RTEMS toolchain for Epiphany?
Joel Sherrill
joel.sherrill at oarcorp.com
Thu Nov 13 21:36:32 UTC 2014
On 11/13/2014 3:25 PM, Hesham Moustafa wrote:
>
>
> On Thu Nov 13 2014 at 9:01:34 PM Joel Sherrill
> <joel.sherrill at oarcorp.com <mailto:joel.sherrill at oarcorp.com>> wrote:
>
>
> On 11/13/2014 2:46 PM, Hesham Moustafa wrote:
>> Hi,
>>
>> Thanks a lot for the patches. I will let you know with the status.
>> One thing I want someone to confirm about is that I will not be
>> able to use/copy any of the GPL licensed code there [1], right?
>>
> Not as part of anything that you submit for inclusion in RTEMS.
> I don't know how this particular setup works. I can see host
> utilities that you would just get from there and I don't know that
> we care about those licenses.
>
> There are two libraries, host library (e-hal), that can reside on
> Linux running on ARM/Zynq chip that's part of Parallella board, and is
> used to communicate with Epiphany, initialize it, and load the program
> to it. This one does not relate to RTEMS.
>
Right.
> The other library (which is interesting one) is run-time one called
> e-lib, and it run on Epiphany HW. This one has some functions like
> interrupt operations, mutex operations, DMA handling, core
> identification ,etc.
>
Just poking through it, I don't see anything terribly difficult to
replace. The mutex is just a test and set spin lock, the DMA setup
looks pretty easy, core ID should be easy.
How much code is it? And consider that the GPL boiler plate at the top
of each file is 23 lines in the 44 line file I am looking at.
>
> And there might be libraries you want to use with your experiments
> that is GPL but not submit to RTEMS. That is OK if you manage how
> it comes in contact with your own application code.
>
> Please avoid it wherever possible. If there is some run-time library
> we need to ask for relicensing, then we can go that round.
>> [1] https://github.com/adapteva/epiphany-libs
>>
>> Regards,
>> Hesham
>>
>> On Thu Nov 13 2014 at 6:47:28 PM Joel Sherrill
>> <joel.sherrill at oarcorp.com <mailto:joel.sherrill at oarcorp.com>> wrote:
>>
>> Hi
>>
>> I couldn't get gdb og gcc to build even for epiphany-elf but
>> I did
>> include patch that I think is all that is needed for RTEMS.
>>
>> I did manage to build binutils from their repo for both elf
>> and rtems variants.
>>
>> At least it is a starting point. You can work with their
>> community to get their tools built as epiphany-elf and
>> then just change the target.
>>
>>
>> --joel
>>
>>
>> On 11/13/2014 12:22 PM, Hesham Moustafa wrote:
>>>
>>>
>>> On Thu Nov 13 2014 at 5:58:25 PM Joel Sherrill
>>> <joel.sherrill at oarcorp.com
>>> <mailto:joel.sherrill at oarcorp.com>> wrote:
>>>
>>>
>>>
>>> On November 13, 2014 11:56:32 AM CST, Hesham Moustafa
>>> <heshamelmatary at gmail.com
>>> <mailto:heshamelmatary at gmail.com>> wrote:
>>> >Hi,
>>> >
>>> >
>>> >I want to let you know that I found their main repos
>>> [1] Can I start
>>> >from there? Imitating what has been done with OpenRISC?
>>> >
>>>
>>> Basically. Except you only need their repos for binutils
>>> and gdb. We use GCC and Newlib from upstream.
>>>
>>> Great, I am waiting for these patches.
>>>
>>> I have binutils patches
>>>
>>> >[1] https://github.com/adapteva
>>> >
>>> >
>>> >Regards,
>>> >
>>> >Hesham
>>> >
>>> >On Thu Nov 13 2014 at 3:29:14 PM Hesham Moustafa
>>> ><heshamelmatary at gmail.com
>>> <mailto:heshamelmatary at gmail.com>> wrote:
>>> >
>>> >On Thu Nov 13 2014 at 2:59:33 PM Joel Sherrill
>>> ><joel.sherrill at oarcorp.com
>>> <mailto:joel.sherrill at oarcorp.com>> wrote:
>>> >
>>> >
>>> >On 11/13/2014 8:07 AM, Joel Sherrill wrote:
>>> >>
>>> >> On November 13, 2014 6:30:48 AM CST, Hesham Moustafa
>>> ><heshamelmatary at gmail.com
>>> <mailto:heshamelmatary at gmail.com>> wrote:
>>> >>> Hi all,
>>> >>>
>>> >>>
>>> >>> I want to ask about the status of RTEMS toolchain
>>> for Epiphany
>>> >>> architecture. I think Joel mentioned that there are
>>> some previous
>>> >>> support for it; and if yes, does the toolchain need
>>> some additional
>>> >>> work?
