On Dec 11, 2008, at 12:49 AM, Patrick Walton wrote:
> Hi,
>
> I've been working on TableML some more and I thought I would see about
> how I should work on getting it into the tree, since it's nearing a
> usable state.
Hi Patrick,
Sorry for not responding earlier. I have a couple of high level
questions about this work. First, is the goal to just generalize and
strengthen the language that is input to tblgen? If so, what
application are you primarily interested in (for example, the code
generator, llvmc, etc)?
> So far I have a (mostly) working interpreter for a subset of Standard
> ML. After interpreting a set of definitions, the interpreter passes
> the
> resulting values off to a backend for serialization. The idea, for
> those
> who didn't see my previous email, is to replace TableGen for use cases
> that require more expressiveness. In particular, I was envisioning
> using
> it in llvmc2 to express compiler driver logic.
I'd really rather not make a new front-end for tblgen. IF you're
interested in making tblgen more expressive (which is great!) I'd
really rather either a) improve tblgen, or b) replace it. Adding
another system that is only used in some cases makes it more difficult
to get to know LLVM because you have to learn multiple different
systems.
-Chris