cos changes

cos changes

[Doug Evans]

Hi.

I checked in a few changes to cos (cgen's object system).
Virtual methods (as implemented) are gone, as is multiple inheritance.
Plus a few other simplifications.

In trying to have something that is portable I now think the thing to do
is keep cos, albeit greatly simplified.
There's a few more cleanups I want to do, plus more portability work,
but hopefully that's it for now.

Re: cos changes

[Doug Evans]

Doug Evans wrote:
> Hi.
>
> I checked in a few changes to cos (cgen's object system).
> Virtual methods (as implemented) are gone, as is multiple inheritance.
> Plus a few other simplifications.
>
> In trying to have something that is portable I now think the thing to do
> is keep cos, albeit greatly simplified.
> There's a few more cleanups I want to do, plus more portability work,
> but hopefully that's it for now.
>   

One data point I found interesting.

Generating m32c opcodes files before /after (on my machine) went from ~8 
min -> ~5 min.

[For completeness sake, it's still too slow, and there are other better 
ways to improve this.  And there were still untapped ways at improving 
the existing cos implementation.  But I thought the data was interesting 
so I'm passing it on.]