RE: Simulator for GCC Testing [was: RE: [Fwd: Re: [avr-gcc-list]GCC-AVR Register optimisations]]

[Paulo Marques]

Quoting "Weddington, Eric" <address@hidden>:
> > -----Original Message-----
> > [...]
> > I've looked at the code for both simulavr and simulavrxx. It seems to
> > me that these are more geared towards people trying to debug problems
> > with their own projects, and not so much automate compiler
> > tests. (more
> > like AVR studio, too)
>
> The goal is for both. Simulavr is already being used for testing the
> compiler and for avr-libc.

Yes, but simulavr isn't being maintained any more, right?

Even more, one of my points is that having a software to handle both 
cases might be harder to maintain than a simple simulator and 
full-featured hardware emulator as separate projects. This is not my 
strongest point, though.

> > Most of the code there is to handle all sorts of peripherals that can
> > be found on avr microcontrollers, as is to be expected from full
> > emulators. However, my idea is much simpler: it is probably just the
> > size of "decoder.cpp" of simulavrxx, re-written in plain C.
> > This should
> > make it really easy to port it to any platform (cygwin, etc.).
>
> So now we're talking about simulavr again. Simulavr is written in C and
> can at least be built for Cygwin. But it's unmaintained. The goal is to
> get simulavrxx working for Cygwin AND for running the GCC test suite.

No, simulavrxx has a _lot_ of code to handle peripherals. In fact it is 
the majority of the code of simulavrxx. The CPU part is "just" 4897 
lines out of a total of 20586.

> > [...]
> And this will further split the community.

You make it sound like a bad thing.

> Why can't we work together, instead of always separately?

Because it's not the way open source works. (or at least not the way it 
works better)

Open source works the other way: projects blossom or die by natural 
selection, with the advantage that we can pick the best parts of each 
project and mix and match as we please (doing a bit of artificial 
selection in the process).

So, I don't like the way simulavrxx works. The pre-decoding of flash 
into instances of "opcode classes" doesn't seem like a good idea to me, 
and it is a fundamental concept of simulavrxx. That _is_ my personal 
opinion and everyone is entitled to his own.

Now, I don't mind at all discussing technical merits of the idea, 
especially if I can show my own code to use as a counter argument. So, 
I was trying to delay my replies (including the reply to Joerg Wunsch) 
to a point where I could show some code instead of the natural 
handwaving that these kinds of discussions inevitably degenerate into.

Attached is a beta version of avrtest. It also has a small "Hello, 
World!" demo that actually runs under avrtest and produces the expected 
output.

It is a single file of C code, 1391 lines long. It must still have some 
bugs in there, but I was actually quite surprised when after writing 
the whole thing it ran "Hello, World!" on the first attempt.

I used the "lookup_opcode" function from simulavrxx to save some time, 
but if the author has a problem with me using it (although I'm legally 
entitled to use it), I'll write my own too, because I believe that 
respecting the author's whishes is more important than respecting the 
actual license.

I'll be polishing it up a bit over the next few days because it still 
lacks a few things:
- a few opcodes are still not implemented at all
- RAMPx registers still need to be handled in a few places
- a few hidden bugs (hopefully) that need to be chased down and shot

So, please, instead of just dismissing this project as a "vanity 
hacker's project" or "NIH sindrome", just take a look at it. If you 
still don't like it, that's fine. But at least now we can talk more 
technical and less handwaving.

At least, I bet you can easily compile it under cygwin ;)

--
Paulo Marques


----------------------------------------------------------------
This message was sent using IMP, the Internet Messaging Program.

avrtest.tar.gz
Description: GNU Zip compressed data