This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-ssa] Maintaining/representing def-use/use-def information
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Chris Lattner <sabre at nondot dot org>
- Cc: Andrew Macleod <amacleod at redhat dot com>, Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>, Jeff Law <law at redhat dot com>, gcc mailing list <gcc at gcc dot gnu dot org>
- Date: Mon, 15 Dec 2003 15:26:58 -0500
- Subject: Re: [tree-ssa] Maintaining/representing def-use/use-def information
- Organization: Red Hat Canada
- References: <Pine.LNX.4.44.0312151353030.26453-100000@nondot.org>
On Mon, 2003-12-15 at 15:06, Chris Lattner wrote:
> I understand why you chose to use the existing tree representation as the
> basis of tree-ssa, but you have to admit that it has some serious
> overheads and costs. I am not at all convinced that you can eliminate all
> of them, even if substantial effort is put towards slimming them down.
>
Oh, absolutely. GCC trees are pretty bulky for lots of reasons. That
needs to be addressed, but that problem is completely orthogonal to
tree-ssa.
It just happens that tree-ssa exacerbates the shortcomings of the tree
data structure. Changing that will probably end up being a mini project
of its own. But in the grand scheme of things, it is just a data
structure change for the sake of memory and cache. There is no need to
alter the algorithms nor the basic design for this.
Something similar happened to the way statements are chained in the IL.
What we had before was extremely unfriendly for iterators. We have
something more decent now.
We didn't want to start by ripping out the tree representation for the
obvious time/effort reasons. How soon it gets done will depend on the
usual factors.
Diego.