Hi Jared,

thanks for your mail :-)

> wouldn't it be better to use
> RTL<http://gcc.gnu.org/onlinedocs/gccint/RTL.html#RTL>,

This is certainly an idea worth to be investigated further.

What I'm doing currently is rather simple and straightforward. It works,
and provides for the most direct hardware access, but at the moment
there is this architecture dependent layer only for x86-64.

I believe implementing that layer for other architectures to be rather
easy. On the other hand, an existing abstraction layer like RTL would
open the door immediately to many other architectures.

Perhaps I should first finish the current design, so see how it all
works out.


> This could make it machine independent. I just found out about it and it
> sounds pretty good, it shouldn't be too hard to change your abstraction
> layer to generate this if you wanted to use it.

Wouldn't we lose some direct control, and thus efficiency?

The reasons why I chose assembly language are (among others) direct
stack manipulations (data type dependent alignments of stack offsets,
pushes and pops determined at runtime), alignment of code fragments to
special tagged adresses, and access to the condition code register
flags.

It would be interesting to find out how to do that in RTL.


> I'm just wondering, in your assembly are you using system calls or stdlib?

Just calling the C libraries. The calling conventions for C functions
are part of the machine dependent layer.


> and how far into the next version you are?

I have around a quarter by now. It all takes longer than expected ;-)

My plan was not to publish a first version before somewhere near the end
of this year, to avoid too much confusion. Many critical parts are still
changing.

But if you are interested, I could send you what I have.


> syntax file that I'll post when it's done, I might also make some plugins
> for it.

Great!

Cheers,
- Alex
-- 
UNSUBSCRIBE: mailto:picol...@software-lab.de?subject=unsubscribe

Reply via email to