sources for the new header files
> > is minimal. See separate commits. The BSP builds completely
> > with all examples. We plan more testing in next days.
>
> A document of what/how testing was done would be nice, and a
> relatively good (English) report about the whole process would help I
> think, if time and energy permit.

I hope that we get to better documentation but the testing is
critical now. Next task is Ethernet support, it is on the Premek's
thesis check list to be defended. He has spent to much time on headers
so he needs to postpone his thesis till end of year to finish that.
Another important task is to implement correct startup but
it is above his planned involvement.

As for other thesis, Jan Dolezal's i386 VESA work, he defends
his thesis over next week and it is written in English so I post
the link as it is published on the university pages. It can be
copied to RTEMS as well then.

Best wishes,

               Pavel
 



More information about the devel mailing list