>>> >> To give you a quick answer, I emailed the people who
>>> did the port.
>>> >There apparently is a github repo with some of it and
>>> some is merged. I
>>> >will dig through the emails and post the proper links.
>>> >>
>>> >> One issue they mentioned was that the gdb port had
>>> many core/thread
>>> >support that made it more than a simple port.
>>> >From Jeremy Bennett:
>>> >
>>> >> piphany tool chain development runs on quite a tight
>>> budget, and its
>>> >> GDB implementation is quite complex (it has to
>>> pretend cores are
>>> >> threads, when they don't completely share an address
>>> space). So we
>>> >> haven't had the effort to devote to upstreaming. And
>>> we were
>>> >> reluctant to push the simulator upstream without a
>>> GDB implementation
>>> >> to go with it. You can of course access the code here:
>>> >>
>>> >> https://github.com/adapteva/epiphany-binutils-gdb
>>> >>
>>> >> Epiphany GDB is still in quite substantial flux, due
>>> to the need to
>>> >> support the Eclipse multicore visualizer with
>>> asynchronous and
>>> >> non-stop support.
>>> >The upstream gcc and newlib are OK. But since binutils
>>> and gdb are
>>> >now in a single repo, it will need to come from the
>>> github site until
>>> >it is merged upstream. And obviously patches just need
>>> to go upstream
>>> >to whereever the code is. :)
>>> >
>>> >Jeremy also encouraged you to openly discuss things on
>>> their forums.
>>> >He thought you would get good insight and advice there.
>>> And I don't
>>> >doubt that.
>>> >
>>> >Thank you, I will.
>>> >
>>> >If it is a relatively low volume place, I may track it.
>>> But my email
>>> >volume
>>> >is already high and I don't have time to poke around on
>>> a bulletin
>>> >board.
>>> >
>>> >> It will not have RTEMS as a target but that shouldn't
>>> be hard to
>>> >address once we know where the master binutils, GCC,
>>> Newlib, and gdb
>>> >are.
>>> >So do you want me to try to build a toolchain and get
>>> you some starting
>>> >patches?
>>> >
>>> >Sure that will definitely help as a starting point. And
>>> if you are so
>>> >busy, you can just drop me HOWTO instructions.
>>> >
>>> >> Then you are porting.
>>> >>
>>> >>> Regards,
>>> >>>
>>> >>> Hesham
>>> >> _______________________________________________
>>> >> devel mailing list
>>> >> devel at rtems.org <mailto:devel at rtems.org>
>>> >> http://lists.rtems.org/mailman/listinfo/devel
>>> >
>>> >--
>>> >Joel Sherrill, Ph.D. Director of Research &
>>> Development
>>> >joel.sherrill at OARcorp.com
>>> <mailto:joel.sherrill at OARcorp.com> On-Line
>>> Applications Research
>>> >Ask me about RTEMS: a free RTOS Huntsville AL 35805
>>> >Support Available (256) 722-9985
>>>
>>
>> --
>> Joel Sherrill, Ph.D. Director of Research & Development
>> joel.sherrill at OARcorp.com <mailto:joel.sherrill at OARcorp.com> On-Line Applications Research
>> Ask me about RTEMS: a free RTOS Huntsville AL 35805
>> Support Available (256) 722-9985
>>
>
> --
> Joel Sherrill, Ph.D. Director of Research & Development
> joel.sherrill at OARcorp.com <mailto:joel.sherrill at OARcorp.com> On-Line Applications Research
> Ask me about RTEMS: a free RTOS Huntsville AL 35805
> Support Available (256) 722-9985
>
--
Joel Sherrill, Ph.D. Director of Research & Development
joel.sherrill at OARcorp.com On-Line Applications Research
Ask me about RTEMS: a free RTOS Huntsville AL 35805
Support Available (256) 722-9985
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20141113/5345be35/attachment-0002.html>
More information about the devel
mailing list