Re: [DotGNU]Reminder - Weekly DotGNU Meeting

[Rhys Weatherley]

Gopal V wrote:

>     * PMCodeGen
>         How's it going ? ... /me wants a peice before
>         I go into my exam season. For example you could handle
>         the class/method/virtual method hassles while the WW
>         take care of the basic operations codegen... (Don't
>         tell me you've finished all that !)

No.  Barely started.  I plan to stub out all of the Parrot code
generation operations with "TODO", get it to the point of
compiling cleanly (even if it won't work), and then the WW can
divide up the work however you see fit.

>     * Meet-a-thon
>         Postponed to next week ...

Regardless of whether there is a meet-a-thon or not this weekend,
I will be unable to attend any session except the first, which
co-incides with the DotGNU regular Saturday meeting.

>     * New Fronts
>         Also stuff like minddog's compiler plugin idea need to explored.
>         Possibly the use of treecc in gcc needs a demo .. could Mdupont
>         port something like the treelang example into treecc based node
>         structure ?. Such a demonstration is going to win more support
>         than just the node structure generation ....

I have some random scribbles on the treecc front also:

    http://www.southern-storm.com.au/download/tc_for_gcc.tgz

Just playing around.  Some changes may be needed to treecc itself.
I envisage a system for gcc with a common set of treecc node types,
that can be extended for specific languages.  Sort of like how
codegen and cscc/csharp interact in pnet.

However, treecc's ability to handle "node libraries" is very shaky
at present.  Hence the "feature" where codegen sometimes rebuilds
even if you don't change anything in it.  Also, the extensions cannot
use %virtual operations at present, which can impact the efficiency
of semantic analysis.  Not sure how to fix these issues just yet.

Cheers,

Rhys.