[patch] Decompose gimplify_expr

[Diego Novillo]

On Fri, Jul 29, 2005 at 12:33:52PM -0700, Mark Mitchell wrote:

> I think this is indeed something that should wait for 4.2.
> 
Yes.

> I'd also like to hear from our GIMPLE experts.  I'm betting that you've 
> already been discussing your plans for gimplification changes with the 
> right people, so I'm not particularly concerned, but it would be good to 
> make sure that we're not trying to move in different directions at the 
> same time.
> 
I'm a bit concerned about some of the proposals in Zdenek's
paper.  In particular the idea of doing scalar cleanups while
gimplifying.  The advantages over straightforward gimplification
would have to be very noticeable.

We already have some modules in the compiler that try to do
diverse things at the same time and they are a maintenance
burden.  The code is typically twisty and full of special case
handling.  I'm a big supporter of small modules with clean
interfaces and few responsibilities.

That said, I am ready to be convinced, if it can be implemented
cleanly and the results are compelling.  The branch will be a
great opportunity for us to experiment with the approach.