Re: SweetAda 0.1 released

[gabriele.galeotti.xyz@gmail.com]

On Wednesday, July 1, 2020 at 6:57:33 PM UTC+2, Stéphane Rivière wrote:
> Hi Gabriele,
> 
> > I feel very sorry, unfortunately the AVR is the ONLY cpu I can't provide even elementary support, because I am still waiting an hardware board.
> 
> No problem. May be it's an opportunity to help each other...
> 
> Any small arduino board could fit at nearly no expanse.
> 
> To give you a perspective of the various existing development
> environments for AVR and Ada :
> https://stef.genesix.org/pub/ada/embedded_AVR-Ada_%20Setup_v29.pdf
> 
> > I promise you effort on that, maybe not just tomorrow morning. I just had no time to build a test emulator environment. Please give me a little bit of time. The AVR toolchain exists because, once I automated the build, I got it nearly for free, and it is an interesting CPU.
> 
> I understand, I have to reactivate mine too :)
> 
> > Being said that, runsweetada is useful only for the QEMU emulator, so no worries if you plan to work on bare hardware. 
> 
> That's the case... Bare Metal only...
> 
> There is also AVR emulators under Linux like
> https://github.com/buserror/simavr for particular needs.
> 
> I know that the manual is missing important pieces, ask me without fear.
> Please explain me your context in detail, so I can help you.
> 
> Many thanks !
> 
> > And thanks for your interest.
> 
> You're welcome ! Your project is a motivating one.
> 
> -- 
> Be Seeing You
> Number Six

Hi Stéphane.

Have a look at the site. Download sweetada-0.1b.tar.gz and re-download sweetada-rts.tar.gz.

The core package has now an "AVR-simavr" platform. There is nothing, but at least you could verify if it generates code. The RTS has now a complete runtime for AVR, something functional.

Some notably things:

- the GCC front-end does not generate atomic instructions, I think this is due to actual implementation or maybe the lack of that instructions in AVR chips. Chances are the even a C compiler will refuse a similar construct.

- nearly the same reason, in RTS, no 64-bit floats, but this a rather impractical thing to worry about now, since there is no even basic support for other most useful things

- you have to comment out in core/core.ads the "Atomic" aspect in the declaration of Tick_Count, simply put a double dash in front of it. It is obviously not used but ADAC will just complains

Have a look in platforms/AVR-simavr/configuration.in and change whatever you want to suit your needs. I had no time to make a full simavr test, I've only check a GDB connection and it shows something interesting. Check the declaration of the ports. Maybe they need adjustment. Just for the record, I am going to buy an AVR board very soon.

Let me know.