This is the mail archive of the
mailing list for the GCC project.
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Steven Bosscher <s dot bosscher at student dot tudelft dot nl>
- Cc: gcc at gcc dot gnu dot org, GNU Fortran 95 development-mailing-list <g95-develop at lists dot sourceforge dot net>
- Date: Thu, 3 Jan 2002 21:15:36 -0500
- Subject: Re: SIMPLE
- Organization: Red Hat Canada
- References: <1010100260.1437.3.camel@steven>
On Fri, 04 Jan 2002, Steven Bosscher wrote:
> Op wo 02-01-2002, om 13:30 schreef Diego Novillo:
> > I'm about to start incorporating these things into the AST
> > branch.
> Does this mean that the decision has been made to go with SIMPLE?
Not necessarily. That's why we'll start working on the AST
branch. The hope is that it will be useful and eventually
incorporated it into mainline.
We need a lower level tree representation with a number of
properties to let us do higher-level transformations and anlyses
like vectorisation, alias analysis, loop optimization, etc.
SIMPLE is a nice starting point because it was defined on top of
GCC trees, so the initial implementation is straightforward. It
doesn't mean that we will not modify it. Particularly since
SIMPLE was designed as a research tool and with C as a front-end
Note that this new IR does not imply the creation of new data
structures. We will keep using trees. The only thing that
changes is that the trees generated by each front end are
re-written to comply with a simplified grammar.
> How does SIMPLE deal with typedefs/structs/classes? Will declarations
> be similair to/the same as those available in the C-family front ends?
> I had some discussion about SIMPLE with Andy Vaught on g95-develop.
Thanks. I'll take a look.
> G95 is getting closer to producing fully parsed, fully resolved g95_code
> trees (G95 ast). The next project is coding a translation pass from
> g95_code to GCC 'tree' structures (including scalarizing). But obviously
> we prefer to know what GCC's new high-level IR will look like before we
> start coding.
That's good to hear. Nothing (or very little) should change from
a data-type point of view. Each front end will need a new pass
that lowers its language-dependent trees into language-independent ones.
> Op do 03-01-2002, om 22:34 schreef Andrew Vaught:
> > Are you sure the GCC people are committed to using SIMPLE? Or rather,
> > does the steering committee have to vote, or does it just take
> > convincing one person of this? My reading of the list looked like
> > people were discussing pros and cons.
> I'd like to an answer to these questions as well, if available, and a
> timeframe, and documentation, and ...
The ast-optimizer-branch will be the initial repository for
implementation and documentation about the li-trees.
I don't know about timeframe. It all depends on how many people
decide to work on it. We have some initial code for the C
front-end. That's what I'm working on now.
> (of course I'm willing to help :-)
Cool! That would be great. Working on other front-ends early
on will let us discover bugs much quicker.