This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
Re: [tree-ssa] Gimplifying Java
- From: Jeff Sturm <jsturm at one-point dot com>
- To: Ranjit Mathew <rmathew at hotmail dot com>
- Cc: java at gcc dot gnu dot org, <gcc at gcc dot gnu dot org>
- Date: Tue, 17 Jun 2003 09:06:22 -0400 (EDT)
- Subject: Re: [tree-ssa] Gimplifying Java
On Tue, 17 Jun 2003, Ranjit Mathew wrote:
> A few dumb questions from a person quite new to tree-ssa:
All good questions, really...
> a. The subject confuses me - should it not be
> "Genericizing Java"? I mean, aren't front-ends supposed
> to generate GENERIC trees that a common "gimplifier"
> will convert to GIMPLE?
Like Andrew says, the Java trees are already a superset of GENERIC. The
frontend never really tries to represent the whole function in canonical
GENERIC form, because nothing useful is done on GENERIC at this time other
than immediately lowering to GIMPLE. That may change later.
Currently all frontends define LANG_HOOKS_GIMPLIFY_EXPR to perform the
conversion of language-specific tree nodes to GENERIC. In spite of the
name, this function isn't required to return a GIMPLE tree.
> c. Are we doing away with the current tree + RTL dichotomy
> of Intermediate Representations and going with a
> single IR after this effort is complete?
GIMPLE, the inliner and SSA optimizers all currently perform
machine-independent transformations. I'd expect that register
allocation, instruction scheduling and peephole optimizations (among
other things) will operate on RTL for a long time to come.
Jeff