Re: Introspector & Treecc (was: [DotGNU]Licence question about GNU and GCC)

[James Michael DuPont]

> Hmm... how about an abstraction layer that allows us
> to write code to access the different ASTs in
> exactly
> the same way?  
That would be possible after the mapping is done and a
union of the two tree models is in place. 
What interface would it support? That has to be
determined via hard thought work.. 
But it is the ultimate goal I aggree.

> My dream would be that for those
> optimizations which make sense for many programming
> languages, the same optimizer code could be used
> both in gcc and in cscc.
Yes, We have to look into a low level
(IL/RTL)representation and optimisation as well, but
that is outside scope of my experience.

> Such an abstraction layer might avoid any need for
> converting between different AST types in
> performance-
> critical contexts. 
True! I think the tranformation has to take place
outside of the inner loop, it has to be done on the
meta level. 

> We'll still want to have a tool for
> testing purposes that can do such conversions, but
> it may be ok to implement that by converting the 
> native  XML representation of the original AST to 
> the native XML representation of the target AST
type.
Somthing on that level, a protyping environment.
Currently I do my protyping in a mixture of perl xml
and database. 

Theoreticaly, it should be possible to optimise the
transformations down into high-speed code, even
generating gcc compatible ast tranfomation code that
talks to the tree directly to implement the
transformations.

I have to think about this some more.

mike

=====
James Michael DuPont

__________________________________________________
Do You Yahoo!?
Try FREE Yahoo! Mail - the world's greatest free email!
http://mail.yahoo.com/