Re: [Etherboot-developers] Summer of Code project (i386-16 backend for gcc)

[Daniel V. <dan...@gm...>]

H. Peter Anvin wrote:
> That would totally not match the goals of keeping the code maintainable,
> and the utility of a 16-bit C compiler is wider than that.
>   
Ok, I can understand that this is an important consideration.  I realize that the translated assembly language is possibly less maintainable than the existing C code, but I still think assembly is not inherently unmaintainable, just more prone to be that way.


> There are two different things that could be accomplished in shorter time:
>
> - First of all, contribute proper support to gcc for emitting .code16gcc
> without any additional hacks.  (Right now, you either have to edit the
> outgoing assembly code to insert .code16gcc, or you do
> asm(".code16gcc"); plus disable any options that can cause code motion,
> which is undesirable.
>
> - Get a stable Unix-hosted version of OpenWatcom.
>   
I realize after thinking about it some more that there is a third
option: creating a totally separate backend from the i386 one.  This
should help quite a bit to reduce the complexity of the code I have to
touch and avoid mucking around in the main i386 code.  I see now that
there has been some effort in this direction already (
http://www.delorie.com/djgpp/16bit/gcc/ ), so I will try this route
next.  It seems quite a bit more doable in a short time period and helps
reduce the scope of the problem immensely.
> I think abandoning this project would be a major loss.  However, I have
> heard on #gcc that there might be someone else out there trying to do
> the same thing, in which case you might want to consider joining forces
> instead of doing it independently.
>   
I will check into this once I look into the separate backend idea -
maybe the other person is also doing it this way, since it seems much
more sane.

Thanks for the insight,
-- Daniel Verkamp <dan...@gm...> / <dve...@pu...>